S03 EP04 with Yohann Doillon - So you think you're ready for AI

Most companies investing in AI are not getting the return they expected. The tools are there. The licenses are paid for. And the results still aren't coming. Yohann Doillon has a clear explanation for why, and it comes down to one missing layer: context.

In this episode of The Habit Architect, Michael Cupps sits down with Yohann Doillon, co-founder and deployment strategist at 10x Partners, to talk about what it actually takes to bring AI into production at the business level. Yohann comes from the Palantir ecosystem and has spent five-plus years building ontology-based data foundations for companies ranging from Fortune 500 firms to growth-stage scale-ups. He and his partners are senior-only engineers, by design. Their work is translating how a business actually operates into a digital structure that agents and AI systems can understand and act on.

The conversation covers the three pillars 10x uses to assess AI readiness: internal operations, product strategy, and competitive risk. Yohann walks through what data readiness actually looks like versus what companies say when asked about it, why buying 10,000 Copilot licenses is not an AI strategy, and what an ontology is in plain language, not the jargon version. Michael ties it back to his own years in the RPA and BPM world, and together they work through a live insurance industry example where 14 disconnected systems are being brought into one coherent picture so that real decisions can get made.

They also get into the question of whether a nine-month consulting engagement is the only way to assess your AI readiness (it isn't), what it means for a company to have a "living" AI assessment rather than a snapshot, and the personal second-brain system Yohann has built that his own partners are tired of hearing about.

The Habit Architect is sponsored by Enterprise Diagnostics and Time Bandit

Check out our Live Show Events here: The Habit Architect Live Show

Subscribe to our Newsletter: The Habit Architect Newsletter

Transcript

0:35

Hello, and welcome to The Avid Architect. This is Michael Cupps, your host as always, and I'm really looking forward to our discussion today. It may get a little bit technical, but we'll tr- we'll try to keep it open for everybody. But it's it's an amazing topic because so many companies are investing in AI, and they may be investing in the right places, they may be investing in the exact wrong places, and we're gonna talk a little bit about how just through simple assessments and understanding of your business you can get to the right conclusion before you dive in, and all of a sudden you're paying token fees that you didn't realize you were paying, and you're getting results that you didn't expect to get results from. CUPPS: So it's a really interesting topic because it is such a big thing with companies today to get... figure out what is their path with AI. And i- there's a tendency to jump in as fast as you can, but there's also needs to be some thought and progress around that. And to do that conversation I'm really excited about our guest today. CUPPS: Yohan's gonna join us, and he's with a company called 10x and I've collaborated with them on a few things, and it's exciting, their perspective and how they're helping their customers. So he's got not only a knowledge of the

space, but also a perspective of customers doing it the right way. CUPPS: So I'm really excited about that. So while we'll bring Yohan from backstage, I'm just gonna mention, do like us, favorite us, send it to your friends, send it to your parents, all of that stuff so you, the podcast gets more coverage. We appreciate all of that all of that love, and continue to do so, please. CUPPS: And also, don't forget to ask questions. If you are sitting online join us with a question and we'll, we'd glad to d- make it a discussion. Hello, Yohan. Welcome. YOHANN DOILLON: Hello, Michael. How are you doing? CUPPS: 1:35

Very good. Good to see you. You're coming to us from Switzerland, just to put a perspective for people. But- YOHANN DOILLON: Exactly. CUPPS: Yeah. Why don't you tell us a little bit about yourself? YOHANN DOILLON: Sure. First of all, it's an honor to be here, Michael. Thank you very much for the invitation. I'm essentially a co-founder at 10x Partners. I lead our commercial operations and also I'm a deployment strategist. I will get a bit more into details in a bit of what that entails. YOHANN DOILLON: We're a company, as you said, Michael, based in Switzerland. We also operate, in the United States now. We'll have operation in the UK a bit all across Europe, and we focus essentially on bringing AI into production. So we are a team of

senior-only engineers. We don't hire juniors, and that's our our USP, I would say. YOHANN DOILLON: So pleasure to be here, Michael, and looking forward to the discussion. CUPPS: Yeah, it's great, and I'm glad you set it up the way you did, th- the... because it's important for people to think about, w- we're past the point of AI just being a tool that somebody chats with ChatGPT and gets cool answers or plans their vacation using it. CUPPS: 2:35

