Recording

Transcript

**
00:08**
Daniel Englebretson
Hey, Simon, how’s it going?

**
00:10**
Simon Walmsley
Hi, Daniel. Yeah, I’m good, thanks. How are you?

**
00:14**
Daniel Englebretson
Thanks for jumping in. Hey, Steven.

**
00:16**
Daniel Englebretson
Hey, guys.

**
00:18**
Stephen Hudson1
Hi, Daniel.

**
00:19**
Daniel Englebretson
Hi, Steve.

**
00:20**
Daniel Englebretson
Nice. Hi.

**
00:26**
Daniel Englebretson
Timeout? No, it’s tough to line all the schedules across all the time zones. So thanks.

**
00:32**
Stephen Hudson1
It’s always easy.

**
00:34**
Daniel Englebretson
Yeah, yeah, for sure. Mika, I think you might be driving right now as you said.

**
00:45**
Bhumika Sachdev
Yeah, Yes, I am, but I’ll be in the office in about 15 minutes.

**
00:53**
Daniel Englebretson
Oh, problem. I’ll give everybody a minute to get in here. Just so everybody knows we are recording the session. Yeah, and I do intend to share that back and I’m not planning on sharing anywhere else. So just so that everyone’s aware, check. Does anybody know if someone we’re expecting is not here yet or do we have everybody that we know is coming?

**
01:26**
Simon Walmsley
I think this is it. From the sales side of Hughes, Georgia and Amber work in the same office as me. Oh, James, Engineering manager.

**
01:38**
Stephen Hudson1
Yeah, everyone’s here as far as I’m concerned. Yeah. I’m not sure Matt’s joining us here.

**
01:42**
Daniel Englebretson
I don’t think.

**
01:44**
Simon Walmsley
I think so. Yeah, I know he’s got GMON with him later.

**
01:52**
Daniel Englebretson
Okay, then we will get started.

**
01:55**
Daniel Englebretson
So again, thank you guys for joining.

**
01:57**
Daniel Englebretson
And a couple people joined after I mentioned it. So I just want to let everybody know we are recording.

**
02:01**
Daniel Englebretson
We’ve got a pretty extensive set of.

**
02:05**
Daniel Englebretson
Stuff to go through today, so I don’t think anybody here is surprised about how extensive this process is.

**
02:11**
Daniel Englebretson
But it was mind blowing for me how much stuff you guys have to.

**
02:14**
Daniel Englebretson
Pour through to do what you do. So I’m looking forward to diving into that.

**
02:18**
Daniel Englebretson
But there’s a we do have an agenda that I’ll try to hold us to. Roughly the one thing I want to call out.

**
02:26**
Daniel Englebretson
And Bamika. This wasn’t totally obvious to me until I got far enough into it.

**
02:30**
Daniel Englebretson
There’s so many files involved in a workflow that it is basically.

**
02:36**
Daniel Englebretson
Impossible to do this with ChatGPT directly.

**
02:40**
Daniel Englebretson
So we are going to demo, we are going to do some hands on and we are going to look at it all. But loading up a project in ChatGPT is it super feasible because of how.

**
02:49**
Daniel Englebretson
Many files there are? ChatGPT has hard limits on how many files you can put in. It’s 20 and we, that’s that. That barely scratches the surface of what we need.

**
02:58**
Daniel Englebretson
So you will see what I mean when we do that. And there is opportunity to play along in ChatGPT if you would like.

**
03:06**
Daniel Englebretson
But I think today is going to be more about seeing and reacting and feedback and discussion. It’s really a collaborative discussion around, you know, what. What do we need to be thinking about and changing as we actually make a workflow?

**
03:23**
Daniel Englebretson
So today I’ve got one made that.

**
03:25**
Daniel Englebretson
We’Re going to be using. But the goal is, after this session, we’ll take all the feedback from today and turn it into something that’s usable by you guys too, so that you can play with it and really start working with it.

**
03:39**
Daniel Englebretson
So this is gonna make a lot.

**
03:40**
Daniel Englebretson
More sense when we get to the end of it, but I just want to go ahead and set that expectation now.

**
03:45**
Daniel Englebretson
And so as we get into it, the goal really, there’s really.

**
03:50**
Daniel Englebretson
Two goals that we’re trying to achieve today.

**
03:53**
Daniel Englebretson
Originally, the.

**
03:55**
Daniel Englebretson
When, when.

**
03:56**
Daniel Englebretson
When bamiga came to me and we.

**
03:57**
Daniel Englebretson
Started talking about what we wanted to.

**
03:59**
Daniel Englebretson
Do, that the first goal was how can we all learn together about what.

**
04:04**
Daniel Englebretson
AI augmentation means, you know, and how we might find opportunities to do that inside the company and try to.

**
04:11**
Daniel Englebretson
Just get on the bandwagon of getting.

**
04:13**
Daniel Englebretson
Some value out of AI?

**
04:15**
Daniel Englebretson
And then alongside that came, well, I believe that it’s really hard for me.

**
04:19**
Daniel Englebretson
To just come in and tell you.

**
04:20**
Daniel Englebretson
That instead we need to actually look at a workflow and somebody needs to actually see it in action so you.

**
04:26**
Daniel Englebretson
Can start to, like, get your gears.

**
04:28**
Daniel Englebretson
Turning on, oh, this is what you mean? And so that’s how we fell into.

**
04:31**
Daniel Englebretson
The safety shower pilot and picking off.

**
04:34**
Daniel Englebretson
A workflow that everyone could look at.

**
04:37**
Daniel Englebretson
And think about while we’re talking about these things. So there’s the kind of business outcomes of how do we get the value out of this workflow for the safety shower example. But then there’s the broader objective of.

**
04:49**
Daniel Englebretson
Like, how do we start thinking about.

**
04:51**
Daniel Englebretson
What we should expect AI to be able to do and how do we start thinking about how that value starts to compound over time and the systems and processes that we need to do it.

**
05:01**
Daniel Englebretson
Those are the two things that I’m trying to achieve today. And after today, it’s going to take me a few days to come back.

**
05:09**
Daniel Englebretson
To you with a revised workflow that you guys can play with.

**
05:14**
Daniel Englebretson
The intent is that you guys will.

**
05:16**
Daniel Englebretson
Have a chance to play around with the workflow, give some feedback on it before we seal the deal on documenting everything and calling the mvp, quote, unquote, done as an MVP goes. So.

**
05:28**
Daniel Englebretson
So that’s just kind of tee us up. Does anybody have any questions about what.

**
05:33**
Daniel Englebretson
We’Re getting into today or the goals of today before I do want to do a quick round of introductions before getting to it. But before we do that, does anybody want to set me straight on anything or add anything to the conversation before we get into it? Right, Great.

**
05:52**
Daniel Englebretson
So I would like to do a quick round of introductions and I know I’ve met almost everybody. There’s a few people I haven’t met yet. But there’s a couple of reasons why.

**
06:04**
Daniel Englebretson
I want to do this.

**
06:05**
Daniel Englebretson
First of all, I would like to.

**
06:07**
Daniel Englebretson
Know your name and your role.

**
06:08**
Daniel Englebretson
But I also would like to understand, just like your quick introduction of what is it about this workflow, the safety shower RFQ process, that you think there’s opportunity to refine and that you think.

**
06:24**
Daniel Englebretson
There’S opportunity to maybe change how you bring value to the customer or how you bring value to the organization and how you do it with AI all answers are valid. I’m just trying to get a grounding around that.

**
06:40**
Daniel Englebretson
I’m going to go last. So if you can give me your name, your role and just a quick what do you think? Where do you think the values at.

**
06:46**
Daniel Englebretson
In this workflow overall? That would be super helpful and then we’ll get on with it. So I’m just looking at my list of names.

**
06:54**
Daniel Englebretson
So I’m going to start with Steven.

**
06:55**
Daniel Englebretson
And then Stephen, you pick somebody when you’re done.

**
06:57**
Stephen Hudson1
I will, thank you. Yeah, I’m Steve.

**
07:00**
Daniel Englebretson
Hi.

**
07:00**
Stephen Hudson1
Daniel. Good to see you again. So I think probably going to get a better answer out of the experts who are on this call. But I feel, you know, when me and Simon first got to talk about this issue a little while ago and when I quickly, oh, by the way, I’m a VP of operation, so I’m not, I don’t work in a sales department. We kind of see the outcome of these activities. And the reason being Simon got talking alongside James was that one occasion we missed something one of the technical reviews. And it was something which is hidden deep within these documentation trails. It was really deep. And it took us ages to kind of find out where this came from. And ultimately we had back and forth to the customers and etc.

**
07:52**
Stephen Hudson1
But we had to kind of hold our hands up, said, yep, we missed it unfortunately. And when Simon kind of walked me through it, I could see why. I mean, it was really some obscure table in some appendix somewhere related to something else. And that gave me a real appreciation of the kind of work that Simon has to kind of work to. But all in that James, who’s on here, he’ll introduce himself, you Know, it’s very, as you know, there’s a lot going on there. So without hogging all the time, I think the, you know, the outcome here is to kind of make Simon’s life a little bit easier. That makes James’s life a bit easier. That makes my life a bit easier right across the chain.

**
08:33**
Stephen Hudson1
But also I think more importantly is giving ourselves a scope to do a lot more with a lot more customers, which will improve our chances of increasing our revenue and growth opportunities. So very significant impact on the business and I think a very significant impact on the people on this call in terms of their daily work.

**
08:54**
Daniel Englebretson
So.

**
08:54**
Stephen Hudson1
So with that, I think, again, a better answer for you, Simon. I’m going to hand the bat on to you and yeah, thank you.

**
09:04**
Simon Walmsley
Okay, thanks, Steve. Hi, Daniel. Yes. So I work in the sales team along with Georgia and. Sorry, Georgia and Amber, who are also on this call. I’m a proposals engineer and sort of main point of contact with the customer during the bid stage. The big problem and what we’re looking for some help really is there’s a lot of heavy lifting involved when we receive this RFQ pack. I think you use the term document triage, which is a good term. You’ve got to very quickly eliminate a focus on what’s, you know, the main important part of the rfq. And as you’ve seen, we. We can literally receive thousands of pages.

**
09:50**
Simon Walmsley
I believe the contractors we’re working with, I’ve used the term EPC contractors, the engineering procurement consultants, the big blue chip companies, and they will send a documentation pack to suppliers of people producing turbines and compressors and lots of complicated process equipment. But we will also receive the same level of documentation. Even though safety showers, there is a degree of complexity to them, but nowhere near as much as some of the other equipment on this site. Nonetheless, we have to comply to everything in the rfq. Once we’ve managed to find out what exactly they’re looking for and we provide a quotation, we’ll go through a technical bid evaluation stage afterwards where we will be asked to provide our full compliance to everything we’ve received.

**
10:45**
Simon Walmsley
And as Steve mentioned before, if there’s something buried in those documents that has been overlooked, you know, there’s a risk to the business that, you know, it’s not been missed and captured in our quote. And by default we’re expected to comply. The other issue we have, obviously the risk of missing things is also the sheer volume of work means that it can take us a long time to respond to RFQs. So a great thing that could come out of this is our speed of reaction and to hopefully pleasantly surprise the customer that we’re able to respond quickly and thoroughly and demonstrate that we have an understanding of their requirements and we fully comply. And as Steve mentioned as well, by responding quicker to RFQs, it frees up time to be more proactive in chasing down these deals.

**
11:42**
Simon Walmsley
So yeah, that’s my introduction and where I see the project helping.

**
11:48**
Daniel Englebretson
Awesome. Thank you for the deep answer there. Who do you want to pass the baton to?

**
11:54**
Simon Walmsley
Georgia. Georgia is a colleague of ours at Hughes.

**
11:59**
Georgia Hogg
Yeah. So sort of following on from Simon, same thing. Simon has more of the high volume paperwork side of the business. I’m more distribution, which is typically leaner, but we can still get complex documents. And I think from a sales team point of view, Simon has an engineering background. I don’t. So it’s also to help assist where some of us in the team that don’t have certain knowledge, the ability to, for a system to flag up things that we need to be considerate of if we might not be aware, speed of review and then also at the end of the chain that the costing is correct. So for example, if we need to change a specification because we missed something technically we could have really under costed the product.

**
12:50**
Georgia Hogg
And then after award stage, when it’s already reached engineering, to claw that money back is very difficult. So that would be my main consideration as well. So a helping us where we might not have the knowledge at the front end that it’ll maybe mean right flag this, siphon it into the engineering team for support or whatever else and then also cost incorrectly and the impact of that further down the chain.

**
13:17**
Daniel Englebretson
Great, that’s really good nuance. I really especially appreciate the kind of clawback statement you’re making. I didn’t, I didn’t recognize that earlier, so I’m glad you called that out. Very helpful. Let’s see who’s got. I think we got a couple more people left. Who do you want to pass the baton to?

**
13:34**
Georgia Hogg
I’ll pass on to James because he sees a lot of the impact here that we’ve spoken about.

**
13:41**
James Hamilton
Hi, Daniel. James Hamilton. I’m the Hughes Engineering Manager. So I look after both the UK and North America. I think everyone’s touched on it really well. Everything that comes into my team tends to be the technical engineered solutions, the high value orders. We typically see a lot come in from Simon where he’s had to go through quite a lot of technical documentation. For me, it’s around making sure that what we receive in is.

**
14:11**
Daniel Englebretson
Is clear.

**
14:13**
James Hamilton
It’s not always easy because we might not have all the right documentation. So maybe some sort of cross referencing to make sure if specifications are referenced within a document that we have received it or if there’s kind of a crossover there so we can quickly and easily cross reference specs. That’d be great. I suppose really it’s more about almost handing over a clean order. So I can just process it based on what’s being quoted and then if anything comes back, maybe I can. It’s easier to articulate back into the sales team if there’s a commercial impact or if it’s. If it’s a technical query that the team can resolve, it doesn’t need to go back to commercial.

**
14:55**
James Hamilton
It’s a little bit like everybody said, it’s really just kind of freeing up some of that resource so that we can process things as efficiently as possible. Not always easy, but yeah, be a nice challenge for us.

**
15:10**
Daniel Englebretson
Okay. Okay. When you say going back to commercial.

**
15:13**
Daniel Englebretson
Is that going back to somebody like Simon or what does that. Correct. What does that mean? Yeah. Okay, that’s what I thought. Yeah.

**
15:19**
James Hamilton
So typically, you know, we may have quoted something based on the documentation that we’ve received. And then all of a sudden the customer will come back and they’ll slightly move the goalposts on exactly what they want. But there’ll be a financial impact in there and it could be. Could be minor. And if it is a minor impact, we don’t send it back. I’ll just make that decision and the team will kind of process it. If it’s a significant impact, we’ll then discuss that with the sales team. And. And again, it might be that we don’t go back to the customer. It might be that we ask for a revised purchase order or requo or something like that. But again, if there’s a way to make.

**
15:56**
James Hamilton
To minimize that and it doesn’t happen a lot, it tends to, but it does tend to be on the more complex orders where, you know, we’ve got 5, 600 pages of documentation to go through. And great term used earlier, sort of like documentation triage. So, yeah, just to kind of almost. Bell embraces that process.

**
16:15**
Daniel Englebretson
Okay. Okay, Very helpful. Thank you. All right, I think we just got Amber left and Demika.

**
16:23**
Daniel Englebretson
Hello.

**
16:26**
Amber Mutwale-Broad
So I work with Georgia in the export team, so she’s basically covered over everything I would need it for. I don’t receive that many documents like Simon would, but. But it would be good to have something that goes, picks out things that I could potentially miss and then saves me spending loads of time with the engineering team having them to go through things. It would be good to have something in place first that would help me go through it before I needed to go straight to them. If that makes sense.

**
16:58**
Daniel Englebretson
Yes, that makes sense.

**
16:59**
Daniel Englebretson
All right, well, thank you.

**
17:02**
Daniel Englebretson
And then Bamika, I don’t know if you want to add anything, I know you’re driving, but you’re welcome to something real quick if you want.

**
17:07**
Bhumika Sachdev
No, I just parked. So for me it’s. I just want to make sure that their lives, people in the operation, sales and everybody’s life gets easier for bringing in some automations. So that’s basically my goal with this and also the knowledge that we learn expand it to like other business units.

**
17:29**
Daniel Englebretson
Absolutely, absolutely.

**
17:31**
Daniel Englebretson
All right, well on my end I.

**
17:33**
Daniel Englebretson
Also have Lane D with me who is a project manager on my team who is here observing to learn from how we’re doing this. So she may not come on the.

**
17:43**
Daniel Englebretson
Call much here, but she’s here.

**
17:45**
Daniel Englebretson
So just wanted to point that out.

**
17:46**
Daniel Englebretson
And then we’ve got me. And so really quick, just about me, my background, I actually grew up in.

**
17:55**
Daniel Englebretson
Manufacturing, if you will. Started my career at a quality test and measurement company that worked with a lot of manufacturing. A lot of automotive, a lot of aerospace, a lot of big manufacturing. Anybody that had tight quality standards and a lot of the big dogs would outsource their quality systems to us. And so I went from that into lean manufacturing, working for a large conglomerate called Danaher, like a $50 billion manufacturing conglomerate.

**
18:24**
Daniel Englebretson
And then from there I went into lighting actually.

**
18:27**
Daniel Englebretson
And if you want to see a complicated process for specking something, lighting is.

**
18:31**
Daniel Englebretson
About as complicated as it gets because.

**
18:33**
Daniel Englebretson
There’S like six layers to the channel. So I worked for the largest LED.

**
18:37**
Daniel Englebretson
Manufacturer for a while and then I moved into BC backed tech and then eventually I left all that.

**
18:44**
Daniel Englebretson
I started my own business. I ran a business for several years and sold it.

**
18:49**
Daniel Englebretson
And then now, the last couple of.

**
18:51**
Daniel Englebretson
Years I’ve been focused on what we’re doing today.

**
18:54**
Daniel Englebretson
And I wanted to give you the history there just so you can kind of see where my perspective is coming from. And then what I do today is I try to help people solve complicated problems with less effort. And we do that with technology but also with just a lot of critical thinking. And so I think one of the biggest things that I can bring to the table today is what kind of critical thinking, what kind of frameworks do we need to look at a process like this process, to be able to.

**
19:27**
Daniel Englebretson
Break that process down into something that AI can actually augment.

**
19:33**
Daniel Englebretson
And I know a couple people mentioned to me they had played around in ChatGPT or they’d done it, done some things here and there. And you probably found that it’s underwhelming or it’s, you know, maybe you get a little bit of speed out of it, but you start to run into.

**
19:47**
Daniel Englebretson
Some really big barriers of success when.

**
19:50**
Daniel Englebretson
You’Re using, you know, kind of just.

**
19:52**
Daniel Englebretson
Off the shelf chat.

**
19:54**
Daniel Englebretson
And a lot of that comes from.

**
19:57**
Daniel Englebretson
Architecture, but also some of that comes from process.

**
20:00**
Daniel Englebretson
And so I say all that, but I also want to flag that there’s nothing inherently expensive about working with AI like the. It doesn’t need to be this crazy expensive thing. And you’ll see that as we get into it today, especially if we take.

**
20:17**
Daniel Englebretson
The time to really break down what are we trying to do with it and why are we trying to do it.

**
20:23**
Daniel Englebretson
So that’s where I’m coming from and what I’m hoping to talk through today. I heard a lot of what you guys referenced, we talked about on the.

**
20:34**
Daniel Englebretson
Last call and I’ve been hearing from.

**
20:36**
Daniel Englebretson
But I did pick up a number of new things. And one thing that you will see as we go through this today is there’s no magic bullet. There’s just getting started and making incremental improvement continuously. And so what we need is how do we actually capture the learning and the incremental improvement with a system that every time you flag something, it recalls that you flagged that so that next time it knows to look for that. Right? So every time Simon’s in the system doing what Simon’s doing, then the system learns and then somebody else who’s got less experience hopefully can benefit from that. But what my belief is, and I think it’ll be clear as we get into it, is that this isn’t a flow that you can just eliminate a human from. That’s not what we’re trying to do.

**
21:31**
Daniel Englebretson
This is a flow that gives the humans working on it the speed and the capacity. Some of that what you guys talk about, to be able to augment the processes but still stay in the realm of thinking critically and thinking creatively about, man, I got these 5,000 pages of stuff. What do I do next?

**
21:52**
Daniel Englebretson
And helping you with some guardrails around, not missing the stuff and not reinventing.

**
21:58**
Daniel Englebretson
The wheel and flags that have been thrown in the past, not forgetting that.

**
22:03**
Daniel Englebretson
Has happened before and you know, to.

**
22:04**
Daniel Englebretson
Look for that’s kind of how I am thinking about this.

**
22:09**
Daniel Englebretson
So Bamika, it looked like it came off mute and wanted to add something. So did you want to add something?

**
22:13**
Bhumika Sachdev
No, actually I am in the office. So if not a network came.

**
22:19**
Daniel Englebretson
All right, no problem. So there’s a couple of things I wanted to speak to ground us before we get into some kind of heavy stuff. And the first thing I want to talk about is when we’re thinking about augmentation with technology, there’s a few ways to think about what level of augmentation we’re aiming for. The reason why I want to start here is because you can’t do the level five stuff on day one if you don’t have the processes and the.

**
22:57**
Daniel Englebretson
Documentation, the systems in place.

**
23:00**
Daniel Englebretson
It’s just as much about understanding where are we today and what’s feasible and what might be feasible in the future. As we break down this workflow today and we look at different segments of.

**
23:16**
Daniel Englebretson
The workflow, there might be some places.

**
23:18**
Daniel Englebretson
Where it’s level zero, there might be some places where it’s level five. And that’s part of what we’re going to be thinking about is what would have to be true, what would we have to know, what has to be documented for us to feel comfortable that.

**
23:31**
Daniel Englebretson
This is an opportunity for a level two or a level three or a level four and so on.

**
23:36**
Daniel Englebretson
And so as we go through the flow, you’ll see what I mean. But just to speak to it out loud so that we’re going to have a level set, levels neuro being no augmentation, pen and paper, basically, or you’re just doing it manually. Rarely is there no augmentation because you probably got a Grammarly or a spell check or a system or something in.

**
23:59**
Daniel Englebretson
There that’s helping you.

**
24:00**
Daniel Englebretson
But as you move up from the level zero and you start thinking about assistive tools, you know, level one, this would be like a summary. So I don’t know if you guys have ever been on a call where.