But in business, it's really an interesting inflection point because there are some massive success stories and there's also some massive failures. And there's a number of reasons for that, so we can't categorize it all into one or two buckets. No. But there is a bit of missing context, I think, when some companies start and they get on their AI journey. CUPPS: Maybe they hire a a, a s- a chief data scientist or chief AI officer, but they still haven't aligned. So can you talk a little bit about that missing context? YOHANN DOILLON: Yeah. AI has been around for more than three years now. I think really it came into play in, at the end of 2023 at, with sorry, 2022 with ChatGPT 3.5. YOHANN DOILLON: It got into the hands of, many different users. We had- we've seen different ways over the past

over the past years, but I don't think that... and, you have a clear demand from the market. Data centers cannot be built, so there's a strong demand on that side. But more and more businesses that are indeed not really seeing this ROI when it comes to, when it comes to AI, right? YOHANN DOILLON: 3:35

They invest a lot- Yep ... on tools lots on tokens even even more than the let's say the actual usage- Budget ... is usually yeah mu- much higher than the demand. What we see usually is indeed this context layer, which does not which is not really present in, in, in many firms and which indeed does not allow, this this return on investment piece to actually happen. YOHANN DOILLON: Yeah ... that's what we've noticed across many different industries, many different customers, Fortune 500 firms, all the way to scale-ups. It's a problem that comes again and again. Of course, you'll have firms that came a bit more prepared because they've invested on the data foundations years ago, and now they are getting all of the benefits from those early investments. YOHANN DOILLON: But the ones that now have to play catch-up also have to invest on the, on on those foundations and build the actual context because that's what essentially

makes artificial intelligence actually carry the context of your business, right? It's all of those- Yeah ... rules, everything that's happening on a daily basis within o- your organization that gets en- encoded in a way that an agent, AI, let's say can understand and act on, right? CUPPS: 4:35

Yep. Yeah, and I mentioned, and there in a minute ago, that pe- companies have gone out and hired a chief AI officer, and I'm not criticizing the role, I'm not criticizing the individuals. Absolutely not. What, what is sometimes disassociated with that, when you create a separate group that is your AI team from the business, th- that context really has a risk of getting completely missed because you still need that knowledge base that's sitting in the, how you run your business today. CUPPS: And I, and I- Yeah ... liken it to one, one example that you and I discussed some time ago, is a company rushes out and they buy 10,000 licenses of Copilot and say they're AI, right? And then- YOHANN DOILLON: Yeah ... CUPPS: they don't see results, and I'm sure you've seen that happen multiple times. YOHANN DOILLON: We've seen that happen multiple times. YOHANN DOILLON: I think it's a bit the easy card to play to be on the safe side, so to say. So you would buy, it could be Claude, could be

Copilot whatever. I think it comes from good intention. You want to bring- Yeah ... the tool, to, to the organization. You want to see what are the usage patterns. You want to see how people are using it. YOHANN DOILLON: 5:35

Maybe you could have some creative use cases that come out of it. But it's still shooting a bit in the dark, if you ask me. So indeed, building this context layer- So that you can have some actual interactions with what's happening directly in the business. You can have actions that is, that are taken by those agents on systems that, those agents would actually understand from the inside and not just from the external documentation. YOHANN DOILLON: All of those topics definitely are playing a big play. CUPPS: Yeah. Yeah. And it seems there's a lot of big consultancies, I'll say whether they're strategy consultants or other big firms, and they're... i... What I see these waves that come every two or three months that say, "Oh, adoption- is down. Success isn't there, ROI isn't there." And I'm- Yeah ... not saying they're all disingenuous, but it also helps them spread a little fear that you're doing it wrong. But I don't know if it really has to be a nine-month assessment engagement. What's your thought on that? YOHANN DOILLON: So I come from management consulting, data strategy,

AI strategy. YOHANN DOILLON: I've been doing that pre-ChatGPT. Yeah. So I know, let's say, the complexity of those organizations, how long it can take, right? To go through, the entire value chain, to go through the entire, business units, really try to di- dissect them through all the way down to workflows. It can take a lot of time. YOHANN DOILLON: 6:35

But nowadays, I think as you said with AI and if you do it right, you can get to a very decent assessment rather quickly. So of course, you could spend definitely nine months and evaluate the business and see how robust it is when it comes to AI. But, in those nine months you might have, I don't know, 10 iterations of of Anthropic in the meantime. YOHANN DOILLON: Yeah. So the, the disruption keeps happening. So I guess the answer to that is, all right, if you can have a living system, living assessment in a way, that understands how your business is evolving, Yeah ... how your industry is evolving, what is the risk that is posed by just, future iterations of of models, that's I would say a big a big win nowadays. CUPPS: Yeah. That's, that, that's exactly it. I think you... And you said something that just is amazing to me. What you do today may be obsolete in four months just because... and obsolete's maybe

the wrong word. Maybe it evolves. It can evolve at a different pace- Exactly ... than we've ever seen before. CUPPS: Yeah. YOHANN DOILLON: You might wanna go beyond the snapshot. I think that's the- Yeah ... that's the question. You might want to have this living system that you can rely on so that you can essentially assess its maturity and see how... are you- CUPPS: Yeah ... YOHANN DOILLON: are your initiatives even in line with your objectives? YOHANN DOILLON: Are they reaching- Wow ... the goals? Do you have the right resourcing for those initiatives, et cetera, right? A lot of this essentially is now feasible with AI and with the right tooling. CUPPS: 7:35

Yeah, absolutely. And so let's talk about the assessment. A- and you and I t- spoke before this and I, there were three key areas that we discussed- as you need to assess these in your business, and there's certainly sub-elements of each of them, but let me just say them. One is how is AI gonna help you run your business? That's Operation Alpha concept that a lot of companies have. And then second is what is... where is AI gonna infuse or help in your product itself- Yeah CUPPS: the product that you deliver to your customers. And the third was manage risk. And so we can talk about each of those individually, but I thought those three pillars were were a good way of framing it because it's internal it's

external, and then it's overall what's happening in your sector. CUPPS: So let's start with h- understanding your own business. How do, how can AI infuse that? Where do you start with the assessment and what are you looking at? YOHANN DOILLON: Yeah that's a good point. So I think on the, definitely on the internal side, it's usually what you know best, right? So it's understanding, what are the, the typical use cases that you would have applied a few years ago that you that you put in your backlog and that now you want to tackle with AI. YOHANN DOILLON: 8:35

Understanding also where do you stand in terms of data readiness. So we talk to a lot of firms, and many of them, like 99% of the case, they say, "Oh, but our data is messy," or, "Our data is not ready." So you can try to assess how advanced you are in terms of, h- how your data is structured. Is it laying in legacy systems? YOHANN DOILLON: Do you already have- Yep ... a bit of a more structured layer that we can tap into? YOHANN DOILLON: And la- last but not least, I think it's a bit more on the where do you stand? What is the reality of your current business? Are people still, making phone calls to update each other, or do you actually have systems that are a bit more advanced running, right? YOHANN DOILLON: Yeah. So that's a bit more understanding the

internal, let's say, way of working. Process. Yeah. Exactly. And I think this is m- that- that's where an actual human assessment can can really support, right? CUPPS: Yeah. And a- and so to recap, you said understanding your data and is it accessible to be into workflows or agentic or whatever it is that, that will benefit. CUPPS: 9:35