**
24:10**
Daniel Englebretson
There’S a recording and then at the end of it you get a little.

**
24:13**
Daniel Englebretson
Recap that says, here’s the 10 things.

**
24:14**
Daniel Englebretson
That were mentioned on the call.

**
24:16**
Daniel Englebretson
You know, that’s what I mean, that’s what Fireflies is. So, you know, you can.

**
24:21**
Daniel Englebretson
Summarize your inbox, you can summarize a document, you can summarize a transcript.

**
24:24**
Daniel Englebretson
You know, those types of things, they keep you from having to read the entire thing. You know, that’s 50 pages long and just Kind of suck out the things.

**
24:34**
Daniel Englebretson
That are interesting, you know, so you’ll.

**
24:36**
Daniel Englebretson
See some of that as we get into it. And then you might get past that up to a level two, where you have what you might think of as cognitive support.

**
24:44**
Daniel Englebretson
And this is like building a research report.

**
24:47**
Daniel Englebretson
So putting this in the.

**
24:49**
Daniel Englebretson
There’s an actual technical thing called deep research. Well, we don’t have to get too.

**
24:53**
Daniel Englebretson
Far into that, but thinking about what we’re doing here today, there’s a lot of opportunity for the quote, unquote, deep research, because you can pour through the hundred files that came over for the.

**
25:05**
Daniel Englebretson
RFQ and surface a bunch of information.

**
25:08**
Daniel Englebretson
Like, you know, what domain are we in?

**
25:10**
Daniel Englebretson
What standards are we looking for?

**
25:11**
Daniel Englebretson
What, what safety risks might there be in this? You know, what you know, so you can start to flag, you know, from that quick review, the highlights or the how did I think about basically the criteria that setting the stage for what.

**
25:28**
Daniel Englebretson
You’Re about to go do?

**
25:31**
Daniel Englebretson
So that is where a lot of the AI tools are today. When you’re working with AI tools starts to stop around there. And what happens is when you’re trying to use an AI tool on top of a process, it doesn’t have the context of the entire thing to be able to accurately move, you know, much past that. And so, you know, it doesn’t have all the documentation, doesn’t know your process, doesn’t know the whole flow to be able to start going much past that, you know, but as you get, as you move up that chain.

**
26:07**
Daniel Englebretson
And I promise this is the most.

**
26:08**
Daniel Englebretson
Abstract thing we’re doing today, but as you move up that chain and you start thinking about cognitive partnership. This is one of my favorite examples.

**
26:17**
Daniel Englebretson
Of this, is a technology called Notebook lm.

**
26:20**
Daniel Englebretson
But this is where you can, and you’ll see this today, you can give it a whole bunch of information and it creates a thing, creates.

**
26:26**
Daniel Englebretson
An executive summary for you based on all this stuff, where it creates a.

**
26:30**
Daniel Englebretson
Podcast review, or it creates something that you didn’t have before out of what you put into it. And so as we move down this chain, the main reason why I want to call this out is there are going to be times where it’s a zero, there are going to be times where it could be a 5, and there are going to be times where.

**
26:50**
Daniel Englebretson
You know, maybe a 5 is just not even possible.

**
26:54**
Daniel Englebretson
So that’s how I wanted to try to tee up some gradation between types of automation. Does anybody, as you’re thinking about what you know about AI and how you’ve used AI in your day to day. Does anybody have any examples of how they’re using AI right now that might.

**
27:16**
Daniel Englebretson
Fit on that spectrum? Just to kind of get our gears turning. Because it’s important, I really think it’s important for us to kind of get level set on what we’re thinking about.

**
27:28**
Stephen Hudson1
I can, I use it quite often these days to generate policies. So it kind of generic boilerplate, you know, policies for quality, ISO 9001. So I kind of, you know, you.

**
27:43**
Daniel Englebretson
Put, put a very brief.

**
27:44**
Stephen Hudson1
Prompt in it and it’ll soon spit out something which is very useful. And then you can just kind of change it up for context. So that’s one thing I use.

**
27:53**
Daniel Englebretson
It for quite a lot. That’s great use case. We’re gonna see that one now. So that’s a good one.

**
27:57**
Daniel Englebretson
Anybody else?

**
28:00**
Simon Walmsley
I’ve used it for a bit of technical research if it’s, you know, conscious of trying to sort of not be burdening engineering with questions all the time. So sometimes I can get a much more nuanced, broad based answer. You know what, I, I’m kind of relatively new to using AI and I kind of think you can find that information yourself by doing a Google search. But I like that you get collated responses back from many different sources so you get a more nuanced, structured response. We do have copilot, like an enterprise type license installed and obviously this whole project has got me kind of thinking about possibilities of what we could be doing with some of these RFQs. And I think it’s great.

**
28:50**
Simon Walmsley
But more recently I put in a data sheet where I had about 100 pages and the layout was exactly the same on every data sheet for a safety shower. And it was 80% were safety showers and 20% were eyewashes. But, but we didn’t have a summary of what the breakdown was. And there was certain parts fitted to some units and not on others which would have involved manually going through counting and totting up and making a summary of each one. So I thought, oh this is great, I’ll put it into AI. So I loaded it into copilot and I said, can you read this document? Yes, I can. Would you like a summary? Yeah, okay, that’s great. And then I said there’s a title which includes a tag number which is a unique identifier. So I said can you tell me how many contain?

**
29:42**
Simon Walmsley
And I put in quotes SS for safety shower and how many contain EW for eyewash. And it said, yes, I can. And it told me, it said it was like 102 safety showers and 18 eyewashes.

**
29:57**
Daniel Englebretson
Great.

**
29:58**
Simon Walmsley
And then I started to interrogate it a bit more to get some more information and then there was something, I can’t remember exactly what it was, but it didn’t quite stack up and it made me doubt it enough to think, right, I don’t want to just rely on this, so I’m going to manually do this. And I did it. I clicked through 100 odd pages of this data sheet and the summary it provided me was incorrect. There wasn’t 102 showers, it was like there was 115, say. And so I asked it, I said, this isn’t right, I’ve manually counted it. And the response was, oh, yes, you’re quite right. And it started to explain about some of the limitations it has about complex numbers of pages and, you know, all the rest of it.

**
30:39**
Simon Walmsley
So I would be interested to know how much we can rely on the responses that we’re getting out of it. When I found a recent example that it kind of let me down a little bit.

**
30:54**
Daniel Englebretson
Yes, that is a. I love that story and I’m going to come back to that story to address that. But before I come back to that story, want to see if anybody else has one they want to throw in the mix for us to think about here.

**
31:08**
James Hamilton
I can so a little bit like Simon, I put things in there from a technical perspective if I’m doing research and things like that. But I’ve also used it for screening CVs when I have a job description and I use it for good, well, good uploading CVs and job descriptions and then screening against the two. And then I normally my part, my interview technique is ask a series of standard questions. Sometimes I go back and put that in with my answers or with their answers to see whether or not they’re good, bad or indifferent. That’s not my sole decision making tool, I will add, but it’s something that I have used previously.

**
31:48**
Daniel Englebretson
Yes, that’s a good one. And I like how you mentioned the process of decision making. So I’m going to come back to decision making as well when we get after this part here. So does anybody have anything else they want to throw in to kind of color the picture here a bit more? Right.

**
32:08**
Simon Walmsley
I’d also say rewriting emails, sometimes that can be helpful. You’re trying to convey a message, you’ll tell it what your objective is and you’ll maybe put a first draft of an Email response and kind of, I guess it’s a bit basic, but, you know, can you make this more assertive yet professional and, you know, have a rewrite prompt, use it for that as well.

**
32:32**
Bhumika Sachdev
I do that a lot for email writing, but sometimes I do get, like. It doesn’t understand what I’m trying to say. And especially, like, if there is a chain of emails, it tries to read through the whole thing. It tries to rephrase what I said, because sometimes it’s really good. Sometimes I just write, type in, like, one line, and it will make up, like, one whole paragraph out of it, which. Which is good if you’re, like, sending an email to an executive or something. But sometimes it just completely does not understand the message that I’m trying to say. And it will, like, address, even change the address that I’m saying, hi, John. And it will change it to like, hi, Hannah, or something like that.

**
33:16**
Daniel Englebretson
Yes, that’s a good one. And just making the notes here so I can come back to it. That’s a good one as well. And one of my first four reasons LLMs was email rewrites, particularly when you had a hot message you needed to send that you wanted to make sure you sent in a professional way. But that’s a great one. So these are all really solid examples, which I’m going to speak back to in just a second here. But does anybody have anything else they want to throw into the mix before I address these?

**
33:47**
Stephen Hudson1
It’s very good at finding the best bars in Amsterdam.

**
33:51**
Daniel Englebretson
I’ll.

**
33:51**
Stephen Hudson1
I’ll certainly give it that.

**
33:55**
Daniel Englebretson
All right. All right. Do you live in Amsterdam? You don’t live in Amsterdam, do you? No, no.

**
34:02**
Stephen Hudson1
That’s why. That’s why I needed some guidance.

**
34:04**
Daniel Englebretson
Oh, fair enough. Yes.

**
34:07**
Daniel Englebretson
I can’t.

**
34:08**
Daniel Englebretson
I can’t remember if I mentioned this before, but I grew up on the border of Holland in Germany, and so. Oh, no way.

**
34:14**
Daniel Englebretson
Yeah, yeah.

**
34:16**
Stephen Hudson1
Where was that?

**
34:17**
Daniel Englebretson
Where was that?

**
34:18**
Daniel Englebretson
I. I went to school in Dusseldorf, but I lived on the border at a town, a small town called Calcar, but close to Nijmegen, but. Okay, I got you. I got you.

**
34:29**
Daniel Englebretson
Yeah, yeah, Cool.

**
34:31**
Daniel Englebretson
So.

**
34:32**
Daniel Englebretson
All right, well, these are. These are really solid examples. And so. But feel free to throw any others at me.

**
34:39**
Daniel Englebretson
Anyone who’s on the call here.

**
34:41**
Daniel Englebretson
So let’s speak to that for a moment because this is super helpful to cue up what we’re going to get into. So the deep research example, Simon, with the hallucinations. And so that’s what they’re called.

**
34:53**
Daniel Englebretson
Technically, they’re called the technical term is.

**
34:56**
Daniel Englebretson
Hallucinations, which is when you ask it to do it and it gives you answer back and it looks right and it smells right and it sounds.

**
35:02**
Daniel Englebretson
Right, but it’s wrong.

**
35:04**
Daniel Englebretson
And so the example where you said you had 110 and you actually had.

**
35:08**
Daniel Englebretson
115 and you had to go back to it.

**
35:11**
Daniel Englebretson
And this is important to call out as we’re thinking about what we’re about to get into, because AI, predictive AI, it’s not thinking, it’s calculating. It’s a math equation. And so the answer it’s giving you is not the solution, it’s a resolution, it’s an outcome, but it’s not the outcome. And so it’s just saying, okay, all the words that came before doing a bunch of math, here’s the answer. And so if you think about it like an algebra equation, every word that’s in the equation changes how it answers. But. But it’s not. It is not doing that in a like one plus one equals two kind of way. It’s doing that in a, well, if this, then this, then this. And how it interprets the if, then how it interprets that is you don’t.

**
36:07**
Daniel Englebretson
Have a lot of control over that.

**
36:09**
Daniel Englebretson
Especially when you’re asking a small question about a big set of information. So the second thing I want to call out here is LLMs. They are so like ChatGPT or Copilot, they don’t give you the right answer. They give you the answer they think you want. Okay? And because they’re trained to give you answers that you will like, normally you like answers that are right, but that’s not always exactly the same thing as the truth. And so if you think about the fact that it’s doing these giant math.

**
36:46**
Daniel Englebretson
Equations on all these variables that every.

**
36:48**
Daniel Englebretson
Variable that you put in is a little bit more likely it’s going to be wrong. And you think about it’s trying to give you what it thinks you want, not the right answer. That’s a really important concept for us to ground into as we get into what we’re getting into. Because what you need to do to address that is you need to break down the complexity of what you’re asking the LLM to do, so that when it’s answering in a much more narrow context. So, for example, instead of saying, what good bars are there? You might say, what good bars are there in Amsterdam? But instead of saying, what good bars are there in Amsterdam, you might say, I like Bars like a. How many of them are in Amsterdam? And instead of saying that, you might say, I like bars like a.

**
37:36**
Daniel Englebretson
That also have this type of beer, and I’m going to be in this neighborhood at Amsterdam. And as you break that down, the likelihood that it gives you answer that’s right goes way up. Because the prediction, the likelihood that the prediction is correct, if you can reduce the complexity of that and improve the likelihood that what it’s referencing is the right thing to reference, then you get a better answer. I know that this is a very big, abstract thing, but it matters a lot. Because when you think about a workflow like what we’re doing, and at the risk of blowing us out here a little bit, you might have a workflow that says, okay, we’re going to get an RFQ packet and we need to give them a quote. If that’s what you say at the top, that’s this huge question, right?

**
38:24**
Daniel Englebretson
But as you start to break that down, well, it breaks down into these different links of activities, which then breaks down further into these functions, which then breaks down further into these tasks, which then breaks down further into these steps that you might take, which then breaks down even further into, how do I QA at this step versus this step? And so we’re gonna. It’s not important what this screen shows, except to get the point that you can turn your question of what good bars are in amsterd into breakdown, A breakdown, a breakdown, a breakdown. And then when it gets to this smaller chunk and asks those questions in a row, then you’re going to get a better answer.

**
39:05**
Daniel Englebretson
This might sound crazy when you’re thinking about the flow that we’re talking about, except that my job and what we’re using the tech to do is help us break it all the way down so that we have. I call them the atomic units, but we have the smallest units that make sense for the AI to be able to react to. That’s a lot of what we’re going to get into as we go through it today. Because then you can say, all right, well, with this chunk of work, this makes sense. But with this chunk of work, that makes sense. And when I’m doing flags for safety requirements and explosive environments, maybe I’m not going to use this level, maybe I’m going to use this level, because you know that’s a higher risk scenario.

**
39:46**
Daniel Englebretson
And so we have to do the work of saying, all right, well, what are the decisions that need to be made? So the second example is screening CVs, uploading, screening and getting answers and helping you make decisions. I think that is a really important one to think about, because some decisions you should be making, other decisions maybe you don’t care as much about making actively. And so part of what we’re also going to see as we get into this is at what point does it make sense to flag. Okay, hold up. It’s time to ask a human a question so that you can support the decision making. As we’re thinking about that right now, and as we’re thinking about something like this big flow, one thing that I want to call out that’s different with AI and natural language processing, which is.

**
40:37**
Daniel Englebretson
What LLMs and ChatGPT can do for.

**
40:39**
Daniel Englebretson
Us, is everyone here is probably familiar with a CRM and CPQ and structured data in the database. And if you’ve got to go into the salesforce and you got to look up the product line or you got to pick it off of a list and you add it to a quote, that process requires a lot of structure, right? You have to have a price book, you have to have a product group, you have to have all the products in there and all that and all these tables, right? And so if you had to structure this whole thing out, that’d be a lot of fields and a lot of tables and a lot of databases. But with AI, the need to structure the data is very different.

**
41:14**
Daniel Englebretson
And instead of structuring it as a table, like with a product feature or something in there, the right way to think about this is structure it as a question. What question do you need to ask of the data and what are you.

**
41:27**
Daniel Englebretson
Going to do with the answer?

**
41:29**
Daniel Englebretson
And so if we start thinking about keeping the bar example, because it’s really easy for us to get our heads around if we are thinking about, well, instead of what do I need to break this down into a salesforce table, what do I need to break this down into? The questions. And why am I asking those questions? That is where you start to get closer to what AI can do for us. A lot of what we’ve done so far as we’ve been working on this is we’ve been thinking about those questions and you guys have given me a ton of feedback on what are you asking of the data and why.

**
42:02**
Daniel Englebretson
If you think about, and I’m going to speak to this equation in a second, but if you think about some of the things you called out from a value perspective, you’re getting tons of data in, you got to go through it all. You Got to understand what you have. You try to mitigate these different flags. Well, if you start thinking about that in terms of questions, okay, what data did I get? How much is it? What is it about? You start breaking that down. You can start to get it to points where the AI can react, actually, before I hit this, because I know.

**
42:33**
Daniel Englebretson
This is a somewhat abstract concept.

**
42:36**
Daniel Englebretson
Is anybody not following what I’m getting.

**
42:38**
Daniel Englebretson
After with this breaking down into questions and kind of breaking down these bigger problems and the smaller problems?

**
42:43**
Daniel Englebretson
Is anybody not following or does anybody.

**
42:45**
Daniel Englebretson
Want me to give a different example or approaches a little differently before I go any further?

**
42:53**
Daniel Englebretson
All right, okay, so I’m going to keep going then, but please stop me. And we are going to get into a demo here in just a second.

**
43:01**
Daniel Englebretson
So it’s going to get really cool.

**
43:03**
Daniel Englebretson
So we talked about rewriting emails, and you mentioned, okay, you might want to rewrite emails. So you say, here’s what I’m trying to do. You give it a draft, you give.

**
43:13**
Daniel Englebretson
It a goal, and you say, and.

**
43:14**
Daniel Englebretson
Then you get something back. That process that you talk through of, okay, here’s my email, here’s what I’m trying to do. Here’s a draft, here’s what I need you to do. And they get it back. That process is you specify a workflow for the AI to do. It’s just a really straight, small one, right? You said, look at this thing. I want to do this thing. Here’s an example of what good looks like. Here’s why I want to do it. Now go. And it works really well on an email because it’s a really small thing. But when you want to blow that out to a really big thing, you can’t just say, give me my RFQ quote. Look at the stuff and give me my quote. Because it has no idea what you mean. It can’t do that calculation.

**
43:53**
Daniel Englebretson
It has no idea what you mean. It doesn’t understand the nuance that you understand. And so what will happen is it.

**
44:01**
Daniel Englebretson
Either can’t do it or it’ll just make it up.

**
44:03**
Daniel Englebretson
And where it’s actually more dangerous where it makes it up, because you can’t.

**
44:07**
Daniel Englebretson
Always tell that it made it up.

**
44:09**
Daniel Englebretson
So we’re almost out of theory section here, but I think it’s an important place to ground. So when you’re thinking about that, when you’re thinking about those questions, when you’re thinking about what questions are we asking of the data, basically it breaks down into these four things that the AI can help us. Do you see a problem or you observe a thing. The problem might be you just received a zip file with a thousand files in it and you need to react to this zip file. My problem is I just got this big zip file of stuff that I need to react to. My goal is this is a big opportunity. I need to get this quote turned around, like this week because it’s this huge deal and we’re late on it or whatever.

**
44:53**
Daniel Englebretson
So that’s my goal.

**
44:54**
Daniel Englebretson
I got to get my RFQ out or my. I got my quote out. So. So, you know, based on your experience, the fastest way for me to do this is do this and this and this. And if you do those things based on your experience, you will be able to achieve your goal and resolve that. That problem. And when you’re done, you say, man, that was crazy. We missed this thing over here because it was in this spreadsheet over there that I didn’t even know to look for, or I missed it because there was a thousand pictures or whatever in here. So next time we do this.

**
45:26**
Daniel Englebretson
Is what we’re going to look for.

**
45:27**
Daniel Englebretson
And that’s that learning piece.

**
45:29**
Daniel Englebretson
Right?

**
45:29**
Daniel Englebretson
And so, and so whatever you’re doing, whatever question you’re trying to answer at any level in this, you know, which bar is in Amsterdam or which bar in the neighborhood, or which bar with French fries, you know, it’s the same. It’s the same process of you’re trying to observe something from the data. And if you know what you’re trying to observe from your data, like, I.

**
45:50**
Daniel Englebretson
Know I need to respond to this.

**
45:52**
Daniel Englebretson
And then you’re trying to set a goal, okay, when this happens, do this, and then you’re trying to respond to that, and then you’re trying to learn from that. And so that flow is what we’re going to be really getting into, except we’re getting into it and progressively more.

**
46:10**
Daniel Englebretson
Like smaller units of work as you go through it.

**
46:13**
Daniel Englebretson
So, so that’s kind of it. On the deep theory piece of this, before we get into the rest of it, does anybody have any questions about how I broke down kind of what this knowledge workflows are? This idea of questions of data and.

**
46:28**
Daniel Englebretson
Doing things with answers and decisions and things like that. Does anybody want me to take a.

**
46:32**
Daniel Englebretson
Step back on that?

**
46:33**
Daniel Englebretson
Are we feeling good?

**
46:36**
Daniel Englebretson
All right, we’re going to get into now something that you guys are. The, the first time I presented it in this way, and, you know, I’ve worked with this stuff a lot, but.

**
46:50**
Daniel Englebretson
It’S the first time I presented it this way. And it’s a little bit overwhelming on.

**
46:53**
Daniel Englebretson
The surface, but I don’t think I have to tell you guys how complex this process is. And so it’s. There’s no real way to not be.

**
47:01**
Daniel Englebretson
Complicated with a complex problem process.

**
47:04**
Daniel Englebretson
So I’m going to give you a link to this, but I thought you might appreciate. What does this process look like? This is what the process looks like when you get information in and you’re trying to react to information. You’ve got these different standard works and you’ve got these different documents that you have to go sift through in order to find all these different things. When you’re trying to react to something, you get these really complicated sets of information that you have to go through. And so what I. I got two versions of this.

**
47:37**
Daniel Englebretson
I’m not happy with the styling of this one, so I’m going to. I’m going to use this other version today and then give you a different one.

**
47:43**
Daniel Englebretson
But we’ve got what I’m about to send you. I’m going to give it to you now as a link where you can come in here and see all of the information. And so, like, kind of blowing it back out. You’ll be able to see the information now. I want to highlight here. You guys gave me, I think, thousands of files, definitely many hundreds of files, and they’re in here, so it’s easy to get lost. So don’t try to make sense of all of the files that are in here right now, because you’ll get lost in it. To do this, if you go to Infinity Eleonox AI, it’s going to ask.

**
48:19**
Daniel Englebretson
You for a password and I’ll put.

**
48:21**
Daniel Englebretson
It in the chat.

**
48:23**
Daniel Englebretson
And the password is just right, all lowercase.

**
48:27**
Daniel Englebretson
And when you get there, you should.

**
48:30**
Daniel Englebretson
Land on the design and build workshop agenda. But if you don’t, you can click on the design and build workshop agenda. So the password, just in case, is just right.

**
48:41**
Daniel Englebretson
So the other version of this I.

**
48:44**
Daniel Englebretson
Intend to give you guys. After all of this, I just, I want to change the styling because I don’t. I just don’t like the styling.

**
48:50**
Daniel Englebretson
So I’m going to change the styling.

**
48:52**
Daniel Englebretson
But this one requires authentication with your email address. So after the workshop today, I’m going to take down the one I just gave you and give you this one.