And then I think those internal processes are a big one. And it's... Understanding processes isn't a new thing, 'cause there was BPM that was a decade worth of- Yeah ... and still going of things that you can build and workflows. And then RPA hit this, hit the market, and it was about, automating from your desktop task and so on. CUPPS: Yeah. And so the processes are maybe not well-defined, but they've been given a good shaking or two just to see what's in there. Yeah. So it's not far from there, but- So you're saying right now if the data doesn't align with that, then it's more difficult. And that's an... Is that an easy assessment, just to understand where they sit operationally in their data? YOHANN DOILLON: Yeah. It's usually not as trivial, so you need the right foundations to essentially do that. So the moment you have, let's say,

those processes that are well documented, you can put agents that will learn from what a human would be doing, capture- ... YOHANN DOILLON: Those, processes and capture those, those learnings. YOHANN DOILLON: That's really what we mean by encoding business knowledge, right? It's why do you take decision A and not decision B, and for what reason, right? So capturing a bit those, let's say, workflow decisions is something that AI is not really good at, but it's... It takes it takes a bit of time. But if you... YOHANN DOILLON: 10:35

Let's say, if all the work that you've been describing, Michael, has been already let's say, documented, you'll have a good idea of, where are the bottlenecks. Then indeed, if you put AI on top, that will definitely increase the the quality, let's say, of the assessment by a lot, and you would of course shrink the timeframe, right? CUPPS: Yeah. Yeah. It's... What's interesting about it to me, I... Coming up I was part of the RPA world and the BPM world. Yeah. And one thing that when we were doing a process analysis and we drew the processes out, but we never really went to the data structure. We just said, "Oh, this data has to be here." CUPPS: So it... I think there is a diligence layer that's new if you really want to- Absolutely ... add AI to succeed. Yeah. So let's move on to the second pillar, and then I'll...

then we'll cover about all three. But the... how do you infuse your products? How... What is your product strategy associated with AI and how you serve your customers, I think is probably the, the best way of putting that. YOHANN DOILLON: Yeah, absolutely. I think that's that's a very interesting topic. I don't think there is one answer to that question, right? From my perspective, what we call AI boosted features. Yeah. So essentially, putting an LLM for the sake of putting an LLM as a feature of your product is maybe not the best idea. YOHANN DOILLON: 11:35

But trying to rethink how your product is essentially operating and put, YOHANN DOILLON: Some some... and bring AI into, in, into the product where it makes sense, I think is something that is more interesting from my perspective. Yeah. And then second, there... Now you can go much faster when it comes to iterations. I think from... Instead of just looking at the product itself, just looking at the... YOHANN DOILLON: on the development side, I think that's- Yeah ... that's actually where things are much more interesting. You can, really quickly pass on all of the analytics about your product- to, to an AI system that would then try to generate some AB testing you know, launch the campaign for specific demographics

and then iterate almost autonomously on, on those. YOHANN DOILLON: All of these areas on the development side I think can be actually quite quite interesting. Of course, there's a lot of point solutions nowadays that try to solve very specific Yeah ... features of this product. CUPPS: Yeah. YOHANN DOILLON: But yeah I think from my perspective it's gonna be very difficult anyhow- Yeah YOHANN DOILLON: Not to be relevant without at least some, some sort of AI component. But I would bet a bit more on the internal development cycle- Yeah ... than just purely on the, CUPPS: Yeah. Totally ... YOHANN DOILLON: on the AI features. CUPPS: 12:35

Yeah, but I even think what's exciting about putting AI around your product or in your product, it... CUPPS: not just for a tech company, but for somebody that sells plumbing fixtures, right? They sell- Yeah ... faucets and things like that. All the stuff that can go around it, how you serve your customers, how you educate the customer to use your product, how to h- where is my shipment, the, all of those things. CUPPS: Not that hasn't- Absolutely ... already existed, but now you can put it together in, in a customer journey or story that is really powerful. And by the way, that's how you're gonna compete. If your product is commoditized, how are you gonna compete on, you're gonna, YOHANN DOILLON: on the service? Absolutely. I, I, that's,

you're right. YOHANN DOILLON: I think I came more from the SAS angle, but, Yeah. Yeah ... for physical products, I think the story is even more compelling because indeed, a lot of those things that someone would be doing, or like you might have even different departments, right? Using a different a different lingual towards the customer- Yeah YOHANN DOILLON: Or, having different ways of treating treating s- different problems. All of this you can now harmonize. You can try to understand a bit things that are, I don't know, maybe the marketing department knows something that customer- Yeah ... service doesn't know. All of those learnings can now be shared much much quicker. YOHANN DOILLON: Absolutely. CUPPS: 13:35

Yeah. It's and that's the fascinating part for me is just how we can have a relationship with our customers is changing or c- or can change. Yeah. And then the third pillar we talked about was risk. And I get internal risk, you've gotta have guardrails and all that stuff. CUPPS: I was more coming from the risk of your industry, your sector. Yes. So if you sell lawn equipment or you sell technology or whatever, your competitors aren't standing still, right? So- YOHANN DOILLON: That's true. That's true. I think from that perspective, AI will definitely help you with, some more up-to-date and also much faster, let's say, competi- competitor analysis. YOHANN DOILLON: You can have some new, I don't know, all of the new startups

typically that are trying to eat your market share- Yeah ... that you can spot a bit quicker. Yeah. All of these, topics definitely can be can help, but you can also use it as a defensive a bit as- Yeah YOHANN DOILLON: 14:35

as a defensive tool, right? Once you understand your landscape much better, you can def- definitely try to prioritize some initiatives that maybe were, really far in your backlog and you didn't really want to to- Yeah ... to bring them up front. But then, if you see that the market is fa- favoring, like, one, one specific feature or one specific way of de- delivering a service, then you might also want to take that trend, right? YOHANN DOILLON: And but the... My, my point is now it's much easier to, let's say, understand everything that's happening within a market than ever, right? And bringing- yeah ... those insights to sales, to marketing definitely is a big plus, and that should be part now of, you don't... Where you used to rely on consultants to give you a snapshot- CUPPS: Yeah YOHANN DOILLON: every six months of a market, now not with all markets, of course, right? And not at the level- Yeah ... of granularity that you always need, but at least to get to 90%. Yeah. You can definitely, assess this, this external risk quite quite nicely just with public

sources. CUPPS: Yeah. Yeah, if nothing else, we're giving good tips on how to cut your McKinsey or Bain or Boston Consulting Group out there. CUPPS: But the, the... But the... But what's interesting about it is, as you were a strategy consultant, there, there's always this notion that you should be the first mover advantage, and all of the things you hear about. But what's exciting to m- to me at least or my perspective when I look across the competitive landscape across any industry is- Now, even if you're not the first mover, you have a chance to catch up and maybe even leapfrog them because AI is available to mid-market. CUPPS: 15:35

It's available to any size company, and it's about the thoughtfulness you put into your strategy more so than YOHANN DOILLON: the, the tech. That's very true. CUPPS: Yeah. YOHANN DOILLON: Yeah. It's very true, and I think, yes, definitely it is leveling the playing field. I would still say that, the more money you have, the more you can invest- Yes YOHANN DOILLON: and the more you can go this last mile. So just to come back to the context question, a bit of a parenthesis. So indeed buying those co-pilots license is absolutely it's maybe step one, but to actually bring AI into production, into things that people are using on a daily basis, you

need to understand, as we said, this context. YOHANN DOILLON: How- Yeah ... how are you gonna do this, right? You essentially... So there is a technological part that we can cover later. We call it an ontology, a knowledge la-layer. Yeah we can find different terms. But you also need- Yeah ... the people who can do both, the talking to the business and then the actual development of that knowledge layer. YOHANN DOILLON: Yeah. That's the work that at NX Partners we, we specialize on, so really being able to translate those business realities into something that is actually gonna be used to bring AI into production, right? YOHANN DOILLON: Yeah. Yeah, parenthesis closed. I forgot for a- CUPPS: 16:35