**
49:02**
Daniel Englebretson
But they’re the same thing.

**
49:03**
Daniel Englebretson
And so that’s what we’re looking at.

**
49:05**
Daniel Englebretson
And so when you come in here, there’s a lot of information I Just.

**
49:09**
Daniel Englebretson
Have to take a minute to ground us in this before we can go further, because there’s a lot of information.

**
49:14**
Daniel Englebretson
So there’s a few things that we’re going to be looking at. First of all, we’re in the design.

**
49:19**
Daniel Englebretson
And build workshop today, which is this folder up here at the top.

**
49:23**
Daniel Englebretson
And in here we have this safety.

**
49:25**
Daniel Englebretson
Shower workflow that we’re going to be talking about.

**
49:27**
Daniel Englebretson
And then we have some documentation that.

**
49:29**
Daniel Englebretson
I had to create to make this possible. We’ve got things like standard works, like you mentioned, creating templates.

**
49:35**
Daniel Englebretson
So all these different templates that start to support, well, what are the decisions.

**
49:40**
Daniel Englebretson
And what does good look like that comes out.

**
49:42**
Daniel Englebretson
So we’re going to go through these. We’re going to go through this in the right level of detail today. But you’ve got the design and build.

**
49:47**
Daniel Englebretson
Workshop, then you’ve got the discovery documentation, which is what you originally gave me.

**
49:52**
Daniel Englebretson
Then you’ve got the reference examples, which.

**
49:55**
Daniel Englebretson
Is all the examples you supplied to me. You can see them all broken down in here.

**
50:01**
Daniel Englebretson
Then we’ve got some trial runs, but I don’t want you to get lost in the trial runs yet, because I need to explain this a bit more. But you can see a few different runs of how did it actually go through and react to these different proposals and where did it end up at the end?

**
50:18**
Daniel Englebretson
So you get to the end and you’ve got outcomes from whatever thing.

**
50:22**
Daniel Englebretson
So we’re going to go through this. But I just wanted you to get.

**
50:25**
Daniel Englebretson
A sense of what is in here before I go any further.

**
50:30**
Daniel Englebretson
And so starting with the design and.

**
50:32**
Daniel Englebretson
Build workshop agenda, which is the page.

**
50:34**
Daniel Englebretson
Here, I spent more time on the.

**
50:37**
Daniel Englebretson
Front end setting this up because I think it’s important to get grounded in kind of the concepts.

**
50:42**
Daniel Englebretson
But we’re going to go through now and think about what is the workflow today, what have we found as we started going through this and how can we.

**
50:50**
Daniel Englebretson
Like the safety shower RFQ workflow?

**
50:52**
Daniel Englebretson
This is how you might would start to actually break it down into.

**
50:56**
Daniel Englebretson
Okay, the first thing we’re doing is RFQ intake and triage.

**
51:00**
Daniel Englebretson
And what does that look like?

**
51:01**
Daniel Englebretson
Well, we got to do all of.

**
51:03**
Daniel Englebretson
These different steps in order to get it done. And then we’re going to get into the next big thing, which is requirements analysis. And then you got to break all that down. And so that’s what we’re going to go through today. And we’re going to do some of.

**
51:14**
Daniel Englebretson
This live, so you can see what happens as we go, as we’re doing It.

**
51:18**
Daniel Englebretson
So I know this is a ton and it’s probably overwhelming, but. But I needed to ground us in.

**
51:24**
Daniel Englebretson
It before we could go any further. So you knew what were getting into.

**
51:28**
Daniel Englebretson
So let me pause there and just see. First of all, did anybody have a hard time getting access to this? Does everybody have it?

**
51:35**
Daniel Englebretson
Is there anything that you want me to address as far as navigating or grounding in this before we go any further? All right.

**
51:46**
Bhumika Sachdev
Can I ask you a question?

**
51:48**
Daniel Englebretson
Yes.

**
51:49**
Bhumika Sachdev
How long did it take you to build this?

**
51:53**
Daniel Englebretson
That’s a good question.

**
51:54**
Daniel Englebretson
I will tell you. It took me longer than I anticipated because there was way, way.

**
51:58**
Daniel Englebretson
More information than I anticipated.

**
52:01**
Daniel Englebretson
But.

**
52:04**
Daniel Englebretson
We’Ve got it, so now we.

**
52:05**
Daniel Englebretson
Can work with it.

**
52:06**
Simon Walmsley
So.

**
52:07**
Daniel Englebretson
So I’m glad you asked that because I mentioned this earlier. There’s no magic bullet with AI. You don’t just magically get your RFQ process. What you do get is a place.

**
52:20**
Daniel Englebretson
To start at a place to start improving.

**
52:23**
Daniel Englebretson
Part of what we’re going to talk about today is where are we starting and how are we going to lock incremental improvement to move up the chain of complexity that we can realistically expect with augmentation? And why I call that out because like I mentioned earlier, the difference between static tables and CSV and a CRM.

**
52:49**
Daniel Englebretson
Or something like that versus unstructured.

**
52:53**
Daniel Englebretson
Less structured information that we’re reacting to when we go to actually build a system that does this, and we’re using the.

**
53:00**
Daniel Englebretson
Microsoft stack, for example, which is what.

**
53:02**
Daniel Englebretson
You guys use, and you want to interface with what we’re about to do through teams, and this is possible, but.

**
53:08**
Daniel Englebretson
Not necessarily the way that we will do it.

**
53:10**
Daniel Englebretson
You could hit a RFQ bot in teams and be like, I got my new proposal, go. And all of this can happen inside of teams. And what we’re looking at on the back end is basically infrastructure behind that. And so we’re going to pretty much.

**
53:25**
Daniel Englebretson
Go through that now.

**
53:27**
Daniel Englebretson
So originally when we talked about doing this, we talked about trying to do.

**
53:31**
Daniel Englebretson
Like a chat, cpt, GPT or project or something to. As a mechanism. But what I learned going through all this is that there’s just way too much complexity to take this whole.

**
53:42**
Daniel Englebretson
Thing and do it. You might decide that you want to take pieces of this and do as a GPT, but I suspect that we.

**
53:49**
Daniel Englebretson
Will end up with something a bit more complicated than that. And so, and so before I go.

**
53:56**
Daniel Englebretson
Any further on this, I want to.

**
53:58**
Daniel Englebretson
Kick off one for you so that you can see something and so.

**
54:04**
Daniel Englebretson
There’S this, what I’m about to kick.

**
54:06**
Daniel Englebretson
Off, this is how I do it locally on my computer. Don’t get too worried about what the technology is here. This is only for grounding.

**
54:18**
Daniel Englebretson
So if I come in here. So what we’re looking at is the.

**
54:25**
Daniel Englebretson
Same set of information that you’re looking at.

**
54:27**
Daniel Englebretson
So this is the same, you probably.

**
54:29**
Daniel Englebretson
Can observe it’s the same set of information.

**
54:32**
Daniel Englebretson
So this would, you would never run.

**
54:33**
Daniel Englebretson
A system like what I’m doing here because this is very brittle. But it will demonstrate the point.

**
54:38**
Daniel Englebretson
So I got my same set of information.

**
54:40**
Daniel Englebretson
I’m working with an LLM. It’s got memory, it’s got capability. And I’m coming over here and I’m like, okay.

**
54:49**
Daniel Englebretson
I’ve got a new RFQ that I got to get done.

**
54:52**
Daniel Englebretson
In this case, we’re going to do it for Techmont Skikba.

**
54:59**
Daniel Englebretson
So in the website that you have.

**
55:02**
Daniel Englebretson
Access to, if you look under trial runs, you should see one that ends with Q32976. You don’t need to go into that right now, but that’s what I’m reacting to.

**
55:15**
Daniel Englebretson
And so for memory, in the reference examples you can find the 32976 reference that you gave me. And in here are all of the.

**
55:29**
Daniel Englebretson
Pre qualification files, RFQ files, the reference documents.

**
55:33**
Daniel Englebretson
These are all the things that you.

**
55:35**
Daniel Englebretson
Guys gave me as an example. So what we’re going to do now.

**
55:39**
Daniel Englebretson
Is say, okay.

**
55:42**
Daniel Englebretson
Let’S react to this example.

**
55:46**
Daniel Englebretson
Now. My AI.

**
55:49**
Daniel Englebretson
And again, you wouldn’t actually do it this way locally. I’m just showing you a back end view of this so that you can see what I’m talking about.

**
55:57**
Daniel Englebretson
Now what the AI is going to do is it’s going to follow all of the standard works and all the process that we have established and it’s going to execute that whole flow and.

**
56:09**
Daniel Englebretson
Come back and give your answer.

**
56:11**
Daniel Englebretson
And so I, we’re not going to sit here, I’m gonna let you see it kick off here, but we’re.

**
56:16**
Daniel Englebretson
Not gonna sit here and watch it all the way through. We’ll come back to it when it’s done. All right, so here it goes.

**
56:20**
Daniel Englebretson
It’s, it’s, it’s gotten grounded and it’s gonna take it a second for.

**
56:24**
Daniel Englebretson
For it to start giving you some response and I just want you to.

**
56:27**
Daniel Englebretson
Kind of see what’s happening and then we’re gonna come back to it. But I wanted to start at the end so that you could see what would potentially happen. So our bot is going through and it’s reviewing the files, it’s reviewing the.

**
56:41**
Daniel Englebretson
Structure, it’s going through the references and it’s so you can see it just.

**
56:46**
Daniel Englebretson
Read the RFQ workflow, it read the.

**
56:48**
Daniel Englebretson
Sequencing, it read the citation style guide.

**
56:51**
Daniel Englebretson
So it’s reading these documents that it’s told this is the process to take and we’re going to come back to what these things mean. So don’t get hung up on that.

**
57:00**
Daniel Englebretson
I’m just trying to show you where.

**
57:01**
Daniel Englebretson
This might would end up. Then it’s going to start executing on this process. The reason why I wanted to start here is because I wanted to show you if we establish what we want the bot to do, this is how it goes about doing it. If this is how it goes about doing it, what do we need to make sure is right so that the bot does it right? That’s what we’re going to get into. You can see it’s going through a compliance matrix, it’s going through a hazards matrix, it’s going through these different things which we’re going to look at because we have to decide what does good.

**
57:40**
Daniel Englebretson
Look like when you’re doing a cross reference.

**
57:42**
Daniel Englebretson
What does good look like when you’re flagging technical issues or whatever that would be. What I did to tee this up is I gave it the 10 examples or whatever that you had given me and I did a few rounds of it teaching itself to get it jumped off. So it’s going through it now I’m going to switch back to the other.

**
58:04**
Daniel Englebretson
Screen and then I’ll pull this back up when it starts making progress so that you can see what’s happening.

**
58:15**
Daniel Englebretson
For anybody who wants to understand the.

**
58:16**
Daniel Englebretson
Deep technical here, which may not be.

**
58:19**
Daniel Englebretson
Anybody here, what it would take for this to be possible is it’s taking all of your standards and all of the SOPs and all the standard works and all the guidance documentation.

**
58:34**
Daniel Englebretson
So let me flip back over to the website so you can see. So I have a slightly different view of this, but I’m going to pull up the same thing you guys have access to.

**
58:44**
Daniel Englebretson
So this is just my look at it on the back end of the.

**
58:47**
Daniel Englebretson
Same thing that you’re looking at on.

**
58:48**
Daniel Englebretson
The website is going through the standard work. So it has like a citation style.

**
58:56**
Daniel Englebretson
Guy and a client profiles and evidence pack and high risk keywords.

**
59:02**
Daniel Englebretson
So for example, if it knows anytime you see these high risk issues, that’s something to pay attention to. Right. So then what should be on the high Risk list. Right, that’s the kind of thing. But my point and the fact that this would learn is that every time you run a cycle, if you append to the list high risk terminology, then it knows to look for it next time. So what we have to do is we’re thinking about building a flow and is we have to know we should.

**
59:30**
Daniel Englebretson
Look for high risk keywords.

**
59:31**
Daniel Englebretson
We don’t necessarily have to know all.

**
59:32**
Daniel Englebretson
Of them, we just have to know.

**
59:33**
Daniel Englebretson
That we should look for them and that way the bot knows to go look for them. So that’s where we are going with this. But I started at the end so that you could appreciate kind of where.

**
59:48**
Daniel Englebretson
I’m going with this so that we can go through it. So I’m going to flip back over really fast so you can kind of.

**
59:54**
Daniel Englebretson
See what’s happening and then we’re going.

**
59:56**
Daniel Englebretson
To get into something that’s easier to get our arms around.

**
59:59**
Daniel Englebretson
It’s still over here running. And it created its first, it started.

**
01:00:03**
Daniel Englebretson
Creating its first things, decision tables, CPQ mapping. It’s starting to create these different documents per the process. You can see it going through it.

**
01:00:16**
Daniel Englebretson
Just showing you that it’s still running. Let me go back to our content.

**
01:00:25**
Daniel Englebretson
And then I’m going to make this easier for us to get our arms around, back to our agenda.

**
01:00:32**
Daniel Englebretson
What I wanted to do up until this point was ground us in what are we trying to do and ground.

**
01:00:38**
Daniel Englebretson
Us in what is the current state of the workflow.

**
01:00:42**
Daniel Englebretson
The current state of the workflow, if.

**
01:00:44**
Daniel Englebretson
You’Re looking from the agenda is this assessment of the safety shower compliance workflow.

**
01:00:50**
Daniel Englebretson
And this is what Stephen and Simon.

**
01:00:51**
Daniel Englebretson
And Bamika already went through with me at the Friction Workshop where we kind of went through this and.

**
01:00:56**
Daniel Englebretson
We said, all right, here’s kind of.

**
01:00:58**
Daniel Englebretson
How this flow might would run. You know, first thing we’re doing is we’re reviewing the inquiry and then we’re analyzing the technical requirements and then and so on.

**
01:01:08**
Daniel Englebretson
So, so you guys have seen that. And so from that then and from the feedback that you guys gave me.

**
01:01:15**
Daniel Englebretson
And I’m going to pause here for discussion, but from the feedback that you.

**
01:01:18**
Daniel Englebretson
Gave me, I then we sent out a questionnaire to the team and we asked about, okay, what should we be looking for?

**
01:01:31**
Daniel Englebretson
What might we need to keep an eye out for? So I think probably everybody has seen this, but just to kind of catch.

**
01:01:37**
Daniel Englebretson
Us up, we heard about a number.

**
01:01:39**
Daniel Englebretson
Of things and you guys added a few more to the list today of.

**
01:01:42**
Daniel Englebretson
Things that we should be Thinking about.

**
01:01:44**
Daniel Englebretson
When we’re going through this process.

**
01:01:46**
Daniel Englebretson
So what we’re going to be hovering on for a while is how this.

**
01:01:51**
Daniel Englebretson
Translated into these atomic units of work or these smaller units of work. And so to that end, and this is where we’ll be pausing, we’re looking at the Safety Shower RFQ workflow. And so if you’re in the, if.

**
01:02:04**
Daniel Englebretson
You’Re on the site, this is under.

**
01:02:06**
Daniel Englebretson
The design and build workshop navigation on the top. And then you’re going to click Safety Shower RFQ workflow. So it should look like this. So I think it’s got a few screens open on me. The way that this site is built right now is it will stack more than one document next to it. And if you want to close one of the other documents out, you can hit the little X down here to close it out. That’s how this one is built. When I turn this over to you, instead of it being like this, it is going to look more like this. So I just wasn’t happy with the styling, so I want to fix that.

**
01:02:55**
Daniel Englebretson
But this is what it will look.

**
01:02:56**
Daniel Englebretson
Like when I turn back over to you.

**
01:02:57**
Daniel Englebretson
So, all right, so going back to.

**
01:03:01**
Daniel Englebretson
My view here, and then we’re going to start talking about it.

**
01:03:09**
Bhumika Sachdev
Just going.

**
01:03:09**
Daniel Englebretson
To turn this into a reading view.

**
01:03:13**
Daniel Englebretson
All right, so the first thing I want to call out is I mentioned.

**
01:03:23**
Daniel Englebretson
This earlier and this is where we’re going to start collaborating on this and making changes.

**
01:03:27**
Daniel Englebretson
And also we can on the fly.

**
01:03:30**
Daniel Englebretson
See how the bot reacts when we make changes. So we’ll do that here in a moment.

**
01:03:35**
Daniel Englebretson
But the first thing I want to highlight is that using our bar example.

**
01:03:39**
Daniel Englebretson
Because it’s easy, instead of saying good.

**
01:03:41**
Daniel Englebretson
Bars, good bars of neighborhood, that’s what were trying to break down.

**
01:03:45**
Daniel Englebretson
And what I call these case files.

**
01:03:46**
Daniel Englebretson
In here, as you think about how.

**
01:03:49**
Daniel Englebretson
This process broke down and my navigation is not. Let me scroll over.

**
01:03:55**
Daniel Englebretson
It was RFQ intake and triage requirements.

**
01:03:58**
Daniel Englebretson
Analysis solution, configuration proposal, handoff contract awarded.

**
01:04:03**
Daniel Englebretson
So that was kind of the highest level flow. And so what we did, if you.

**
01:04:08**
Daniel Englebretson
Recall, I had like all the boxes and then breaking the boxes down as I took those and broke it into.

**
01:04:15**
Daniel Englebretson
All right, what do. What are we doing in this step that’s valuable and why is it valuable.

**
01:04:21**
Daniel Englebretson
To the next person in the process?

**
01:04:23**
Daniel Englebretson
So step one.

**
01:04:24**
Daniel Englebretson
So first hour of contact with new.

**
01:04:26**
Daniel Englebretson
Rfq, remove it from a state of chaos to organize readiness. So we’re transforming a chaotic dump of up to 5,000 pages of stuff. And what we’re trying to do is get clarity and confidence of what we’re looking at so that when we kind of pass it down the chain and this, we could be in the AI, it could be a person.

**
01:04:46**
Daniel Englebretson
Proposal engineer is able to understand, where.

**
01:04:48**
Daniel Englebretson
Am I starting with this? And so it sits kind of step one.

**
01:04:52**
Daniel Englebretson
So proposal engineering is going to do.

**
01:04:54**
Daniel Englebretson
The initial RFQ assessment and so signal goal, steps, outcome, email, zip file shows up. You know, the goal is to achieve a confident go notice it go, no go decision. Do we have everything that we need to even be able to react to this or what am I missing here? You know, that’s kind of the first thing. And so to do that we’ve got to get the documents out, we gotta hunt through it and find the cane.

**
01:05:17**
Daniel Englebretson
I know this is supposed to say materials request, the main.

**
01:05:21**
Daniel Englebretson
Mr. We’ve got to gut check if it looks like we got everything. We got to get a feel for the size. Is this a one day thing? Is this a huge project? How long is this going to take? We’re looking for any corrupted files or anything like that and then we’re making a call. Are we ready?

**
01:05:35**
Daniel Englebretson
Can we actually process this thing or not?

**
01:05:37**
Daniel Englebretson
You might be able to do that off the top of your head. Maybe that’s something that comes really fast or maybe that’s something that only a really experienced person can do and the junior person doesn’t really know. And so, okay, well how would we break this down?

**
01:05:49**
Daniel Englebretson
And I’m going to kind of lay.

**
01:05:51**
Daniel Englebretson
This out, then we’re going to speak to it. Well, you got to have procedural knowledge. You got to understand, okay, what do I got to do to break this thing down? You got to have conceptual knowledge. So what’s the difference between a high qual, a high priority, a low priority. And you’ve got to be able to.

**
01:06:05**
Daniel Englebretson
Understand the facts, like, what am I looking at?

**
01:06:07**
Daniel Englebretson
You know, what do these things mean? And so realistically, being able to remember last time I saw this, being able to understand and being able to apply, what do I do? That’s what we’re trying to get after. So when you break it down to this kind of level, you start seeing, okay, well, you might have a proposal engineer, you might have a project orchestrator, you might have an executive sponsor.

**
01:06:28**
Daniel Englebretson
And this is kind of how they’re connected to it.

**
01:06:30**
Daniel Englebretson
And so our goal then here, what we think we want to do with AI is let’s make the work easier, you know, so the primary pain isn’t the quality of the final. The experts, the experts know how to do it, right?

**
01:06:43**
Daniel Englebretson
It’s just the huge volume of stuff.

**
01:06:46**
Daniel Englebretson
And so, and so I’m just going to kind of walk us through and then we’ll. We’ll start at the top. So if you break that into chunks of work that the AI could do, if you think back to questions that the AI could reasonably answer for you know, step one, you know, do we have like, is there new unhandled information?

**
01:07:05**
Daniel Englebretson
Yes, it just came in the inbox. You know, so what am I going to do?

**
01:07:08**
Daniel Englebretson
How am I going to understand?

**
01:07:09**
Daniel Englebretson
And so on.

**
01:07:10**
Daniel Englebretson
So that’s how we’re breaking it down to what the AI would do. And then at the end, here’s how you might want automate. And then when you’re done with that.

**
01:07:18**
Daniel Englebretson
And then we’re going to kind of piece this one at a time.

**
01:07:20**
Daniel Englebretson
You know, the compounding advantage is, hey, if you can get a pattern of.

**
01:07:24**
Daniel Englebretson
Clarity and control from the very beginning.

**
01:07:26**
Daniel Englebretson
Every time, that will help reduce the anxiety, reduce the stress load, speed up these things, make it more repeatable and.

**
01:07:33**
Daniel Englebretson
Scalable, and so on.

**
01:07:34**
Daniel Englebretson
So this step is kind of the most obvious, easiest step in the process because you’re reacting to the materials coming in. But we break that down then into.

**
01:07:43**
Daniel Englebretson
Number two and then number three and number four, and off it goes.

**
01:07:48**
Daniel Englebretson
And so if you are navigating this.

**
01:07:51**
Daniel Englebretson
And then we’re going to go through.

**
01:07:52**
Daniel Englebretson
It, you can see it breaking down.

**
01:07:55**
Daniel Englebretson
Into all of the things that the bot would need to go do or the human to be able to run it. And so that’s where you’re getting into what the overall flow would look like.

**
01:08:10**
Daniel Englebretson
So, so that’s what we’re going to be spending a lot of time going through now. And the reason why I want to.

**
01:08:17**
Daniel Englebretson
Queue this up is because if we flip back over to our bot.

**
01:08:21**
Daniel Englebretson
Friend over here.

**
01:08:24**
Daniel Englebretson
We can see that the bot just finished going through whatever.

**
01:08:27**
Daniel Englebretson
Process and it came back and it said, all right, I need your. I need you to answer these questions.

**
01:08:33**
Daniel Englebretson
You know, confirm tiller inclusion, proceed with.

**
01:08:35**
Daniel Englebretson
This, or do you prefer that, you know, IP target? Should we assume this spec unless APC.

**
01:08:40**
Daniel Englebretson
Concerns other, or do you want to preemptively select, you know, do you have.

**
01:08:45**
Daniel Englebretson
EMC decoration, so on, so on to drop in, you know, and.

**
01:08:49**
Daniel Englebretson
Here’S a summary of what I’ve been through so far. So, like, we’ll just keep it going here. So we’ll say, first of all, I’m going to say, prefer this, I’m just going to make this up. But if you want to give me. If you want to tell me to do a different answer, we’ll do a different. Then we’re going to say preemptively select.

**
01:09:07**
Daniel Englebretson
I don’t know why I would say this, but you know why you would say this.

**
01:09:11**
Daniel Englebretson
Then we’re going to say, you have. I’m going to say, no, I don’t. Just to keep us going.

**
01:09:34**
Daniel Englebretson
All right, so this, the fact that it came back and asked for feedback is because it knows when I went through these things and I said, okay.

**
01:09:44**
Daniel Englebretson
I first got to do this and.

**
01:09:45**
Daniel Englebretson
Then I got to do all these.

**
01:09:46**
Daniel Englebretson
Things and I got to look at all these things and this is how I’m going to go about doing it.

**
01:09:50**
Daniel Englebretson
This isn’t it, deciding how to do this. This is what we just broke down in the other screen of.

**
01:09:56**
Daniel Englebretson
Here’s how you break down how you’re.

**
01:09:58**
Daniel Englebretson
Going to go do this.

**
01:09:59**
Daniel Englebretson
And here’s what we want you to be checking for as you do that.

**
01:10:02**
Daniel Englebretson
And so off it goes.

**
01:10:05**
Daniel Englebretson
So flipping back to my screen over here so that we can speak to. Here we go. All right, so here is my workflow.

**
01:10:26**
Daniel Englebretson
So.

**
01:10:27**
Daniel Englebretson
So for the.

**
01:10:28**
Daniel Englebretson
For people who are more familiar with working with ChatGPT or Copilot, let me show you how.

**
01:10:35**
Daniel Englebretson
How this starts to come to life in a smaller way so that we can start making changes and reacting.

**
01:10:43**
Daniel Englebretson
Right now, I’m just hooked up to the data. You could do this with teams. We’re over here on the right is teams, and it’s sitting on top of.

**
01:10:49**
Daniel Englebretson
All your data and your systems and your processes.

**
01:10:53**
Daniel Englebretson
Imagine that this is just teams because.

**
01:10:55**
Daniel Englebretson
This is totally feasible in teams. And I just say, read this document, then explain it to me. It’s just going to read my RFQ and explain it back to me.

**
01:11:08**
Daniel Englebretson
Now, my ChatGPT or my teams or my whatever is reading this Safety shower workflow. If were going to build this and deploy this on a stack that you were just going to run with, you wouldn’t have to tell it to go ground. And then it would know that this is the process that we’re going to go do. It’s going to read these documents and it’s going to start coming back and saying, okay, well this document is related.

**
01:11:32**
Daniel Englebretson
To these things, it’s got these requirements and so on. And so we’ll let it kind of answer. And then just so you can see how this plays out, I think it’s important to understand how it plays out. If I then hit all right, we’ve got these three documents, so we’re going to give it this one, and this one.

**
01:11:53**
Daniel Englebretson
So these are the standard works for the flow.

**
01:11:58**
Bhumika Sachdev
So even if the team says like you don’t have copilot license, will it still read it or it won’t.

**
01:12:06**
Daniel Englebretson
You will need to have a copilot license in your stack to be able to do this.

**
01:12:10**
Daniel Englebretson
If you want to do it in.

**
01:12:11**
Daniel Englebretson
Teams, you don’t have to do it in teams. I just know we’ve talked about doing in teams. So when we get through today, and.

**
01:12:17**
Daniel Englebretson
I realize this is a long setup.

**
01:12:18**
Daniel Englebretson
For this, we’ll have more context on like what would be the right place to do this, you know, but it.

**
01:12:27**
Daniel Englebretson
Could be done in teams. But if for you to be able.

**
01:12:29**
Daniel Englebretson
To do it in teams, you would.

**
01:12:30**
Daniel Englebretson
Have to have a copilot license for the user who’s using it, which carry.

**
01:12:35**
Daniel Englebretson
So there’s two ways you could do.

**
01:12:36**
Daniel Englebretson
That just from a cost perspective.

**
01:12:37**
Daniel Englebretson
You can have a per user cost.

**
01:12:39**
Daniel Englebretson
Of like $30 a month, or you can do what’s called pay as you go, where it just eats credits up.

**
01:12:44**
Daniel Englebretson
And when we’re done today, I can.

**
01:12:46**
Daniel Englebretson
Show you how much money we spent running this process. I expect it’s going to be, you know, if I was going to swag it’s probably going to be something like $10 when we get done here. But I’ll be able to, I can show you that.

**
01:12:59**
Daniel Englebretson
The workflow. So what I’m demonstrating on the right, before we get into it here, is imagine you had standard works that were.

**
01:13:09**
Daniel Englebretson
Written for how you process a workflow.

**
01:13:12**
Daniel Englebretson
And that standard work or that guidance or SOP that you have is this.

**
01:13:16**
Daniel Englebretson
Safety shower RFQ workflow at the highest level, right?

**
01:13:20**
Daniel Englebretson
So if you were to come in here and say, you know what, we.

**
01:13:22**
Daniel Englebretson
Actually need to insert another step because there’s a step between RFQ intake and triage requirements analysis. We would enter that step and now you have a different step.

**
01:13:31**
Daniel Englebretson
Or if you were to come in.

**
01:13:33**
Daniel Englebretson
Here and say, you know what, when we break this down into the incremental steps that we want to do, we actually want to change it so that there’s one more step in here. So that’s what we have here.

**
01:13:42**
Daniel Englebretson
And a human could follow the process.

**
01:13:44**
Daniel Englebretson
Or a bot could follow the process.

**
01:13:46**
Daniel Englebretson
So it’s just reacted to this and it said, okay, here’s all these things.

**
01:13:50**
Daniel Englebretson
So just to get us one more.

**
01:13:52**
Daniel Englebretson
Prompt here so you can see where we’re going with this now I’m going.

**
01:13:55**
Daniel Englebretson
To Say your task is to execute the RFQ workflow. Start with CF01 and I’m just going to add CF01 and then we’ll talk about it. All right?

**
01:14:16**
Daniel Englebretson
So I needed to get it to this point so you guys could see why all these components matter and how.

**
01:14:21**
Daniel Englebretson
We start to make changes to it.

**
01:14:24**
Daniel Englebretson
In this case, the bot that’s doing the work with your collaborator in this, if you will, it just read through the standard works of what’s the RFQ workflow? What’s the MVP of this overall process? What’s the ideal sequence we should run it in? What should I be paying attention to it? Read those documents, which are living documents that you could change at any point. So if you learn something, you said, you know what, we’re going to change this, or we’re going to change the sequence of how we’re going to do it, or we’re going to change what the checks are. Right there, it’s changed and off you go. So it’s like, okay, we’re going to do it. So it says, okay, well, step one is I need you to answer the, I need the data that’s going to tell me this. What’s the client?

**
01:15:07**
Daniel Englebretson
What’s the blah, blah.

**
01:15:08**
Daniel Englebretson
So that’s kind of, it’s just queuing.

**
01:15:10**
Daniel Englebretson
Up so they can actually do the first step.

**
01:15:13**
Daniel Englebretson
So I’m just going to give it.

**
01:15:14**
Daniel Englebretson
The example that you guys gave me.

**
01:15:16**
Daniel Englebretson
On the front end so you can.

**
01:15:17**
Daniel Englebretson
Start to see how this unfolds and then we can start speaking to it. So in the discovery artifacts, you guys gave me this adnoc example.

**
01:15:28**
Daniel Englebretson
So if I just take the examples that you gave me, so these are.

**
01:15:31**
Daniel Englebretson
The different examples which you also should be able to see. And I give it to it here, just kind of giving it these examples.

**
01:15:40**
Daniel Englebretson
So if you were going to do this in a non chatgpt way, you.

**
01:15:44**
Daniel Englebretson
Were going to do this like in a teams bot.

**
01:15:46**
Daniel Englebretson
Instead of having to come in here.

**
01:15:47**
Daniel Englebretson
And add all these things like I.

**
01:15:49**
Daniel Englebretson
Just did here, you could just point it at a client folder. Go look at this folder. Go look at this RFQ folder and.

**
01:15:55**
Daniel Englebretson
All these files that are in the RFQ folder.

**
01:15:57**
Daniel Englebretson
These are the files, right? I’m just telling it which files to look at. But that, so that you can see that here.

**
01:16:03**
Daniel Englebretson
But the way that the black screen.

**
01:16:05**
Daniel Englebretson
Is doing it is it just knows.

**
01:16:06**
Daniel Englebretson
To go look at the folder.

**
01:16:08**
Daniel Englebretson
So we’re going to say use this.

**
01:16:10**
Daniel Englebretson
Context as the example, then continue.

**
01:16:15**
Daniel Englebretson
All right, so we’re getting to a point. Where we can actually start picking this.

**
01:16:18**
Daniel Englebretson
Thing apart so that you can start to weigh in on this.

**
01:16:21**
Daniel Englebretson
So.

**
01:16:24**
Daniel Englebretson
Let me see.

**
01:16:25**
Daniel Englebretson
So while this one is kicking off.

**
01:16:27**
Daniel Englebretson
I’m just going to flip back over so you can see how the other one’s going, so you can pace with me on this.

**
01:16:32**
Daniel Englebretson
So we gave us some feedback. It came back and asked us some.

**
01:16:34**
Daniel Englebretson
Questions and we said, no, here’s my answer. So it went back through the materials and starts making changes.

**
01:16:40**
Daniel Englebretson
It went back to the compliance matrix.

**
01:16:42**
Daniel Englebretson
And made some changes, you know, and.

**
01:16:44**
Daniel Englebretson
Said, all right, here’s how we edit.

**
01:16:46**
Daniel Englebretson
It based on your feedback. You know, here’s kind of the end. So it’s saying, all right, we locked in this chiller.

**
01:16:52**
Daniel Englebretson
We locked in this thing.

**
01:16:55**
Daniel Englebretson
The weakest link flag remains true. You know, so this is the evidence. So if you find deviations, basically it’s. It’s logging them, updated the RFI so what other information it needs, and so on.

**
01:17:08**
Daniel Englebretson
So this process that it’s kind of flowing through it, read all these different.

**
01:17:14**
Daniel Englebretson
Documents and move through the process of doing it.

**
01:17:18**
Daniel Englebretson
So again, if you were going to fully automate the flow, that’s what would.

**
01:17:23**
Daniel Englebretson
Be happening on the back end.

**
01:17:25**
Daniel Englebretson
It’d be following these processes of doing.

**
01:17:27**
Daniel Englebretson
That on the back end.

**
01:17:29**
Daniel Englebretson
But on the front end, as a human in the loop, who’s thinking about this and not automating this, your interface might be more like something that’s giving you flags back of, hey, you know, do you agree that we shouldn’t continue.

**
01:17:43**
Daniel Englebretson
With this because we don’t have all the information that we need?

**
01:17:46**
Daniel Englebretson
Do we need to consolidate a new.

**
01:17:48**
Daniel Englebretson
Request for information based on these findings? So I use this term probably the wrong way here. This is where it’s coming back, asking you for information.

**
01:17:58**
Daniel Englebretson
But. And then, and then this is where it gets interesting. It links back to all of the things that it checked. So if you need to go check.

**
01:18:04**
Daniel Englebretson
Okay, what was the material requisition for.

**
01:18:06**
Daniel Englebretson
Emergency safety sours and eyewashes? Well, boom, it’s giving you that.

**
01:18:11**
Daniel Englebretson
It’s kind of ugly in this system because I had to parse it from your PDF.

**
01:18:14**
Daniel Englebretson
So we’re going to say we’re going.

**
01:18:16**
Daniel Englebretson
To just tell it so you can see what happens here. One, I agree, but let’s continue for the demo and to proceed with the draft.

**
01:18:33**
Daniel Englebretson
It’s taking these actions based on the guidance that it’s given, based on the documents that it’s referencing, based on the process that it’s following. So if you were doing this in teams or In a bot or whatever, you’d be able to click on these things and go to those files. You say, oh, okay, well, let me go look at this file, or let me go look at this thing.

**
01:18:54**
Daniel Englebretson
And so on here.

**
01:18:56**
Daniel Englebretson
It’s doing what we asked it to do and it came out with our RFI draft. So, hey, safety showers, Team Simon. We need these things. I need this and I need this.

**
01:19:07**
Daniel Englebretson
And I need this in order to be able to do this, to do this.

**
01:19:13**
Daniel Englebretson
I will leave this on the screen here for a second, but this is where you start getting into.

**
01:19:19**
Daniel Englebretson
And then I can take a breath and we can look at this and get into it.

**
01:19:23**
Daniel Englebretson
How did it know, how did it know to ask these questions? How did it know that it was missing, you know, this information? How did it know that it shouldn’t proceed with the task because it doesn’t have the information that it needs to.

**
01:19:38**
Daniel Englebretson
Proceed with the tasks?

**
01:19:39**
Daniel Englebretson
Right up here it came through and.

**
01:19:42**
Daniel Englebretson
It kind of went through it and says, hey, I went through everything.

**
01:19:45**
Daniel Englebretson
But we shouldn’t move forward because the.

**
01:19:48**
Daniel Englebretson
Quantities of high level hazard errors, rating period of line, there’s significant blockers, uncertainties for any moving forward with the CF02 requirements, you know, missing these documents.

**
01:19:57**
Daniel Englebretson
So if the proposal engineer was sitting on top of this system and kicked it off and pointed it at the folder and went through the whole thing and said, go look at this folder and tell me if we have everything that we need and tell me what we don’t need. It will go look at that folder, follow your standard work and say, okay, I looked at this, this, this, here’s what you don’t have that I think you need. What do you want to do basically? And you could stop here and go do other things, or you could keep going. But one thing I want to call out in this is you saw how fast it went through this. One of the big changes that AI brings you is the speed, the pace at which you can run.

**
01:20:40**
Daniel Englebretson
So if you let this thing run all the way through to the end, which I’m not going to do now, and see what the result is and see what the findings were, you could do that in the space of a few minutes and then take kind of step back and say, all right, well, here’s how this whole flow went. Do I want to rerun it? Do I want to give it different context? Did I see something I didn’t realize? And so if we’re going back to the value drivers here is something buried. Are there risks to the Business. Is this a really complicated thing? We miss something? There’s just a lot of volume. Well, what’s happening in this case? So if I look at this flow.

**
01:21:19**
Daniel Englebretson
So I’m going to go back to my safety shower workflow here.

**
01:21:24**
Daniel Englebretson
If I look at this flow that we’re running, I also can see all of the other documents in the system that are related to the document I’m looking at. If I take my example quote here, I can see all of the other documents in our system that were related to this document. I don’t know what all these things.

**
01:21:42**
Daniel Englebretson
Are, but you know what these things are.

**
01:21:44**
Daniel Englebretson
That is all me queuing up. What would an interface actually give you and how would you lock in what it’s doing? You’re doing it by specifying, you know, what is the actual flow that you want it to run.

**
01:22:02**
Daniel Englebretson
So back to the safety shower rfq and how do you want to break it down?

**
01:22:06**
Daniel Englebretson
And so the question I’ll tee up.

**
01:22:09**
Daniel Englebretson
To you as we start to pick.

**
01:22:11**
Daniel Englebretson
This part is I would never put.

**
01:22:15**
Daniel Englebretson
Something like this in production on day one and expect it to be right. Right?

**
01:22:19**
Daniel Englebretson
I just never. But after 10 or 15 or 20.

**
01:22:23**
Daniel Englebretson
Iterations of it, which I think I did seven or eight, or actually I.

**
01:22:26**
Daniel Englebretson
Think I did nine already, and you ask it to look back and what did you miss and what did you flag?

**
01:22:31**
Daniel Englebretson
It starts learning and building. And you might have seen it in the trial runs down here there’s a quality check folder.

**
01:22:38**
Daniel Englebretson
And in the MVP updates, you can.

**
01:22:41**
Daniel Englebretson
See, okay, it went through these different episodes, that’s what I call them, and.

**
01:22:45**
Daniel Englebretson
It found these different things.

**
01:22:47**
Daniel Englebretson
And as a result of finding these.

**
01:22:49**
Daniel Englebretson
Things, it’s going back and saying you need to update these different standards in.

**
01:22:54**
Daniel Englebretson
Order to be able to refine the system.

**
01:22:58**
Daniel Englebretson
And so what I would likely advocate.

**
01:23:00**
Daniel Englebretson
For as you start thinking about deploying something like this is you’d run it.

**
01:23:03**
Daniel Englebretson
In parallel to your processes today and you would accrue learnings and you’d have a feedback loop.

**
01:23:10**
Daniel Englebretson
And I recommend doing it on a.

**
01:23:12**
Daniel Englebretson
Weekly basis that checks what was learned and what was changed and approves those.

**
01:23:17**
Daniel Englebretson
Changes as you’re going through the different cycles until you get to a level of quality that you want.

**
01:23:24**
Daniel Englebretson
So teeing the question up for today.

**
01:23:26**
Daniel Englebretson
Then, and what we’re going to spend most of the rest of the time.

**
01:23:29**
Daniel Englebretson
On is as you start breaking these.

**
01:23:31**
Daniel Englebretson
Chunks down, we have to start really.

**
01:23:34**
Daniel Englebretson
Understanding, you know, what is the goal.

**
01:23:37**
Daniel Englebretson
For the chunk, what are the steps for the chunk? What are the standards that you need.

**
01:23:41**
Daniel Englebretson
And how, and how far do you need to take that before you have.

**
01:23:46**
Daniel Englebretson
A system that starts to be valuable to you? You know, that you can start to work from at the level of granularity that you need.

**
01:23:54**
Daniel Englebretson
So I realize it took me like 30 minutes to go through all of that. I’m going to pause here because I know it’s a lot and just kind of get your top of mind thoughts as we’re teeing this up and pointing.

**
01:24:07**
Daniel Englebretson
Us in the direction of, you know, what assumptions do we need to be.

**
01:24:10**
Daniel Englebretson
Checking, you know, what value drivers.

**
01:24:13**
Daniel Englebretson
Are we looking for now that you’ve kind of got a sense of how this might would flow to tailor this further? I guess I even start with, are you following what I’m up to here? And do you have any like structural questions about what we’re up to here?

**
01:24:36**
Simon Walmsley
No, I can certainly, I can see the value by having these sort of defined processes. So you’re sort of running step by step. And each of those steps is almost like a document in its own right, which you can then tweak and build on and strengthen over time. I think like what we touched on the previous session and what Steve said at the start there was this all kind of kicked off from one example where we missed a key piece of information buried in a very big rfq. It was about the bolting grade. And you know, we talk about lessons learned and you know, right. We’ve got to make sure we don’t get caught out by something like this again.

**
01:25:21**
Simon Walmsley
And you know, it’s all very good saying that, but everybody in the sales team, you could have somebody new who’s joined the business and they’re not going to know about a quality issue that happened three years ago that cost us a lot of money. Whereas if you can build these things in and document them as these are our, you know, I think I saw something earlier where it was like some key hazards or things that it’s going to be running through the documents and checking. Well, you know, you put that in, you know, bolting grades or material specifications or something and then that’s, it’s a documented process that the AI isn’t going to miss or skip over, whereas human error would allow that.

**
01:26:07**
Simon Walmsley
So I think that’s a really good thing that it’s, you know, by constant use, it’s, it gets stronger and more, more thorough the more it’s improved upon.

**
01:26:20**
Georgia Hogg
Just a question. Obviously we use systems like SAGE for our bill of materials and not saying it should link up to sage, but is there a way of almost having these documents to compare to what our scale standards are? So it can flag. Oh, the you. Our bolts are typically a 2. This is saying a 4. It’s flagging, but it’s comparing back to our typical. And whether we then move forward and deviate and say, oh no, this is our standard, but it’s referenced here and we’re kind of flagging it to the client to that we won’t be supplying it or not. But is there a way of calling back to what we do? Because especially for me, if I’ve got material or electrical specs, I can read through it, but I would then have to go speak to and go, right, what’s this?

**
01:27:05**
Georgia Hogg
What’s the norm? Especially on electricals? Okay, what’s our, what would be our standard gland? What would be our standard terminal? What’s the material? You know, is this a big deal? Is it not? So yeah, just wondering about that as a comparison tool as well. Why does this function?

**
01:27:23**
Daniel Englebretson
Yes, yes. So that’s a phenomenal question. So the risk of getting technically nerdy here, one of the cool things that you can do with AI is what’s called semantic search. And semantic search is where you’re searching through a whole bunch of data based on meaning, not based on the word itself. But then there’s also what’s called full.

**
01:27:44**
Daniel Englebretson
Lexical, which is like going to Google and typing in some words and getting a response which is keyword based. Right.

**
01:27:50**
Daniel Englebretson
But then there’s a third thing which.

**
01:27:51**
Daniel Englebretson
Is relationships of the data.

**
01:27:54**
Daniel Englebretson
I know all of these things are related to each other because this safety.

**
01:27:59**
Daniel Englebretson
Thing is related to this product, this customer and so on.

**
01:28:02**
Daniel Englebretson
If you think about going to Google and typing in bars in Amsterdam and it gives you answer back, well, in this case the AI, you can type in a question like what you just asked about with standards and pre existing or whatever and it can look across all of the data that it can see into and it can search, okay, how is this query related to all these things? But also what is semantically related to this? So you talked about, I’m going to make this up explosive environment. So it’s going to look at everything related to that, but then also keywords. So if you use a specific spec, it’s also looking at that. The reason why I want to call that out here is we just created this example over here and it did some stuff over here.

**
01:28:41**
Daniel Englebretson
Here I am, I got this new.

**
01:28:42**
Daniel Englebretson
Thing working with an AI system. If I hit refresh up here, it’s now going to re index the file based on this and you might notice that the stuff just popped up. What just happened is it now looked across all the data has access to and it found all of the other places in the system that’s related to this. This might not be exactly what you were just asking for, but if you knew you were looking for a certain standard or you knew this message just came back from the system and it says you need to pay attention to.

**
01:29:15**
Daniel Englebretson
Ingress protection, IP rating, you know, and.

**
01:29:17**
Daniel Englebretson
You want to go look for that, it can go find anywhere in the data that those things exist. And it can do that based on those three criteria. But also if there are high priority.

**
01:29:29**
Daniel Englebretson
Things that you wanted to flag. So for example, coming back over to.

**
01:29:33**
Daniel Englebretson
Our high risk keywords, if you have high risk keywords and you turn these into, well the way you do them, they’re called tags here. But if I turn it into a tag, so if I just take for.

**
01:29:47**
Daniel Englebretson
Example, it doesn’t matter, I’ll just pick one, I’ll take IIC as a tag.

**
01:29:55**
Daniel Englebretson
Once I’ve tagged that, I can now search anywhere that has that tag too. And so that’s where you start getting into what would you want to tag and why. And that would be a great use case for high priority stuff. And so, and so like taking a.

**
01:30:13**
Daniel Englebretson
Step back on this then to kind of show you what I mean with relationships and that.

**
01:30:19**
Daniel Englebretson
You don’t necessarily need to see this view. But if you’re in here looking at something like I’ve got this Q32859 NMDC proposal, I can see this and see all those things that are related to it.

**
01:30:37**
Daniel Englebretson
And so can the bot too.

**
01:30:38**
Daniel Englebretson
That’s what I mean about being able to work backwards on it. But no one has to go put that link. That’s the thing with the AI. Nobody has to actually go link that. You just have to know that you want to link that. You have to know that you want.

**
01:30:51**
Daniel Englebretson
The answer to that question.

**
01:30:52**
Daniel Englebretson
And as soon as you specify that.

**
01:30:54**
Daniel Englebretson
You want the answer to that question.

**
01:30:55**
Daniel Englebretson
It could do it across the entire database. And so the way that you start thinking about specifying that is getting into okay, what would high risk be?

**
01:31:05**
Daniel Englebretson
What would a checklist be?

**
01:31:06**
Daniel Englebretson
You know, what is the past.

**
01:31:08**
Daniel Englebretson
Deviations library that I’m going to go.

**
01:31:10**
Daniel Englebretson
Look up, you know, and so it really just comes down to back.

**
01:31:14**
Daniel Englebretson
To the critical thinking of what would.

**
01:31:15**
Daniel Englebretson
You need to know and why? And so Partly you already know that from your experience, especially you know.

**
01:31:23**
Daniel Englebretson
Simon, all that you’re doing.

**
01:31:25**
Daniel Englebretson
But also every time the system runs, if you include a retrospective look at.

**
01:31:31**
Daniel Englebretson
That, if I come in here, my.

**
01:31:34**
Daniel Englebretson
Quality checks and you include a retrospective look at that, it’s going to say, oh man, you missed all these things. Here’s how we should change it next time. And then the expert can say, oh yeah, we need to commit that change. Every time this thing runs and you learn something, you commit the change, it knows to look for it next time. That’s how over time you start getting into a direction where it knows what to catch. That’s also why I say you wouldn’t do this on day one. You wouldn’t roll out a system on day one expecting it perfect. But what you do need to roll.

**
01:32:08**
Daniel Englebretson
Out is the scaffolding to be able.

**
01:32:10**
Daniel Englebretson
To catch all the stuff that you.

**
01:32:11**
Daniel Englebretson
Want to be able to catch.

**
01:32:15**
Daniel Englebretson
Then to be clear, everything that we’ve just looked at can be done on the Microsoft stack in a secure environment, you know, on a pay as you go model. And it’s not particularly expensive to do. So it’s really not. This isn’t a deep technical cost, it’s more of a critical thinking cost of.

**
01:32:36**
Daniel Englebretson
What do we want and why do we want it.

**
01:32:37**
Daniel Englebretson
So I know that’s a long winded answer, but that’s how you.

**
01:32:41**
Daniel Englebretson
Start to be able to do those kind of surface, those kinds of things. Yes, Stephen, Yes.

**
01:32:47**
Stephen Hudson1
So I think to George’s question as well that we’d be able to upload our standard specifications like a standard safety shower for it to cross reference bikes. I think what George is asking is if we use a specific bolt and it’s calling that out.

**
01:33:10**
Daniel Englebretson
That’S where we’ve.

**
01:33:11**
Stephen Hudson1
Got a gap in the specification. I think we played with that as well. I think what can do Georgia is refer back to drawings or data sheets. This is our normal and this is where the variance lies. I think that’s, that’s another way of doing it as well.

**
01:33:25**
Georgia Hogg
Yeah. So it’s like for example, if there’s a pack of, you put the quote together and kind of put in all your sub component data sheets that we might have like switches, junction boxes, chillers that it. Yeah. Like you say can, it can look at the RFQ docs and then call back to what our potential quote is and then if we need to modify that especially. So like here’s what our standard scope supply would be to kind of meet the overall. But are there Tweaks that then need to be made to better match the rfq.

**
01:34:01**
Daniel Englebretson
Yeah. In fact, even as you’re saying that actually might would be the highest.

**
01:34:07**
Daniel Englebretson
Value, lowest lift thing to do first, because you’re not asking it to make the proposal, you’re asking it to check your work and you know what you’re checking for. So if you worked backwards on this process and you said let me deploy a workflow that checks the final product.

**
01:34:24**
Daniel Englebretson
Against these like thousands of pages and.

**
01:34:26**
Daniel Englebretson
Flags any potential misses or flags like the client asked for this, but actually.

**
01:34:31**
Daniel Englebretson
That or whatever that would be all of the things we looked at are feasible, but that would be the most feasible because it’s not making any decision.

**
01:34:40**
Daniel Englebretson
So let me put this in a slightly different frame so that it’s easier.

**
01:34:47**
Daniel Englebretson
So the reason why I say that is because the question of go double check this for me is a level one.

**
01:34:54**
Daniel Englebretson
Whoops.

**
01:34:55**
Daniel Englebretson
Is a level one question that is not hard for AI to do that and so it’s a much safer bet. But the question of what should my RFQ be is a level four question because it’s got to go through all this stuff and predict, which is a much harder question. Not that you can’t get there, it’s.

**
01:35:13**
Daniel Englebretson
Just harder and riskier and going to.

**
01:35:14**
Daniel Englebretson
Take more time to get there. So anything down here at the bottom is easier for the AI to do first. So as we’re thinking about what we’re looking for, but yes, that is a phenomenal use case. And so if you go back into what we’re doing today and I’m glad we’re recording and you think about the process, you would say, all right, well in this flow, at the end when you’re doing proposal handoff, so we.

**
01:35:39**
Daniel Englebretson
Go all the way down to number four down here at the end.

**
01:35:43**
Daniel Englebretson
All right, what are we doing as valuable? We’re going through all these discrete technical.

**
01:35:46**
Daniel Englebretson
Decisions into a persuasive package.

**
01:35:49**
Daniel Englebretson
Why is it valuable? So the sales team can get trustworthy.

**
01:35:52**
Daniel Englebretson
Technical foundation, so on.

**
01:35:53**
Daniel Englebretson
So how are we going to do that? We’re going to do these things? Well, if I came in here and add a step down here for step five and it said go look for the kind of thing that you just flagged. Now that’s baked into the process and.

**
01:36:07**
Daniel Englebretson
It knows to go look for the kind of thing that you just flagged.

**
01:36:09**
Daniel Englebretson
So I’m not as aware of how.

**
01:36:11**
Daniel Englebretson
This actually works for you. So this might not make total sense.

**
01:36:14**
Daniel Englebretson
But that’s how this type of thing would happen. And so what happens is the expert sitting here, you make one change here, and every time it runs, it now has that, it now knows to look for that. Right. The point that I want to then make on top of that, because I think we’re headed in the same direction with this, is you don’t even have to be the expert to ask the question. You can just say, go back and check your work based on everything you’ve seen before and validate it. Somebody needs to spec this on the front end. But having the AI just do a quality check and go back through here and be like, yeah, here’s how we made these decisions and here’s what it was connected to and here and so on. This is a specification.

**
01:36:59**
Daniel Englebretson
It didn’t decide to do this gate.

**
01:37:01**
Daniel Englebretson
You gave it that gate. Right?

**
01:37:03**
Daniel Englebretson
So. So that’s where like feedback loops and like a committee of people thinking about this process.

**
01:37:08**
Daniel Englebretson
Improving it gets it smarter and smarter. And so when your SMEs are working.

**
01:37:12**
Daniel Englebretson
With the tech in the platform that.

**
01:37:15**
Daniel Englebretson
Might would look something more like this.

**
01:37:18**
Daniel Englebretson
And it’s running through it and it’s giving you answers back, it will then say, hey, per step number three or.

**
01:37:25**
Daniel Englebretson
Whatever is in here.

**
01:37:26**
Daniel Englebretson
Here’s, here’s.

**
01:37:27**
Daniel Englebretson
Let me see if I can find it. So it’s coming up here. Let’s just go up to the top where I actually did it. Yeah. You see how it’s kind of going through. I will now execute step two. So here’s step two. So here’s how I’m going to do it. I’m going to go through, I’m going to look at this, I’m going to look at these documents. I’m realizing we’re missing this.

**
01:37:48**
Daniel Englebretson
So what happens once you change that.

**
01:37:49**
Daniel Englebretson
Spec is it just starts to show up in the chat over here and then the SME who’s working on it.

**
01:37:54**
Daniel Englebretson
Will see, oh, it just did this.

**
01:37:57**
Daniel Englebretson
Step and here’s what it found, you know, but it has to have that step to do that thing.

**
01:38:00**
Daniel Englebretson
So hopefully. I’m connecting the dots on this.

**
01:38:03**
Daniel Englebretson
I know it’s a little bit abstract.

**
01:38:04**
Daniel Englebretson
But does that answer your question?

**
01:38:08**
Georgia Hogg
Yeah, it does. I think from our end, there’d probably be limitations in the example of assembly bolts. We don’t have that detail within a quote or on our data sheet. There’s sort of that sort of more minute detail that we don’t typically share, but maybe it’s something that if we pulled a bill of materials for a standard tank shower, that it could be there. Simon I didn’t. What do you think on that one? Almost like have go run through, put together a quote of tank chiller, alarm proxies, junction box, have a file put together and then work it backwards. I don’t know if you’ve had a go at that before.

**
01:38:50**
Simon Walmsley
Yeah, yeah, how that’s how that is formatted and inputted into this workflow. But yeah, I think it’s a good point of listing almost like a library of what is our standard. And like Georgia said, you know, if were to select any electrical equipment, we know that our standard material of the cable plans is nickel plated brass. We wouldn’t explicitly say that in our quotation because otherwise it, you know, it would become very complex and long the quotation. But we just know our standard is nickel plated brass. And if we receive an RFQ and something in there says all cable glands must be 316 stainless steel, well there’s a flag to say, right, you need to look at this doesn’t meet your standard. Or yeah, the bolting grade, Our standard is A2, which is 304 stainless steel.

**
01:39:49**
Simon Walmsley
And again there could be something in an RFQ that says all bolting must be grade A4 which is 316. So again, flag that up as that’s not a mismatch look at. And then yeah, the more we build on what is our list of standards, the more information we’ll get back if something flags that it’s not quite there.

**
01:40:13**
Daniel Englebretson
Yes, that’s exactly right. Yeah, that’s exactly right.

**
01:40:18**
Daniel Englebretson
It’s updating the standard to know, to check for that in a cross reference so that next time it knows to.

**
01:40:22**
Daniel Englebretson
Check for that and the data has to be somewhere for it to be able to access it. But if you have bills of material that say that in your system, it goes back to, well, it could do the semantic search. And so if it knows it’s looking.

**
01:40:37**
Daniel Englebretson
At the material, like the metal material, that’s one of the flags, it can.

**
01:40:42**
Daniel Englebretson
Go look for any place in the.

**
01:40:44**
Daniel Englebretson
Linked contextually relevant information that references that and surface it.

**
01:40:49**
Daniel Englebretson
And so it goes back to what.

**
01:40:53**
Daniel Englebretson
Question are you asking of what data and why?

**
01:40:55**
Daniel Englebretson
And as long as you know, okay, I need to ask the question of is there a metal concern here and where do I ask that of.

**
01:41:03**
Daniel Englebretson
The bill of material and why do I ask that?

**
01:41:05**
Daniel Englebretson
Because it’s not in the quote and it’s only in the bill of material.

**
01:41:08**
Daniel Englebretson
And it’s we’re going to get zinged if we don’t Know that.

**
01:41:10**
Daniel Englebretson
Well, once you have that, then it can become a spec, you know, and so it just, it just. Somebody has to think of that on the front end. So again, that’s actually where I go.

**
01:41:19**
Daniel Englebretson
Back to the beginning where I was talking about, we’re not trying to replace a human. We’re trying to let you be more creative.

**
01:41:23**
Daniel Englebretson
That’s what I mean by creative. That’s in a creative example. You’re thinking, connect this dot to this dot. And now I have a solution, you know, and so that’s what we’re looking for.

**
01:41:31**
Daniel Englebretson
So that when this thing comes back, you’re like.

**
01:41:34**
Daniel Englebretson
And you do a retrospective. You’re like, oh, yeah, it. Instead of, instead of having the machine.

**
01:41:38**
Daniel Englebretson
So let me just look at this so that we can really hammer the point here.

**
01:41:41**
Daniel Englebretson
So when the machine comes back and it says, okay, I mean, I made these changes.

**
01:41:45**
Daniel Englebretson
Let me look at this one instead.

**
01:41:46**
Daniel Englebretson
I made these changes. And here’s what I think we need to do differently. You’re not just committing whatever the machine says. You’re saying, does that make sense? Like, you know, should we do that? You know, and so I’m trying to.

**
01:42:01**
Daniel Englebretson
Find a better example. So that’s the spin up. So if I come up here to.

**
01:42:04**
Daniel Englebretson
Where it finished and it’s coming back.

**
01:42:07**
Daniel Englebretson
Here and it’s saying, actually, let me go up a little bit further. I want to give you a good example so that I can really make the point here. I might go back to an earlier one. All right, so it’s coming in here and I’m asking it to do retrospectives. Sorry, veil contacts. And proceed back up here. I’m trying to find a good example so we can speak to this.

**
01:42:54**
Daniel Englebretson
So.

**
01:42:58**
Daniel Englebretson
I shouldn’t be doing this on the fly.

**
01:43:00**
Daniel Englebretson
So like right here where it’s asking.

**
01:43:02**
Daniel Englebretson
You know, seeking your feedback, approved power decision. Should I do this or should I do this or, you know, is signage lightbox in scope? If yes, confirm this or whatever.

**
01:43:13**
Daniel Englebretson
So what a human would do in this process is they would get to the end of a flow, they would see what the bot flagged, and the human would say, oh man, we don’t have a check for light boxes in our process. We should put a check for light boxes in our process. And then the human would be able to go in and put a check for light boxes in the process. Right. And then bam, every time it runs moving forward. You got that? And so I want to flag here then, because it’s a good point to flag okay, what’s the burden of that?

**
01:43:42**
Daniel Englebretson
Bamika kind of alluded to this earlier.

**
01:43:44**
Daniel Englebretson
How long did it take you to do this? What’s the burden of that? And who’s going to have to go do that? So we didn’t really talk about this because I wasn’t trying to go super deep on it today, but I have what I call a method grammar for.

**
01:43:56**
Daniel Englebretson
How you go about doing this.

**
01:43:57**
Daniel Englebretson
And so in the method grammar section.

**
01:43:59**
Daniel Englebretson
There is this thing called the Infinity.

**
01:44:01**
Daniel Englebretson
Method grammar, which explains how this whole thing breaks down. And the reason why I call this out is because once you find that with your AI, you say, AI, follow my method grammar, follow my standards, and go write the spec for this question. And it will write the spec for the question based on the grammar and then insert it into the spec for you so you don’t have to worry about how do I write a tote loop. Right, because that’s in the specific spec. You just tell it, go insert this into the spec. And it looks at the grammar, it looks at the frameworks, it looks at your heuristics, it looks at whatever other templates that you might have in here of how you’ve done it before, and it says, here’s your spec. And then the human says, agreed. And now you have respect.

**
01:44:47**
Daniel Englebretson
So the reason why I want to call that out is because the difference between using something like ChatGPT is that it’s not going to remember that you can’t do these iterative incremental cycles of.

**
01:44:59**
Daniel Englebretson
Continuous improvement, because it doesn’t remember that. Like, not really.

**
01:45:02**
Daniel Englebretson
But when you have a system built with semantic search, which you can do on the Azure stack and interface into that, and some standard works that you’re looking against now, you can start to make those changes. And so that’s where you start to.

**
01:45:13**
Daniel Englebretson
Get the compounding advantage over time.

**
01:45:15**
Daniel Englebretson
And it really comes down to a.

**
01:45:18**
Daniel Englebretson
Process of feedback loops, of like sitting on top of this, paying attention and just iterating over time. So that’s a long dive into how that plays out.

**
01:45:28**
Daniel Englebretson
But that’s how that plays out.

**
01:45:30**
Daniel Englebretson
And so coming back to our process here, rolling back up because we just finished number one, the intake and triage.

**
01:45:47**
Daniel Englebretson
Kind of.

**
01:45:49**
Daniel Englebretson
Or go ahead, Tamika.

**
01:45:50**
Bhumika Sachdev
All right, so just to break it down, like you just talk with.

**
01:45:56**
Daniel Englebretson
Doing.

**
01:45:57**
Bhumika Sachdev
This with Copilot, I just started asking the question, are just putting in a file, just evaluate this. Let me just try to like break down different steps. How, how is it gonna work?

**
01:46:12**
Daniel Englebretson
You’re breaking up a little bit.

**
01:46:14**
Daniel Englebretson
I didn’t Hear exactly what you said. Can you say it again? Or if it’s simple, throw it in the chat. It’s like a. Your mic is distorted.

**
01:46:22**
Bhumika Sachdev
Yeah. So can you hear me better now?

**
01:46:29**
Daniel Englebretson
It’s like a. It’s like a technical thing, but we can try again. Okay, sorry.

**
01:46:40**
Daniel Englebretson
I think you were, you.

**
01:46:41**
Daniel Englebretson
Were asking for clarification on what I just stated and I think you were asking, okay, let me say back to you, how would we operationalize this for ourselves? And you’re going to put this in the chat and see if I was right. But if I’m on the right track here, if you go back to what’s my goal for this workshop and what’s my philosophy? It’s grassroots innovation. It’s teaching you how do you innovate your processes and what frameworks do you need to innovate those processes on top of this tech?

**
01:47:10**
Daniel Englebretson
So it’s one part what tech and one part what process.

**
01:47:13**
Daniel Englebretson
And so the process is this idea.

**
01:47:15**
Daniel Englebretson
Of breaking it all the way down.

**
01:47:16**
Daniel Englebretson
To these questions to ask of the data, which is what we’re observing here.

**
01:47:20**
Daniel Englebretson
Yeah.

**
01:47:21**
Daniel Englebretson
Yes, from my understanding, I am trying to break it down. Yes, exactly.

**
01:47:25**
Daniel Englebretson
So the process is what I owe you in this is observing what you.

**
01:47:30**
Daniel Englebretson
Have told me and giving you what I think your process is as a jumping off point so that you can start to run cycles like what I’ve been doing here, to observe outcomes of those cycles, to think, oh, we need to make this change, we need to add this standard, whatever.

**
01:47:45**
Daniel Englebretson
But, but instead of having to do.

**
01:47:47**
Daniel Englebretson
That in the abstract, it’s already broken down, okay? I’ve got standard works, like references, I’ve got SOPs and I got templates like these are my components, right?

**
01:47:57**
Daniel Englebretson
I’ve got these kinds of documentations and.

**
01:47:58**
Daniel Englebretson
I’ve got to broke it down this way.

**
01:48:00**
Daniel Englebretson
So you don’t have to start from nothing.

**
01:48:02**
Daniel Englebretson
You start.

**
01:48:03**
Daniel Englebretson
What I’m trying to do is get.

**
01:48:04**
Daniel Englebretson
You started from a flow that’s logical.

**
01:48:06**
Daniel Englebretson
And grounded, that’s meaningful, that starts to give you value and speeding you up so that you can start to run.

**
01:48:12**
Daniel Englebretson
The loops for yourself.

**
01:48:13**
Daniel Englebretson
Because it’s the whole teach the fish thing and because AI and I can’t tell you what the important questions are, only you know, what the important questions are to ask. But what I can do and what AI can do is surface the information.

**
01:48:28**
Daniel Englebretson
For you so you can get it documented.

**
01:48:30**
Daniel Englebretson
So and so. So to that end, what I want.

**
01:48:34**
Daniel Englebretson
To give you as part of this.

**
01:48:35**
Daniel Englebretson
Pilot, you’re going to get this entire set of documentation and A flow and a path to using the flow, which.

**
01:48:47**
Daniel Englebretson
We will have to decide how you would want to deploy and whether you’d want to deploy, which is the other half of this workshop.

**
01:48:55**
Daniel Englebretson
And then get you into a rhythm where you can actually run it and start seeing what’s coming back. And what I’m trying to do on the front end is run every example that you can give me.

**
01:49:07**
Daniel Englebretson
So you gave me a bunch of examples.

**
01:49:09**
Daniel Englebretson
I’ll run every example through that. I can in the same way and log the deviations and log the refinements.

**
01:49:15**
Daniel Englebretson
And put it into those retrospectives.

**
01:49:18**
Daniel Englebretson
And you’ll be able to go back and see. When you ran this one, you learned this. When you ran this one, you learned this. And then what we want to be able to do is take one that you’ve already done, throw in the raw inputs, and say, build it, and then see if it builds it as well.

**
01:49:31**
Daniel Englebretson
As you built it.

**
01:49:32**
Daniel Englebretson
And then after time of doing this, and maybe we’re close now, and you start to see that it’s close enough, then you start to ratchet up your trust level of which pieces of this.

**
01:49:43**
Daniel Englebretson
Are you going to let it do?

**
01:49:45**
Daniel Englebretson
If you think about which pieces and you start breaking this thing down. The reason why I put this value chain, value pair at the front of each step is because I think what we’re trying to assess is what in this step is valuable and why is it valuable to the person who receives it. And then if that’s true, all right.

**
01:50:05**
Daniel Englebretson
How do we achieve that?

**
01:50:06**
Daniel Englebretson
And so you might learn, you know.

**
01:50:08**
Daniel Englebretson
What, this is actually slightly off. Or you might learn that you want to put a step in between.

**
01:50:12**
Daniel Englebretson
And the tighter you get on that.

**
01:50:14**
Daniel Englebretson
The smarter the system gets.

**
01:50:16**
Daniel Englebretson
But coming away from this conversation and coming away with this documentation you don’t like. I didn’t write any of this documentation. Nothing in this vault that I write, I didn’t write any of this. You guys gave me thousands of files that I just ripped it into the vault. What I did write are the standards and the process and the approach and.

**
01:50:40**
Daniel Englebretson
The guidelines so that it will write it, right?

**
01:50:42**
Daniel Englebretson
And then you’re going to have those, too. So everything that I can do, you can do, because you have the exact.

**
01:50:48**
Daniel Englebretson
Same standards and the exact same SOP that I use. So.

**
01:50:52**
Daniel Englebretson
So that’s. That’s how. And then. And then it becomes a question of.

**
01:50:55**
Daniel Englebretson
And I’m just trying to really answer your question here.

**
01:50:58**
Daniel Englebretson
Then it becomes a question of, okay, well, is there enough value at each step in the chain, you know, to.

**
01:51:07**
Daniel Englebretson
Push it to a one or a two or a three or four.

**
01:51:09**
Daniel Englebretson
That’s what, that’s what we try to answer as you get into it. And people like Simon are going to have to work with the tech and say, yeah, that was valuable. I actually did save time on that. Or we actually did and you start.

**
01:51:21**
Daniel Englebretson
To get more appreciation for like how hard you want to hit it.

**
01:51:25**
Daniel Englebretson
But Simon, you can probably imagine a world where if you just had the.

**
01:51:28**
Daniel Englebretson
Keys to this kingdom and you just.

**
01:51:29**
Daniel Englebretson
Had it, and you just had it.

**
01:51:31**
Daniel Englebretson
Alongside you, two years from now, this thing’s just humming for you and it’s made your life a lot easier.

**
01:51:36**
Daniel Englebretson
And that might be as far as it needs to go. Like it’s up to you. But then rolling that out across the organization, how do you roll it across? It’s the same. That’s why it’s the same problem. Goal, response, learn, breakdown. Doesn’t matter what flow you’re looking at.

**
01:51:50**
Daniel Englebretson
It’s the same mechanics. Right?

**
01:51:52**
Daniel Englebretson
That’s, that’s the other half of what.

**
01:51:54**
Daniel Englebretson
We’Re trying to do.

**
01:51:54**
Daniel Englebretson
We’re trying to teach the mechanics, but.

**
01:51:56**
Daniel Englebretson
Do it in the context of a workflow. And so if you were reflecting on this after this conversation, you would recognize that the process and structure applies to any flow. We’re just, we’re just applying it to this flow. So. And so hopefully that answered your question. Vamika.

**
01:52:14**
Daniel Englebretson
Did it, did it answer your question.

**
01:52:15**
Daniel Englebretson
Or do you want me to try that differently?

**
01:52:19**
Bhumika Sachdev
No, it did. It did. Thanks.

**
01:52:21**
Daniel Englebretson
Okay.

**
01:52:21**
Daniel Englebretson
Okay. All right. So it actually is a good opportunity to take a step back and going.

**
01:52:33**
Daniel Englebretson
Back into our site here. I hate how it’s splitting this up here.

**
01:52:37**
Daniel Englebretson
As we’re thinking about what we’re going.

**
01:52:38**
Daniel Englebretson
To go through now and we get into this method grammar and we look at these heuristics here. There’s this mapping what’s possible. And so we’ve kind of touched on this a little bit, but types of.

**
01:52:52**
Daniel Englebretson
Knowledge that you must have access to.

**
01:52:55**
Daniel Englebretson
And to do what kind of things.

**
01:52:57**
Daniel Englebretson
And so if you think about that, like your question about bill of materials and nickel plating, well, that knowledge has to be in the system for the bot to be able to use it.

**
01:53:06**
Daniel Englebretson
Right.

**
01:53:06**
Daniel Englebretson
And if it’s not in the system, it can’t use it. Conceptual knowledge, it has to be these dots need to be connected or the bot can’t do it. You know that you need to go look at the bill of materials to check that because you know that the data lives there. The System doesn’t know that. But if it’s written down and it.

**
01:53:24**
Daniel Englebretson
Knows, go look at bills of materials.

**
01:53:26**
Daniel Englebretson
Now, it knows that, it knows how.

**
01:53:28**
Daniel Englebretson
They’Re relate, they’re related.

**
01:53:29**
Daniel Englebretson
And then procedural knowledge, you know that the RFQ process and the CPQ process and all that is X. And if we get that written down.

**
01:53:36**
Daniel Englebretson
The bot knows it too, and off it goes.

**
01:53:38**
Daniel Englebretson
And then, and then with the knowledge.

**
01:53:41**
Daniel Englebretson
You can start to do these kinds of things, you know, so retrieve some facts or understand some things or apply some procedures.

**
01:53:48**
Daniel Englebretson
So even in our example earlier about email writing, right, like, one of the steps in here when you get to the end is what does the package.

**
01:53:55**
Daniel Englebretson
Look like that gets sent to the.

**
01:53:56**
Daniel Englebretson
Client at the end of the day? Well, if we specify the template for.

**
01:54:00**
Daniel Englebretson
What that package looks like and we.

**
01:54:02**
Daniel Englebretson
Break it down, which I did in.

**
01:54:04**
Daniel Englebretson
Here, we just haven’t got that far yet.

**
01:54:06**
Daniel Englebretson
It will then follow the process and write it for you, and then your work becomes.

**
01:54:12**
Daniel Englebretson
If I go into. Let me just find an example. So it’s not abstract, let’s actually take a different one. Here we go. This one’s a full one. So if I get into. Let me close some of these out so I can see better.

**
01:54:28**
Daniel Englebretson
So I get into a full one.

**
01:54:30**
Daniel Englebretson
You know what, I’m gonna switch my screen to the other one because it’s easier for me. All right, here we go.

**
01:54:39**
Daniel Englebretson
So if I get into one of these ones that ran all the way.

**
01:54:41**
Daniel Englebretson
Through and got me my answers. So pick this one. Yeah, here we go.

**
01:54:45**
Daniel Englebretson
So when it was done, that’s like, okay, here’s my technical handoff summary. So I’ve got, by the way, if.

**
01:54:53**
Daniel Englebretson
You’Re looking at my screen, you’re trying to find this in the website, if you use the search bar at the top and you type in like one of the keywords, it will find the document for you. You don’t have to, you don’t have to navigate through the whole tree and everything to find it. So you could take, you know, a technical handoff summary and put it in the chat and you’ll find it.

**
01:55:10**
Daniel Englebretson
It’s, it’s doing okay. Here’s my email draft.

**
01:55:14**
Daniel Englebretson
Okay, please find our submission for this. Here’s the highlights, here’s the deviations, here’s the attachments, whatever.

**
01:55:19**
Daniel Englebretson
So these are made up from templates.

**
01:55:24**
Daniel Englebretson
That are in your standard works.

**
01:55:25**
Daniel Englebretson
And so the question you’re trying to answer when you’re deciding do I want.

**
01:55:30**
Daniel Englebretson
To put time into a template for.

**
01:55:32**
Daniel Englebretson
My standard work is if your bot could Write your email for you at the end of this whole process. Would that save you enough time to.

**
01:55:39**
Daniel Englebretson
Make it worth it to you to go in there and document what a good email looks like?

**
01:55:43**
Daniel Englebretson
Right. And so as we’re thinking about the.

**
01:55:45**
Daniel Englebretson
Different chunks of work that come out. So for example, the client packet or this is kind of rolling up all the things or the spec crosswalk. So my spec crosswalk, you know, it went through and did my specification crosswalk. So this is just what I observed from what we saw.

**
01:56:05**
Daniel Englebretson
But you might say, you know what.

**
01:56:06**
Daniel Englebretson
We also need to include this other thing too in a spec crosswalk or whatever.

**
01:56:11**
Daniel Englebretson
And so the question we’re trying to answer as we’re going through this is first of all, do we have the data and the system to even.

**
01:56:20**
Daniel Englebretson
Be able to do it? So going back to our feasible impossible.

**
01:56:26**
Daniel Englebretson
Do we have the data?

**
01:56:27**
Daniel Englebretson
Do we have the concept, do we.

**
01:56:28**
Daniel Englebretson
Have the procedures and if not, do we need to document? So all of that is what’s in the graph and then what do I.

**
01:56:34**
Daniel Englebretson
Want to do with it?

**
01:56:35**
Daniel Englebretson
And so to answer the question of.

**
01:56:36**
Daniel Englebretson
What I want to do with it, there’s two things. One is friction. So we’re trying to eliminate friction in the process, which I know we spoke a lot of time about that at a workshop already. So bottlenecks and delays and rework. And the other is value.

**
01:56:50**
Daniel Englebretson
So how do I get value?

**
01:56:52**
Daniel Englebretson
And I break it into, you know, streamline, enrich, accelerate.

**
01:56:56**
Daniel Englebretson
So I call this out here to ground us back in the rfq. That’s how this whole thing is written.

**
01:57:02**
Daniel Englebretson
It’s saying, okay, here’s how we’re going to get value.

**
01:57:05**
Daniel Englebretson
Here’s the function, here’s the steps, and.

**
01:57:08**
Daniel Englebretson
Then down here it’s like, here’s how we’re getting value out.

**
01:57:10**
Daniel Englebretson
And so because it’s all written down, you can then just ask the AI, go look at my friction and my sea. Take this question and then.

**
01:57:21**
Daniel Englebretson
Help me figure out, is this worth doing or not? So there’s a, I have a step that we’re not doing today which writes the business case for you of like, should you put the time into it or not? But that’s not what were trying to do today.

**
01:57:33**
Daniel Englebretson
I know, I just hit a lot of concepts fast.

**
01:57:37**
Daniel Englebretson
Let me pause there and put back to you.

**
01:57:41**
Daniel Englebretson
Looking at what we’re doing here and how it’s flowing and thinking about this.

**
01:57:45**
Daniel Englebretson
As being a living thing over time.

**
01:57:48**
Daniel Englebretson
And having a feedback loop in place.

**
01:57:51**
Daniel Englebretson
Maybe that’s cross functional to look at.

**
01:57:53**
Daniel Englebretson
Okay? We ran 10 of these through what do we want to change in our spec? You know, and to answer the question of how much time is it worth.

**
01:58:01**
Daniel Englebretson
Putting into fixing this thing up?

**
01:58:03**
Daniel Englebretson
Well, that’s. How much time is it saving you?

**
01:58:05**
Daniel Englebretson
And how many flags is it throwing and all those kinds of things? And so the question I put to you as we’re thinking about this is what. What would you.

**
01:58:14**
Daniel Englebretson
What would you need to see in the data?

**
01:58:17**
Daniel Englebretson
What process improvements would you need to realize as you’re going through it to.

**
01:58:21**
Daniel Englebretson
Make it worth it?

**
01:58:22**
Daniel Englebretson
And that’s a rhetorical question. I’m not expecting you to answer that off the top of your head. But it plugs up to these chunks of work.

**
01:58:30**
Daniel Englebretson
And so.

**
01:58:33**
Daniel Englebretson
I’ll pause there and kind of give you a chance to react and digest to that. And if there’s anything you want me to dig into or ideas that are popping to mind here.

**
01:58:48**
Daniel Englebretson
How are you feeling, Simon? Yeah.

**
01:58:56**
Simon Walmsley
I can see the logical approach and how it’s breaking down, and you’re almost telling it how to think. So it becomes less of an abstract kind of answer and more methodical and following documented steps. I. Yeah, it’s kind of. It’s making sense. I know you’ve touched on it before, but what are the typical sort of actual outputs that would come out of this?

**
01:59:26**
Daniel Englebretson
Okay. Yes. So let’s. Let’s.

**
01:59:29**
Daniel Englebretson
So I thought you might go there. So all we did over here on the right was step one. So now I threw at step two in the process. So if we’re looking at that, we’re going to. We’re going past RFQ triage and intake and into requirements analysis. So it’s kind of showing me it’s gonna start. It’s gonna start cranking that here. But if you’re on the site looking at trial runs and you pick out one of these trial ones, we’re going to this case file number two, and you got outputs.

**
01:59:56**
Daniel Englebretson
So some of these aren’t as thorough.

**
01:59:58**
Daniel Englebretson
As others because I was testing different things. So. So the main thing it’s trying to.

**
02:00:03**
Daniel Englebretson
Do at this step, when it.

**
02:00:05**
Daniel Englebretson
Or when it gets into case file two, is it’s doing this requirements analysis. So I’m gonna pull up a couple different ones. So you can kind of see where these are at. So I was doing these crosswalks, and as I ran more of these, it made refinements. But what it just finished doing here then in this. In this example is it said, all right, I’m at step. Step four, CF02, extraction of the MVP sequence. Okay, I need to analyze all this information to identify following high risk of forceful requirements. So it’s kind of going down the list and saying, okay, here are the requirements that I found. High risk requirements. You know, here’s the category, here’s the source, where I found it, here’s why it’s a high risk, you know, and, and then, you know, you can.

**
02:00:53**
Daniel Englebretson
You can let it make the decision or you can give a decision.

**
02:00:56**
Daniel Englebretson
So it’s coming back and saying, all right, review what I just came back with. Do you accept them? Do you want to add them, do you want to delete them? So I’m just going to say, assume they are all accepted for this demonstration.

**
02:01:15**
Daniel Englebretson
And, and as you, the knowledge worker.

**
02:01:18**
Daniel Englebretson
Are working with a system like this.

**
02:01:20**
Daniel Englebretson
It’S telling you, okay, I got this information from this sheet.

**
02:01:25**
Daniel Englebretson
I followed this process when I did it. So you can kind of work backwards. And they’re not linked here because of how this interface is set up. But you could actually go into the documents themselves. But one thing I will call out, so if I took this answer really fast, it’s hard to do in this view because this isn’t like a production environment. But if I take the answer it just gave me in that chat and imagine that you had access to it.

**
02:01:52**
Daniel Englebretson
And I just indexed that answer.

**
02:01:56**
Daniel Englebretson
It just found all of the stuff that I have access to related to it. What it would do is it would give you a short list of all of the documents or all of the requirements or whatever that you could go into.

**
02:02:10**
Daniel Englebretson
That’s where you start moving through. All right, well, now what I’m looking for in this is a short list.

**
02:02:16**
Daniel Englebretson
Of the high risk things that I.

**
02:02:17**
Daniel Englebretson
Want to flag based on my specifications.

**
02:02:20**
Daniel Englebretson
For what high risk is.

**
02:02:23**
Daniel Englebretson
Then I’m.

**
02:02:24**
Daniel Englebretson
Going to ask it to, I’m going.

**
02:02:26**
Daniel Englebretson
To prove that, yes, these are the.

**
02:02:28**
Daniel Englebretson
High risk things to go look for.

**
02:02:30**
Daniel Englebretson
And it’s like, great, you approved it, so let’s keep going.

**
02:02:33**
Daniel Englebretson
So it’s going down the list and.

**
02:02:34**
Daniel Englebretson
It’S saying, are we compliant? No.

**
02:02:37**
Daniel Englebretson
Yes. No. Yes, yes. Need clarification players? No. And it’s answering the reason why is coming from here.

**
02:02:44**
Daniel Englebretson
So recall that I don’t have all the data in the world in this system. So it may not.

**
02:02:50**
Daniel Englebretson
Have what it needs to be able.

**
02:02:51**
Daniel Englebretson
To answer the questions, but that’s how it’s doing it.

**
02:02:54**
Daniel Englebretson
So now you’re saying, all right, here’s my draft compliance matrix. Now do you want to trust this?

**
02:03:02**
Daniel Englebretson
Not today. You would not trust this outright Today.

**
02:03:04**
Daniel Englebretson
Because you don’t have any experience with it.

**
02:03:06**
Daniel Englebretson
But every time you run this, if you find a deviation or you ask it to a retrospective and you change.

**
02:03:12**
Daniel Englebretson
The spec, it’ll get tighter and tighter, but at least instead pointing you in.

**
02:03:16**
Daniel Englebretson
The direction of, okay, here are the.

**
02:03:17**
Daniel Englebretson
Ones to go look at.

**
02:03:19**
Daniel Englebretson
This is where I was saying. When you get to the end of.

**
02:03:21**
Daniel Englebretson
This and you finish doing all your.

**
02:03:22**
Daniel Englebretson
Work, even if you just had the machine check your work, and you’re not letting it make the decision that in and of itself could solve that one problem.

**
02:03:30**
Daniel Englebretson
So it’s saying, all right, are you.

**
02:03:32**
Daniel Englebretson
Ready to proceed to the next one?

**
02:03:33**
Daniel Englebretson
We will address the gaps. We’ll say, okay, we’ll proceed to the next one. Just add it in here.

**
02:03:44**
Simon Walmsley
Yeah, I think I touched on it before that. Sometimes it’s almost as important to say where we’re not complying as opposed to just speccing out the product that we’re offering. And I think that’s good that it’s showing you where in the document, which document, and where it’s found a deviation. That would make life easy as well. Rather than just giving you a list saying, oh, you don’t comply here, here, and here. At least it’s signposting you where to go. So, you know, you’d at least then be able to maybe screenshot that particular paragraph. And then you have a conversation with your client and. Yes, that’s good.

**
02:04:29**
Daniel Englebretson
Yes, yes. Actually, I love that you called that one out, because one of my favorite steps on this is, okay, you can ask the bot. This could be an outcome. Queue up my shortlist for my phone.

**
02:04:41**
Daniel Englebretson
Call with the client. Have some standards for that.

**
02:04:45**
Daniel Englebretson
Jump on the phone with the client, record the conversation.

**
02:04:48**
Daniel Englebretson
I mean, tell them you’re recording it in the conversation.

**
02:04:51**
Daniel Englebretson
Bot pulls a transcript and refactors everything.

**
02:04:54**
Daniel Englebretson
Based on the transcript. So as long as you ask the.

**
02:04:57**
Daniel Englebretson
Questions and the client answers it will go back and pull it and.

**
02:05:00**
Daniel Englebretson
Come back and update all your information based on that call.

**
02:05:03**
Daniel Englebretson
And then you could say, okay, check.

**
02:05:05**
Daniel Englebretson
Your work and tell me how you know.

**
02:05:06**
Daniel Englebretson
And I use that process all the time for what I do. And so. So here it did. Is. Is it came back and it says, okay, we looked at this item. Deviation required.

**
02:05:15**
Daniel Englebretson
Here’s my rationale. Okay, here’s the draft for that, you know, analysis, the item deviation. Analysis item.

**
02:05:22**
Daniel Englebretson
So it just went down that list based on that. And so importantly, it’s not deciding what qualifies as a deviation. That’s in the spec and that’s in your Feedback up here of, yes, go.

**
02:05:34**
Daniel Englebretson
Hunt these ones down. Right.

**
02:05:35**
Daniel Englebretson
And so, and so you can keep pushing through. And so now it’s coming down to.

**
02:05:41**
Daniel Englebretson
The end and it’s saying, all right.

**
02:05:43**
Daniel Englebretson
Please review the draft. Do you accept these or not?

**
02:05:46**
Daniel Englebretson
Once you have, let’s proceed to the next one. So we’ll just keep this one going. I approve.

**
02:05:53**
Daniel Englebretson
Proceed.

**
02:05:54**
Daniel Englebretson
And I’m just going to give it. Have to manually give it here because of what we’re doing here. Just add it here as CPAT04 and drop that one off and we’ll hit.

**
02:06:05**
Daniel Englebretson
Go as it’s answering the question here. Just to reiterate so that people are tracking from a technology perspective, imagine that over here on the Left is a.

**
02:06:19**
Daniel Englebretson
SharePoint folder or your OneDrive or your.

**
02:06:22**
Daniel Englebretson
Entire Azure stack of data.

**
02:06:25**
Daniel Englebretson
Imagine that’s what’s on the left over here.

**
02:06:27**
Daniel Englebretson
Now, in Microsoft, different people have permissions.

**
02:06:30**
Daniel Englebretson
To different folders and files.

**
02:06:31**
Daniel Englebretson
Microsoft will respect that. So you can sit Microsoft on top of all of your data and say, only the people who have permission to look at this data can get answer that includes this data. You can bake that into it. Then you can say, all right, I’ve got these standard works system that I.

**
02:06:49**
Daniel Englebretson
Want you to follow.

**
02:06:50**
Daniel Englebretson
And those standard works are baked in here.

**
02:06:52**
Daniel Englebretson
We don’t have to rehash that.

**
02:06:54**
Daniel Englebretson
It just knows that those are there. You don’t have to tell it. It knows that you kick the flow off and instead of having to prod it through like I’m doing here, it just takes the step surfaces, the flags, feature feedback. You give it the feedback and off you go through. So here. Here we go. All right, so it just finished. All right, here’s my executive summary.

**
02:07:14**
Daniel Englebretson
So, you know, we’re pleased to present our comprehensive technical blah, blah.

**
02:07:18**
Daniel Englebretson
So you might decide that instead of.

**
02:07:20**
Daniel Englebretson
Executive summary at this point, you want a checklist and a set of questions to ask the customer on a phone.

**
02:07:24**
Daniel Englebretson
Call instead of this step.

**
02:07:26**
Daniel Englebretson
And that’s just a matter of what.

**
02:07:28**
Daniel Englebretson
Step do you want down here at the end, here’s my final deliverable.

**
02:07:33**
Daniel Englebretson
Does it show what you want it to show? Are there any specifications or whatever?

**
02:07:38**
Daniel Englebretson
This flow, then it allows you to map back to what decisions were made. If I take this whole conversation thread.

**
02:07:51**
Daniel Englebretson
Let me see if we’ll. Let me do it. I don’t have this one set up super well.

**
02:07:55**
Daniel Englebretson
If I take the whole conversation thread.

**
02:07:58**
Daniel Englebretson
As an example here and I’ll just index it. All right, so I’m gonna start a new chat and just ask it about this chat. Whoops. What standards were used to arrive at this outcome? I’m just giving you an example so you can kind of get your head around what might be possible here.

**
02:08:36**
Daniel Englebretson
So I can, it can look back.

**
02:08:38**
Daniel Englebretson
At all of this context that we gave it and ask it a question and it can look across what it has access to, which is what I just asked it to do, and it’s going to come back and give you a short list of everything that it looked at.

**
02:08:51**
Daniel Englebretson
So you might not actually run this query, but if you ran a query at the end of all this or.

**
02:08:56**
Daniel Englebretson
On a proposal and it said, you.

**
02:08:59**
Daniel Englebretson
Know what safety standards are relevant.

**
02:09:02**
Daniel Englebretson
To this process and you know, which ones did you cite when you did this?

**
02:09:08**
Daniel Englebretson
And it comes back and gives you.

**
02:09:10**
Daniel Englebretson
An answer, which you’ll see here in a minute.

**
02:09:11**
Daniel Englebretson
It will give you a table of all the standards the site, like the document where that was cited, like the standard itself, and then where it found it in the document you’re looking at. So you can cross reference. Here’s the standard, here’s where I cited it and here’s where it is in the document. And so like here, just to kind.

**
02:09:30**
Daniel Englebretson
Of show you an example of that, I just asked it, what did you use to do this? And it’s like, okay, here’s the procedural standards that we used here. Here’s the standards catalog that we use in here. There’s a standards catalog that rolls up all the standards if you want to see them all in one spot. It looked at these different artifacts and deliverable templates. So I use this template, this template.

**
02:09:54**
Daniel Englebretson
It allows you to go dig into how are these decisions made, what governed it and do I agree or disagree? So this might not be a.

**
02:10:06**
Daniel Englebretson
Query you’d actually run, but just trying.

**
02:10:07**
Daniel Englebretson
To paint a picture of like how.

**
02:10:10**
Daniel Englebretson
Can you kind of get value out.

**
02:10:13**
Daniel Englebretson
Of what happens at the end of.

**
02:10:14**
Daniel Englebretson
This and then kind of work backwards on that.

**
02:10:17**
Daniel Englebretson
So, so as far as deliverables outcome, not only can it give you the.

**
02:10:23**
Daniel Englebretson
Deliverables we just went through, but it can help you work back how you got those answers in the first place and agree or disagree. So I pause there before I go any further because I can keep going on this, but I want to make sure that you’re tracking with me and see if there’s anything you want me to dig into before I go any further.

**
02:10:49**
Bhumika Sachdev
Can we have Simon like share his screen and I don’t know, Simon, do you have Access to copilot or anything, you’re on mute.

**
02:11:17**
Daniel Englebretson
Technical difficulties.

**
02:11:27**
Bhumika Sachdev
And so Daniel, I want, I just wanted to see if Simon can like cast some of these files and create some of the standards like you did. I don’t know if that’s what you wanted to do, but just a thought.

**
02:11:43**
Daniel Englebretson
One, absolutely.

**
02:11:44**
Daniel Englebretson
I mean, all these files are coming to you.

**
02:11:47**
Daniel Englebretson
Two, the reason why I couldn’t do.

**
02:11:49**
Daniel Englebretson
That today is because I didn’t appreciate how many files this was going to be.

**
02:11:53**
Daniel Englebretson
But when were scoping this and there was just no, there’s no way.

**
02:11:59**
Daniel Englebretson
To use off the shelf premium tools to demonstrate this because of how many.

**
02:12:05**
Daniel Englebretson
Files it has because both of these.

**
02:12:06**
Daniel Englebretson
Tools are capped at 20.

**
02:12:07**
Daniel Englebretson
So like if you were going to.

**
02:12:08**
Daniel Englebretson
Do this right now and you were going to attach some files to it will only let you attach a maximum number of files and it’s going to be something like 10 or 20, depending on what you’re on.

**
02:12:17**
Daniel Englebretson
And so by the time you even.

**
02:12:19**
Daniel Englebretson
Just attach the standards, you’re at 20.

**
02:12:21**
Daniel Englebretson
And before you even have all of the RFQ files. And I just didn’t know that about the process when we first started. So. So we could try to simulate it.

**
02:12:31**
Daniel Englebretson
Here today, but it’s going to be underwhelming because. Because it’s just. It’s not going to be set up to do that.

**
02:12:39**
Daniel Englebretson
Well, however, when it is time to do it, because today it was really.

**
02:12:43**
Daniel Englebretson
About talking about what’s possible, refining it. Then when we come back after we scope something up, I can give you.

**
02:12:52**
Daniel Englebretson
A package of stuff that you can load up and use so that you can do this on your side as part of this. But it will require at least one.

**
02:13:01**
Daniel Englebretson
Copilot license and some configuration include copilot to be able to do that. But yeah, you could do that. But yes, I fully intend to give you the whole stack of documentation.

**
02:13:16**
Daniel Englebretson
So the question.

**
02:13:22**
Daniel Englebretson
That we’re trying to get into at its core is what are the chunks of work that bring the most value for the least effort out the gate? And from what we have seen, there is opportunity to set some standards and some templates against some data and ask some questions using some standards and some templates and get a response. And all of those that I’ve come up with so far, the process are all on that site that you can get, but they’re also. It’s also a file directory that I will give you.

**
02:13:55**
Daniel Englebretson
So.

**
02:13:57**
Daniel Englebretson
So it’s because I’m trying to answer your question and it’s just not.

**
02:14:01**
Daniel Englebretson
It’s just Not a.

**
02:14:02**
Daniel Englebretson
It’s not, it’s not an easy thing to do.

**
02:14:04**
Bhumika Sachdev
Yeah, no, I get it. Thanks.

**
02:14:13**
Daniel Englebretson
Oh, so that’s that said then.

**
02:14:16**
Daniel Englebretson
Let me just jump back over to our agenda really quick here so we.

**
02:14:22**
Daniel Englebretson
Can take this a few different ways.

**
02:14:24**
Daniel Englebretson
I recognize there’s a lot of data and a lot of critical thinking here, and we’re pretty deep into it, so I’m happy to take the conversation how you’d like. My main goal for today was to queue up how you break these things down and how these standards come to life and how that interface might, would look and then show you real examples of what comes out when the.

**
02:14:51**
Daniel Englebretson
Bot does it, and then think about together. If I give you a stack of.

**
02:14:57**
Daniel Englebretson
Documentation and a means to execute on it, what would you need to be able to start your kind of cycle of learning and improving and getting value out of this? And so I can tell you from what I heard already, the one that’s retrospectively looking back at a final product and flagging things is a pretty good opportunity from what I’ve seen. But there could be even like, generating the questions to ask the customer or things like that. And so as we’re thinking about the back half of today’s time, we’ve. We’ve now gone through the target workflow. We’ve talked about the current state of it, we’ve talked about how you would break it down, we’ve talked about what the flow is overall. You kind of saw it roughly.

**
02:15:47**
Daniel Englebretson
And what we baked time into the agenda for was to actively make changes or actively insert new logic or new templates or something like that, if we want. So we could take the time to go through the RFQ flow that we have and just talk about each one and see what we want to get documented, or we can really take the time however you’d like. But what I will do after this then is take all the feedback from you, make revisions to this and package it up in a way that’s usable on your stack. That’s where it’s going to go next. So I’ll pause there and see. Do you have a preference on how we spend the time? Do you want to go function by function, or do you want to get into some of the.

**
02:16:39**
Daniel Englebretson
Like, do you want to see more examples of specific things running, or do you want to. Knowing that I got like thousands of files here, do you want to test something? You know, what if we tried this or that, or what if I asked that we can really do whatever is going to help you kind of get your head around where can you start to get value from this? You’re on mute too. Yeah.

**
02:17:15**
Simon Walmsley
It’S hard for me to sort of conceptualize because obviously we’re creating something new that we didn’t have before. So it’s kind of hard to see what the end goal is going to be. I’m certainly appreciating and understanding the workflows that are being built behind the, you know, the structure of how it’s going to analyze these documents and come out with some outputs. Yeah, I’m not sure where, yeah, I’m not sort of verbalizing this properly, but it’s, it’s hard to know what, where the tool could go next, if you know what I mean, as.

**
02:17:58**
Daniel Englebretson
Yeah, yeah, yeah. Well, it’s a big question to ask you on the fly, so that is totally fine.

**
02:18:03**
Daniel Englebretson
That’s your answer. I, I can shepherd the conversation, no problem. I just also want to, if anybody has something they want to focus on, I would be glad to do that. So.

**
02:18:13**
Daniel Englebretson
But, and if you have ideas as.

**
02:18:14**
Daniel Englebretson
We’Re going through it, that’s totally cool too. But yeah, that’s. It’s to be expected that, you know, it’s a tough question to answer. Steve, I saw you came off or came on camera. Did you have a question you wanted to throw in before we go further?

**
02:18:29**
Stephen Hudson1
No, not in particular. I’m having to drop off the camera because the WI fi is not great where I am, so I’ve just had to save bandwidth and now I just think, you know, what we’ve seen so far is first of all extremely impressive. I would say that it’s a lot, there’s a lot more to it than I initially imagined. I mean, wrote a 10 sentence prompt prior to this to try to achieve the same outcome. So it’s, yeah, it’s quite incredible, the detail going through. And I think, yeah, please guide us further and kind of take us down the pathway because I think it’s probably bigger than any of us expected, really. So I guess we’re probably on the verge of being a little bit overwhelmed.

**
02:19:17**
Stephen Hudson1
At least I am, yes, I’m, I with it, I’m with it just about, you know, and I think that’s probably where we’re at. And I think, you know, at some point we will see it being tested and sort of run it through its paces and I think that’s what it kind of all comes together a little bit.

**
02:19:30**
Daniel Englebretson
Doesn’t.

**
02:19:32**
Daniel Englebretson
Does. Yes. And I totally appreciate. It’s overwhelming.

**
02:19:36**
Daniel Englebretson
It’s, It’s a really complex process.

**
02:19:39**
Daniel Englebretson
I don’t have to tell you guys.

**
02:19:40**
Daniel Englebretson
That, but it is probably the most.

**
02:19:43**
Daniel Englebretson
Complex process that I have broken down thus far.

**
02:19:48**
Daniel Englebretson
You know, because really, each one of.

**
02:19:51**
Daniel Englebretson
These case files is a process in.

**
02:19:53**
Daniel Englebretson
It of itself, and then even within.

**
02:19:54**
Daniel Englebretson
That as process, and then the volume is just so big.

**
02:19:58**
Daniel Englebretson
But, but yes, I’m totally tracking with you. And so let me.

**
02:20:05**
Daniel Englebretson
Well, actually, before I hijack the conversation, anybody else have anything else they want to throw in or direction they want to put on this before I move into the kind of, the next thing here.

**
02:20:17**
Daniel Englebretson
All right, so before we, before I saw anything from you, I had a conversation with Viveka before we had even like, signed on the dotted line. And I gathered from the conversation and created this artifact to help me try to understand and convey, you know, what, what are we doing? And, and so you have a target knowledge, workflow, you’ve got some friction in the process. You’ve got kind of, you’re trying to assess what’s a reasonable level of collaboration with AI that’s maybe feasible for this process. You know, what kind of maturity do you guys have already in terms.

**
02:20:57**
Daniel Englebretson
Of your stack and tools and practice and things like that?

**
02:21:00**
Daniel Englebretson
So I was trying to understand kind of that scenario then where did I think we might would get some value if we started looking at this.

**
02:21:08**
Daniel Englebretson
Through the lens of streamline, accelerate, enrich.

**
02:21:12**
Daniel Englebretson
And then really just trying to break that down into, well, what would be meaningful gains.

**
02:21:21**
Daniel Englebretson
As an overall whole?

**
02:21:23**
Daniel Englebretson
And the way I think about gains is if you can have a process today and then look up tomorrow or.

**
02:21:31**
Daniel Englebretson
A month from now and have another process.

**
02:21:33**
Daniel Englebretson
And that process is appreciably different from what it was in such a way that it has reduced your effort, made.

**
02:21:41**
Daniel Englebretson
The outcomes better, accelerated the pace.

**
02:21:44**
Daniel Englebretson
As you stack those incremental gains, you.

**
02:21:46**
Daniel Englebretson
Start to see enough difference between how.

**
02:21:47**
Daniel Englebretson
You were doing it and how you’re now able to do it. That’s where you start to be able to say, here’s the value that we.

**
02:21:53**
Daniel Englebretson
Got out of this. Right? Here’s. Is the juice worth the squeeze? Right?

**
02:21:57**
Daniel Englebretson
And so, yes, revenue and costs are.

**
02:21:59**
Daniel Englebretson
The easy trackers for stuff like that, but those are usually trailing indicators, and they’re really hard to map into a specific thing.

**
02:22:08**
Daniel Englebretson
And so as we’re taking this next.

**
02:22:09**
Daniel Englebretson
Kind of chunk of work here in.

**
02:22:11**
Daniel Englebretson
Front of us, and we’re thinking about.

**
02:22:12**
Daniel Englebretson
Measurable innovation or incremental change. We’re thinking about streamline risk, accelerate, can we reduce the effort? Can we get higher quality outcomes? Can we accelerate the pace of this overall cycle?

**
02:22:24**
Daniel Englebretson
You guys already signaled, oh, you could.

**
02:22:26**
Daniel Englebretson
Get more throughput for people like Simon. You could respond faster to deals because. And then basically beat the competition because of how fast you respond. You can reduce your operational risk by clawbacks from costs.

**
02:22:43**
Daniel Englebretson
So you’ve kind of dropped these big opportunities. And if we start thinking about those.

**
02:22:50**
Daniel Englebretson
Big opportunities through the lens of what we’re doing and working backwards on that.

**
02:22:54**
Daniel Englebretson
Of how do we make that possible. And then we start thinking about the questions we ask of the data, it becomes.

**
02:23:03**
Daniel Englebretson
And the most kind of fundamental sense.

**
02:23:06**
Daniel Englebretson
The questions that we’re asking of the.

**
02:23:08**
Daniel Englebretson
Data are basically the standards and the answers are basically the templates.

**
02:23:14**
Daniel Englebretson
And so we have these standards.

**
02:23:18**
Daniel Englebretson
I kind of put all this into a single index called standards catalog. So if you’re on the site and you just go into the search and you type standards catalog, you’ll see this.

**
02:23:26**
Daniel Englebretson
But I tried to. Some of them I deleted out.

**
02:23:28**
Daniel Englebretson
So they’re not. Some of the links might not work because I’ve been making refinements here.

**
02:23:32**
Daniel Englebretson
But as we think about that, you’ve.

**
02:23:35**
Daniel Englebretson
Got, you’ve got different standards and different templates.

**
02:23:39**
Daniel Englebretson
And so if you think about a.

**
02:23:41**
Daniel Englebretson
Template as like a compliance matrix or a deviation report or a deviation library or an RFI template, let’s take this one for example. So if I go to an RFI template, this template is basically where the answers go, right?

**
02:23:56**
Daniel Englebretson
You ask questions of the data and you send those answers somewhere and it.

**
02:24:00**
Daniel Englebretson
Goes to a template, right?

**
02:24:02**
Daniel Englebretson
And so the way you get a better answer back is by having a.

**
02:24:06**
Daniel Englebretson
Place to send it.

**
02:24:07**
Daniel Englebretson
And so if we’re thinking about getting.

**
02:24:10**
Daniel Englebretson
Value out of what we’re doing and.

**
02:24:12**
Daniel Englebretson
Appreciable measurable improvement, we have to start thinking about what are the templates that we need that we don’t have or that we could improve. And as we’re doing that, what kinds.

**
02:24:26**
Daniel Englebretson
Of things would we be potentially, what.

**
02:24:29**
Daniel Englebretson
Kinds of questions basically could we be.

**
02:24:31**
Daniel Englebretson
Asking of the data? So what I’m really asking you here.

**
02:24:36**
Daniel Englebretson
Is if we think about what we’ve gone through so far and you take a quick look at the kinds of.

**
02:24:41**
Daniel Englebretson
Templates that I’ve spotted, these were just.

**
02:24:44**
Daniel Englebretson
Intuitive to the process, but if there were other cross reference docs or other templates or other potential outcomes that maybe.

**
02:24:55**
Daniel Englebretson
Now your mind is kind of jogged because we’ve looked at this, that would.

**
02:24:58**
Daniel Englebretson
Be helpful to have like if I.

**
02:25:01**
Daniel Englebretson
Look at the metrics, for example, what.

**
02:25:04**
Daniel Englebretson
Kinds of things are we trying to.

**
02:25:06**
Daniel Englebretson
Flag as we go through it? I recognize that this is a lot. So keeping it as simple as possible, you know, to your question earlier, Simon, what kind of outcomes can we actually expect to get? Well, you can expect to get whatever outcomes that we template up, basically.

**
02:25:22**
Daniel Englebretson
And so, and so the way that.

**
02:25:25**
Daniel Englebretson
This whole thing has been orchestrated so.

**
02:25:27**
Daniel Englebretson
Far is in the design and build.

**
02:25:30**
Daniel Englebretson
Workshop, in the MVP documentation. I’ve got these workflow documentations which are these CF012334. And so if I.

**
02:25:41**
Daniel Englebretson
This readme right here explains what all these are.

**
02:25:45**
Daniel Englebretson
And so it just says, okay, we’ve got these prompt packs, we’ve got this matrix, we got this deviation report, we got this past deviation library.

**
02:25:54**
Daniel Englebretson
How would we use it?

**
02:25:55**
Daniel Englebretson
Originally I was going to do this with ChatGPT projects.

**
02:25:58**
Daniel Englebretson
You would take, you would create four.

**
02:26:00**
Daniel Englebretson
Projects, one for each of these things.

**
02:26:02**
Daniel Englebretson
You, you drop in the relevant files.

**
02:26:06**
Daniel Englebretson
Sorry, you drop in the relevant files to run it and then you can run the process. And so that process being at the.

**
02:26:14**
Daniel Englebretson
Highest level, very lightweight.

**
02:26:18**
Daniel Englebretson
First you have to say, okay, what.

**
02:26:19**
Daniel Englebretson
Am I reacting to?

**
02:26:21**
Daniel Englebretson
And then on down the list, the machine executes the process. So we don’t have to look at this in detail. This is just kind of the sequence.

**
02:26:28**
Daniel Englebretson
But where the human would look is.

**
02:26:30**
Daniel Englebretson
Okay, case file one, RFQ intake, triage prompt package role, what makes it successful, what are the expected inputs, what do we need to make sure we’re calling out? How do we get the value, what are the parameters? And then off it.

**
02:26:44**
Daniel Englebretson
This is what the machine was doing.

**
02:26:45**
Daniel Englebretson
When we’re saying, go do CF01, okay, go look at these things, look for this, do this and give me these outcomes.

**
02:26:53**
Daniel Englebretson
So, yes, this is a ton of.

**
02:26:55**
Daniel Englebretson
Information and I don’t expect us to get into the fine print on all of it right now.

**
02:27:00**
Daniel Englebretson
But knowing that this is how it.

**
02:27:02**
Daniel Englebretson
Starts to come together into essentially a standard operating procedure that the AI is executing, this is where our critical thinking lands of what do we want to know and why do you want to know it? So flipping back to. Let me pin this one, I’m just going to grab. I lost my link. Hold on, I think it’s over. Here we go. Even I have a hard time navigating all this stuff.

**
02:28:01**
Daniel Englebretson
So as we think about these templates.

**
02:28:04**
Daniel Englebretson
And schemas in the context of the prompt packs, which is just what I call them here, question I have for you, where we might would go next is from what you have seen and.

**
02:28:16**
Daniel Englebretson
From some of these ideas that have.

**
02:28:17**
Daniel Englebretson
Popped up, for example, the retrospective look back, what did we miss? Or the materials dive into, you know, nickel plated bolts or whatever? A question that we could spend some time thinking about here is like, what, what else would we want to request of the, of the data of the machine to go ask? You know, so if were going to do the materials request or the material dive on the bolt, the stickle plated, what would that look like? I’m not saying that’s a good one, so you can stop me on that one, but that’s what I’m trying to get my arms around. Because when I package this up and give it back to you and it runs the flows, it will give you everything we ask for and it will do it by following the standards and following the templates.

**
02:29:01**
Daniel Englebretson
I’ll pause there.

**
02:29:03**
Daniel Englebretson
Are you tracking me with what I’m driving for here in terms of what templates might we want, what standards might we want to capture? What other kinds of things would we potentially want to think about right now? Are you tracking me on that? Even if went into things like high risk keywords, are there other things like this that we would hunt for? So I’ll pause there and see how that’s sitting with you and where you might want to take it. Yeah.

**
02:29:41**
Simon Walmsley
I think building on the, like you say, the high risk keywords.

**
02:29:46**
Daniel Englebretson
It’S.

**
02:29:46**
Simon Walmsley
Probably going to take a bit of digging back on previous quotes that we’ve done, previous RFIs we’ve received. I can see that it’s obviously building this using the examples that I provided, the dozen or so, but obviously we’ve got hundreds that we can start to populate and pull more information from to build out on that. I think, as well as the keywords for the hazards, the things that we want to pick out on, I think we need to look at how we can build. Yeah, sorry, the high risk keywords as well as those kind of things. We also need to be telling it what is our standard, what are the standard materials and selections of components and materials that we use. So you need to be sort of teaching it what our products look like, what they’re made up of.

**
02:30:51**
Simon Walmsley
And then when it’s got a library of all of our standard component choices, it can then do the exercise, I guess, of saying, well, the RFQ is looking for this and I’ve compared it to your list of standard components or sub components and here’s the mismatches. So you have to address these or it might come back and say everything that’s being asked for in the RFQ you already supply so that you can proceed. But I can see there would be a bit of work and time involved in populating what our standards are, you know, because at the moment it’s obviously it’s built, it’s pulled in a lot of information using the examples we provided.

**
02:31:39**
Simon Walmsley
But as were touching on earlier, I think George was saying, you know, we wouldn’t necessarily put in our quotation that we use this particular type of electrical cable which is 1.5 millimeters in core size and we use nickel plated brass cable glands from this manufacturer, you know, or it’s a particular grade of stainless steel pipework that the showers are made from and the thickness and the powder coating. Our standard is this particular document and it’s so many microns thick and the more you can put in, I guess all of these things that we take for granted because we just, you know, we’ve got years of experience and product knowledge that is dotted around in our system but we need to be, I guess putting it into this to give it as much a bigger library of a reference point to begin with.

**
02:32:45**
Simon Walmsley
So I can see there would be some work to do there.

**
02:32:50**
Georgia Hogg
I think as well. Just to build on that. I kind of think that starting point of our information is good because obviously this is an ADNOT spec, but we have specs for different end users and therefore they have different requirements. So I understand the system will start to learn and look out for certain keywords. But for adnoc we’re seeing requirements for a certification called ecas. For Saudi we see requirements for ulfm, csa and it’s sort of little nuances as you go through different countries if you like Europe’s then different to. If were to supply to Russia, but Russia for example, or maybe Asia and then offshore and onshore.

**
02:33:33**
Georgia Hogg
So the, yeah, the more we talk almost that starting point seems to be us and then it can go from there to kind of spread out for the different contractors and end users that we come across and their own nuances.

**
02:33:57**
Simon Walmsley
I see Daniel’s just put in the chat. His, his n just dropped.

**
02:34:02**
Georgia Hogg
He’s very still. He was.

**
02:34:07**
Stephen Hudson1
Can always tell.

**
02:34:08**
Daniel Englebretson
Kai.

**
02:34:10**
Georgia Hogg
Welcome back.

**
02:34:12**
Daniel Englebretson
Sorry about that.

**
02:34:13**
Daniel Englebretson
It just randomly went down on me. But we’re back. You had just started talking and then.

**
02:34:18**
Daniel Englebretson
I missed what you said. I apologize.

**
02:34:20**
Georgia Hogg
Oh no, I was just basically saying this, the idea of having this basis of our own standards. Because the example we’re looking at now is obviously a really big tender, but it’s for adnoc, which is an end user in Abu Dhabi, but we deal with end users in Saudi, Oman, Asia, Europe, all over. And the requirements can vary. So obviously certain things, certain parts of the RFQ mirror each other, but there might be nuances that could come up. So this is sort of where I’m sort of thinking having our embedded information. This is our standard, this is what we tend to use as a starting point, might be the most beneficial method.

**
02:35:07**
Daniel Englebretson
I like you called that one out.

**
02:35:09**
Daniel Englebretson
I kind of picked up on client baselines.

**
02:35:12**
Daniel Englebretson
So like when it looks at one, this is most like this one over here. And so because it’s most like this one over here, these are things that.

**
02:35:20**
Daniel Englebretson
We know we need to look for in this one. So back to like dynamic scenario setting.

**
02:35:25**
Daniel Englebretson
Maybe, maybe all you need is regional.

**
02:35:28**
Daniel Englebretson
Or maybe you need industry or maybe you need whatever that might would be.

**
02:35:31**
Daniel Englebretson
But, but yes, if the scenarios change.

**
02:35:34**
Daniel Englebretson
Significantly based on the profile of customer.

**
02:35:37**
Daniel Englebretson
Having a library of those profiles, it.

**
02:35:42**
Daniel Englebretson
Immediately hones the focus into that most representative example.

**
02:35:49**
Daniel Englebretson
It’s not saying, okay, this is like adnoc, so do adnoc.

**
02:35:52**
Daniel Englebretson
It’s just saying when customers like ADNOC think about these things. So that’s a great example as well. I’m trying to my notes up here as well.

**
02:36:02**
Daniel Englebretson
So is that in line with kind.

**
02:36:05**
Daniel Englebretson
Of what you were thinking there as well?

**
02:36:10**
Georgia Hogg
Yeah, I think, I mean there are some of the stuff that’s sort of looking there in the adnoc. A lot of that will cross over. But there might be certain certification standards that might differ. Like adnocs might use the Atex, whereas in Saudi they want American standards, for example. And that’s kind of something I would maybe think to look for. But as Simon said, if we had someone new to the business or maybe we, even if we got documents, we might not even know where. It’s for little things, we sometimes they get blocked out, we don’t, we’re not told the end user information. So it’s sort of, yeah. Learning and nuancing those things as well.

**
02:36:54**
Daniel Englebretson
Okay, okay. All right, that’s a good one. I’m just taking my notes over here. All right, so I’ve got two scenarios you guys gave me so far. I’m writing down here. So I got thrown off because my stuff went down.

**
02:37:10**
Daniel Englebretson
So. All right, so those are two good ones.

**
02:37:14**
Daniel Englebretson
So I am going to react to them once we’ve collected a few more. So you can kind of see where we could go with this. But this Is good. So other scenarios that might come to mind in terms of setting the stage, you know, when a new RFQ comes over, you might be setting the stage or when you want outputs coming out of this process, you know, things that might would be nice to have in terms of you have to sink a lot of time into them or high risk or something like, could even be things like what metrics might we want to track.

**
02:37:59**
Daniel Englebretson
So like coming out of step one, it’s tracking any file integrity issues, any missing documents, any anchors is what it calls like the most important documents to look at that came in basically or you know, how many things high risk things were spotted, you know, so you could almost think of this as like if you had a bi dashboard that was sitting on top of the data that was coming back and after you ran this it gave you a snapshot that said assess how high risk this one is or how much deviation this is or how complicated. Like you could surface it in that type of way. So it could even be like if you were trying to assess the performance or the quality or the risk, you know, what might would you want to know?

**
02:38:45**
Georgia Hogg
There could be cases of contradictions. Sometimes you get files and in one of them it says everything has to be stainless steel. You get another one, it says you can use a GRP for example. So maybe flagging contradictions within the files.

**
02:39:01**
Daniel Englebretson
Huh.

**
02:39:06**
Georgia Hogg
And that allows us then to the client to go. Right. Please clarify, what are we following here?

**
02:39:13**
Daniel Englebretson
Huh. I’ll just make a note in. Okay, all right, that’s a good one. All right.

**
02:39:37**
Daniel Englebretson
So then.

**
02:39:43**
Daniel Englebretson
Trying to think what, trying to decide where I would go. I have some ideas on where to go next but I want to make sure that we’re using the time as smartly as possible. So we talked about client profiles, we talked about the RFQ looking for X scenario profiles. If we look at some of the outcomes. Let me just collapse this quality. I look at some of the change log. So I’m going to start at the bottom because it’s the more recent stuff. So if I look at some of the things that came out, for example, let’s see if I can find a good one that, Let me see if I can look at something like this what work within standards deltas. Okay, yeah, here’s a good one.

**
02:40:58**
Daniel Englebretson
So one of them ran and it.

**
02:41:00**
Daniel Englebretson
Got to the end and it said hey. In our crosswalks moving forward, we’re now adding explicit section title and page slash line for IP minima. The spec Area class minimums, variant power and chiller IP power. So in the hazards table, we added explicit, you know, weak link flags for this evidence or whatever.

**
02:41:21**
Daniel Englebretson
So as you think about like, oh, it just went through and looked at this thing and it found, okay, next.

**
02:41:27**
Daniel Englebretson
Time we do a crosswalk, we should be looking for this specifically, you know that’s a kind of an example.

**
02:41:33**
Daniel Englebretson
Some of these things will start to.

**
02:41:35**
Daniel Englebretson
Level off as more go through. But you know, as we’re thinking about that kind of stuff, anything that I know this is super abstract. I’m trying to make it nice. Reference to codes, index community. Yeah, like IP certificates not attached and the declarations or ambiguity around the spec or you’ve already called out a lot of these. All right, So I just wanted to see if there was anything that was going to jump out of us to help us kind of think about other examples. So I’m going to edit then, unless you have another one and show you how we can start to do this. So switching back then. All right, so what are we about to do?

**
02:42:43**
Daniel Englebretson
What I’m trying to convey in this part right now is how much effort does it take to do, set aside from knowing how to do this and.

**
02:42:55**
Daniel Englebretson
Just how much time and effort does it take to react to those three.

**
02:42:59**
Daniel Englebretson
Scenarios that you just came up with, you know, and how much time would it take to change the spec and what would the outcome of something like that be? And so what I’m going to do is I’m going to come in here, I’m going to say I was reviewing this outcome with my SME and they asked me the following scenario. I’m going to take that one scenario that you gave me first. RFQ is looking for X. I need to compare a list of standards of subcomponents, how they are made. I need to generate a list of mismatches for the material itself, ag these different things and where the deviations are. So then, based on our process and what we have observed, how can I best refine our workflow to react to this?

**
02:44:00**
Daniel Englebretson
So, so the reason why I wanted to ask you this question is once you have this kind of flow in place and you have a feedback loop system where you’re asking these questions.

**
02:44:20**
Daniel Englebretson
What I’m advocating that you would do.

**
02:44:21**
Daniel Englebretson
Is periodically people who are related to the process would meet and they would have this kind of conversation and they’d.

**
02:44:28**
Daniel Englebretson
Say, man, based on what we’ve seen.

**
02:44:29**
Daniel Englebretson
This week or based on what we saw this month, we really need a different step in the process to react to when this happens. You’d have that conversation and then you’d be equipped to put that scenario to your system. And the system comes back and says, oh, based on the scenario that you’re trying to resolve, here’s how you should update your flow. Then what you would say is, yes.

**
02:44:53**
Daniel Englebretson
I agree, or no, I don’t agree.

**
02:44:55**
Daniel Englebretson
And then I want to show you how it actually work, make that change.

**
02:45:01**
Daniel Englebretson
Because your intuition about how much time.

**
02:45:05**
Daniel Englebretson
And energy it would take to go make a change like this, it’s not wrong.

**
02:45:09**
Daniel Englebretson
It is really complicated. But if you have tools and you.

**
02:45:13**
Daniel Englebretson
And you and you have the experience to be able to react to this, it doesn’t have to take a long time. And so here it goes. It’s like, okay, we need a reliable way to compare RFQX to a standard list of sub components and so on. Currently, scaffolding already gives us the right anchors and gates. So it’s looking at DS02 and 03 gates. So step two and three, you know, it’s looking at CSV coupling evidence mirroring standards. So that’s where it’s looking. So it’s like, okay, rigor and citation style already designed. So it’s saying, okay, we already have an evidence pack checklist, and we already have a citation style guide for how it’s going to create the evidence and cite it. So hazardous matrix and decision table patterns exist and can be mirrored for materials.

**
02:46:00**
Daniel Englebretson
So it says, okay, we already have a hazards component matrix and we have a decision table. So we can, we can use that kind of logic for what we’re talking about here. So now we want to look at a materials Component matrix pass in 03. So parallel to hazards driven by standard subcomponent library vendor baselines and RFQ extracted specs.

**
02:46:19**
Daniel Englebretson
So blah, blah.

**
02:46:20**
Daniel Englebretson
So how are we gonna go about doing that? Well, first, okay, we’re gonna look at this step, then we’re gonna prescribe this action. We’re gonna think about these things, and as a result, we’re gonna make sure that the standards are specific and the citation style and the evidence pack.

**
02:46:35**
Daniel Englebretson
All right, so how are we gonna do that?

**
02:46:36**
Daniel Englebretson
Well, we’re gonna define the standard subcomponent library. So it’s going through the data sheets and all that stuff. It’s structuring by product subcona attributes, all that stuff store as a symbol markdown table per model of repo. That’s just what we’re using here. And then, okay, CF02 requirements, no schema change.

**
02:46:54**
Daniel Englebretson
So it’s not changing the schema, but.

**
02:46:56**
Daniel Englebretson
It does need to expand the extraction to capture these things and tag them in this way. And then the crosswalk also needs to add these things. And then new additions in 03 is we need a new matrix file, but no changes in the CSV itself. So materials component matrix, this is where.

**
02:47:13**
Daniel Englebretson
We need, we need these columns, these.

**
02:47:15**
Daniel Englebretson
Things and so on and on down it goes. So it gets to the end, blah, blah, and it says, ok, here’s what we look at, here’s what we need to do next. So it’s going to say which component should be included? Do you want materials component matrix created now? Any EPC specific tolerances, because so on.

**
02:47:32**
Daniel Englebretson
So now it’s reacted to that.

**
02:47:34**
Daniel Englebretson
And I’m going to say, okay, please review our rfq workflow, our standards and this example. Then make annotated list of assignments documents.

**
02:48:05**
Daniel Englebretson
So again, without getting hung up on.

**
02:48:09**
Daniel Englebretson
On, on the, on what I’m doing more just kind of demonstrating the.

**
02:48:16**
Daniel Englebretson
Idea.

**
02:48:19**
Daniel Englebretson
And we’ll just hit go.

**
02:48:22**
Daniel Englebretson
So you would have the capacity to have critical conversations about what did we.

**
02:48:29**
Daniel Englebretson
Observe that we want to change. Then you would have the standards and the standard operating procedure to go back to your system and say, here’s my scenario.

**
02:48:38**
Daniel Englebretson
Then the system is doing all the heavy work to go figure out how that fits in. Then it’s bringing you back answer and saying, okay, based on your scenario and based on our process, here’s how.

**
02:48:51**
Daniel Englebretson
I propose that you update.

**
02:48:54**
Daniel Englebretson
I am doing this on an ugly backend system, but it could be made.

**
02:48:58**
Daniel Englebretson
In a pretty way where we’re just focused on this continuous improvement. Let me let it finish. And then we can kind of react to that.

**
02:49:11**
Daniel Englebretson
As it’s finishing and we’re thinking about, okay, what are we trying.

**
02:49:14**
Daniel Englebretson
To do today and how are we trying to get to an outcome?

**
02:49:17**
Daniel Englebretson
You guys have already given me a.

**
02:49:19**
Daniel Englebretson
Ton of context to refine what we’re doing. I expect that as you think back on some of this, you might have some other ideas as well.

**
02:49:26**
Daniel Englebretson
And then what I aim to do is get back to you the package.

**
02:49:31**
Daniel Englebretson
Of stuff and the standards so that you can do this for yourself and then we can decide what you want to do from there. This is how that might would flow. It’s like, all right, well I look.

**
02:49:41**
Daniel Englebretson
Through these things, I suppose we’re going to do this. Here’s how I do it, here’s where I would add these different things and.

**
02:49:48**
Daniel Englebretson
Then you get to the end and you can say do you want to do this or not?

**
02:49:53**
Daniel Englebretson
So we don’t have to read this in great detail right now because I know it’s a ton of detail but this is where it will just go insert into those documents that one thing.

**
02:50:03**
Daniel Englebretson
That needs to get inserted into that document based on what we learned and then off we go.

**
02:50:08**
Daniel Englebretson
So does that. Well, I just kind of talked through.

**
02:50:13**
Daniel Englebretson
In terms of like a process of feedback loops and continuous improvement and documentation update and how the AI can help you do that, you know, with less effort.

**
02:50:22**
Daniel Englebretson
Is that, is that like making sense to you guys?

**
02:50:27**
Daniel Englebretson
Is it helping to paint a picture of how something like this would evolve? Yeah, yeah.

**
02:50:36**
Simon Walmsley
I can see how that would work.

**
02:50:40**
Daniel Englebretson
Okay.

**
02:50:42**
Daniel Englebretson
Okay. So I know we are three hours deep and we’ve gone through a lot of stuff and I can appreciate that it’s heavy. We don’t have to run the full four hours. We had planned extra time to do some chat pt Goodness. But we observed, we see why that’s.

**
02:50:57**
Daniel Englebretson
Going to be tough to do today.

**
02:50:59**
Daniel Englebretson
We don’t have to stay the whole.

**
02:51:00**
Daniel Englebretson
Four hours but I am more than happy to stay the whole time and keep going through it. Taking us back to our agenda really quick.

**
02:51:22**
Daniel Englebretson
This is what were spending a lot of time on.

**
02:51:24**
Daniel Englebretson
So how could we co create the process, integrate best practice automations into this MVP and kind of think about it. So we spent some time thinking about that.

**
02:51:32**
Daniel Englebretson
You saw a little bit how some changes like that can play out but.

**
02:51:39**
Daniel Englebretson
Really it’s just getting the gears turning to think about this so that as our next step in this is I can take the high value stuff we’ve identified and make those kinds of changes and document it all up so that you can have it and start kind of playing with it. And then coming along with that is I also owe you basically guidance on how do you do what we just did so that you can do it should you want to. So so that being said.

**
02:52:11**
Daniel Englebretson
How would.

**
02:52:11**
Daniel Englebretson
You guys like to. Do you guys want to wrap up.

**
02:52:14**
Daniel Englebretson
Now at the 3 hour mark or.

**
02:52:15**
Daniel Englebretson
Do you want to spend the time going further into some of this stuff? I’m just trying to respect that. It’s, it’s heavy mental work. So totally up to you.

**
02:52:25**
Stephen Hudson1
Yeah, I think break it off now.

**
02:52:27**
Daniel Englebretson
Good.

**
02:52:28**
Stephen Hudson1
So we can got time to reflect everything that we’ve been through so far. I don’t know what you’re thinking, Simon.

**
02:52:36**
Simon Walmsley
Yeah, I think I can totally understand and I mean it’s been a lot to take in but I can see how we’re Building the architecture behind how this will work even so much in that, you know, you’ve got a section on, I think it was heuristics. You’re even telling it how to think, how to interpret the language that we’re going to use. And I can imagine for this to work we’re going to have like a sandbox area of information that we need to help you by populating. So like I said, it doesn’t know what it doesn’t know.

**
02:53:16**
Simon Walmsley
And there’s all this information that we have in our business of like I said, all these subcomponents that we use, we need to be, I guess listing all as much detail as we can to put into this program to get some really good use out of it. Because unless it understands these things, it’s never going to know what’s a deviation or what’s an issue. So yeah, I think we’re going to have to give you some more information about our products. Whether that’s just a list of things, I’m not sure how that would, how you need it. But it’s certainly been very eye opening and interesting today.

**
02:54:08**
Daniel Englebretson
Well, I’m glad that you’re tracking me and you’re getting the value out of it and I appreciate that. It is a very heavy thing. And so as you’re, you know, we’re kind of started on the front of this as we start to wrap up.

**
02:54:22**
Daniel Englebretson
This is as you’re thinking about this.

**
02:54:25**
Daniel Englebretson
And you can take obviously as much time as you want to think about this and you’re having a chance to kind of sit with it a little bit.

**
02:54:32**
Daniel Englebretson
We’re thinking about what you just mentioned.

**
02:54:35**
Daniel Englebretson
About the data that has to exist.

**
02:54:37**
Daniel Englebretson
We’re thinking about what facts does it.

**
02:54:39**
Daniel Englebretson
Need, access to what conceptual knowledge, what methods, what kind of metacognitive knowledge. This is not something I made up. It comes from somewhere else.

**
02:54:47**
Daniel Englebretson
But what I have tried to do is curate everything that I can into.

**
02:54:52**
Daniel Englebretson
What’S called a knowledge graph, which is what this is to show you. Here’s what the world looks like. This is what your RFQ world looks like based on the data that you gave me.

**
02:55:03**
Daniel Englebretson
And here’s how it’s all related. Now, whatever you want, you can come in here and navigate through as you’re thinking about this and come in here.

**
02:55:12**
Daniel Englebretson
And say, oh, okay, well back to our workflows. What was the process or what was the goal or what was the thing to get you exposed to that?

**
02:55:21**
Daniel Englebretson
But you’re right, we have to curate into a place in this case it.

**
02:55:25**
Daniel Englebretson
Would likely for you would be on your Microsoft stack and these, the kinds of things that you’re curating so that you can do these kinds of things.

**
02:55:34**
Daniel Englebretson
This is super abstract right here and it translates into these are the things that we’re doing right here and the data and the inputs that are being.

**
02:55:45**
Daniel Englebretson
Used to do it. And it’s all governed by standard works and documentations.

**
02:55:50**
Daniel Englebretson
And so what we have now though is a good enough understanding of your world that you can say things like.

**
02:55:57**
Daniel Englebretson
Hey, I need a process that I can go dive in and look at these materials and the bot can go.

**
02:56:02**
Daniel Englebretson
Through everything it knows and bring back to you.

**
02:56:05**
Daniel Englebretson
Here’s what I think your process should.

**
02:56:06**
Daniel Englebretson
Be for you so that you can.

**
02:56:08**
Daniel Englebretson
Just weigh in on that.

**
02:56:09**
Daniel Englebretson
And that in and of itself is like, oh my God, wow, I can.

**
02:56:14**
Daniel Englebretson
Now have it go do all the critical thinking for me and come back with the standard. And then I just weigh in on that standard and that’s where I’m.

**
02:56:19**
Daniel Englebretson
Trying to get right. So to kind of land the plane.

**
02:56:22**
Daniel Englebretson
For us, I will leave this up for you guys to have access to and I will come back to you with some feedback based on what we heard today. And I can make some revisions. It will update what this looks like and then I will get direction from your team and Vamika for how you.

**
02:56:41**
Daniel Englebretson
Might would want to move forward or not move forward, however you guys want to do it.

**
02:56:45**
Daniel Englebretson
And I’ll give you some ideas on.

**
02:56:47**
Daniel Englebretson
How that might look so that we can kind of get it back to.

**
02:56:51**
Daniel Englebretson
You in whatever the most meaningful way is. And so that’s what we’re ultimately hoping to walk away with today is having an appreciation for, you know, how do you break these things down and how do you start to look for value and how do you start to look at, you know, where AI may be able to help.

**
02:57:09**
Daniel Englebretson
So also we have this recorded so.

**
02:57:11**
Daniel Englebretson
I can share that as well. And I’m happy to share all the materials too to kind of help break it down.

**
02:57:17**
Daniel Englebretson
So I guess I will come back to you with some follow up questions.

**
02:57:24**
Daniel Englebretson
Based on what we talked about here.

**
02:57:26**
Daniel Englebretson
If there are any.

**
02:57:27**
Daniel Englebretson
I’m sure there will be to see if I can jog you guys of anything. The there’s no time constraint on my side. It’s really as fast as you guys want to move. I know Bomika has some time constraints specifically, so I mean, I’ll let her drive that, but you guys can take whatever time you want with the material to think about it and then we’ll make another iteration. If you have more examples that you.

**
02:57:49**
Daniel Englebretson
Want to throw at me, I can.

**
02:57:52**
Daniel Englebretson
Process more examples through to have it kind of like you see in those episodes over here on the trial runs, I can run more through and you.

**
02:58:00**
Daniel Englebretson
Can see then, okay, after it runs through, how does it change?

**
02:58:04**
Daniel Englebretson
So we can start to observe that. And so that’s something I’m happy to do as well.

**
02:58:14**
Bhumika Sachdev
So I will be out office starting end of the day tomorrow. Daniel so from a next steps perspective, you are going to come back to the team. I know Steve also is out office. So I don’t want the, you know, I want Simon to move forward. What we, what I was actually hoping is at least have, if Simon has to do it on his own. I know you said you’re going to give us like the standards and everything. So if you were to use our own like Microsoft Stack as copilot or anything that we have, what will be the next step? Is that what you’re going to cover in the next meeting?

**
02:58:55**
Daniel Englebretson
Yeah.

**
02:58:55**
Daniel Englebretson
Yeah.

**
02:58:56**
Daniel Englebretson
So part two of this is now.

**
02:58:58**
Daniel Englebretson
That I have the feedback from you, I’m going to give you all the documentation for how it is as well as a recommendation for how you would deploy it on the Microsoft Stack. That’s kind of what I owe you as the second part of this. Based on what I’ve heard, you obviously can give me feedback and we can work with that. But what I want to make sure that you have coming out of this first phase of work is documentation on a flow that makes sense to you and a means to deploy it. Then as you may recall, if you want me to help you deploy it, we can go that route. Or if you want to deploy it.

**
02:59:31**
Daniel Englebretson
Yourself, we can go that route too.

**
02:59:34**
Daniel Englebretson
But that’s kind of the next big chunk of work. So I will crunch this, I will synthesize what I heard, I’ll bring it back, you can react to that at your own speed. Then I’ll finalize what we have and finalize recommendations for how you might would deploy this on your stack. And not just technically, but also from a process perspective. Then my goal is always to teach you to fish.

**
02:59:59**
Daniel Englebretson
If you.