Yeah... for a point. That, the... CUPPS: No, that's, no, that's good. But what I wa- also what I would wanna tie to that is that we can use AI, and you can... When you and I talked about, when you look at EBITDA, and for those that- YOHANN DOILLON: Yeah ... CUPPS: don't know it's basically your earnings and things like that measure... It's a measuring stick of how your company's doing. CUPPS: Yeah. It's looking at costs, it's looking at revenue, it's looking at all these things. I think with EBITDA and customer health you can use AI to diagnose what's really happening and then find the variance of where you can actually make a difference, right? YOHANN DOILLON: That's correct. That's correct. So

if I can jump a bit into the how you would do this. YOHANN DOILLON: Yeah. So typically, what, how would you assess that on a typical ba- consulting engagement? Yeah. You focus on one metric, maybe a couple of metrics, and you try to drill down, okay, for this business unit this and this is going yeah. YOHANN DOILLON: 17:35

This and this is going less well. Let's focus on what's going less well, and we try to fix it. The beauty of, the systems that are now available and that we are building together, Michael, is that we can bring intelligence, we can surface intelligence from, multiple points across the organization, bringing that into a picture that makes sense, right? YOHANN DOILLON: And then you can have actual alpha, because that's really what your company is doing on a daily basis. It's pure... It's the intelligence of your business that you try to bring into a coherent picture. Yeah. And once you start to have that and to juice that- Then it, it becomes interesting, right? YOHANN DOILLON: So typically what we see, I think you work a lot with mid-market firms- ... where a lot of, I would say especially like the in- let's say industrials or, you know- Yeah ... CapEx heavy CapEx heavy industries, usually they have... They've been relying on, let's say,

legacy systems for many years. It's really hard for them to, let's say, make those systems talk with one another without a lot of integration work. YOHANN DOILLON: Sure. And nowadays and w- with the methodologies that we- that we put in place, it is possible to actually bring... bridge those, those gaps, so to say- Yeah ... and then get the actual return on investment when it comes to extracting again- Yeah ... the intelligence from various systems and various parts of the org- organization that usually don't really talk to one another for various reasons. YOHANN DOILLON: Yeah. CUPPS: 18:35

Yeah. What's amazing with that, I'm gonna show you my age'cause I've got the gray hair to prove it. But I year- years ago, I was with a company called WebMethods, and we sold ESP- Yeah ... or Enterprise Service Bus, which was- Yeah ... integrating systems. And it was still powerful then, but it still baffles me that 20 years later, there's still companies, like you said, are struggling to get their legacy systems to get the information. CUPPS: And to me, when I saw this concept of ontology, which we're gonna talk about here n- next, is that, that was the light bulb. That was like, if you can get your systems and your... not just your systems, the way your business operates into this ontology. And for

those that may not know, ontology is an archaeological term that you can look up. CUPPS: But in, in this context that we're gonna talk about it's basically your business in a digital format that allows you to act quickly, learn quickly, all of that stuff. So maybe we can s- maybe you can kinda give us a, a view of what the ontology is, and then we can talk about practical use cases YOHANN DOILLON: of it. YOHANN DOILLON: 19:35

Of course. Of course. So as you said, Michael, so ontology comes from the Palantir ecosystem. That's where we come from. Now you see that the market is, picking up of the term. You have it Microsoft is launching an ontology. I think Databricks, Snowflake also have something very similar in place. CUPPS: Yeah. YOHANN DOILLON: So everyone is understanding that, okay, this ontology, which essentially, as you said, is a digital twin of your business, it's a representation of not tables and let's say the IT world, but really things that both... I me- let- we call them business entities that both the business and and the IT and data scientists of this world can understand. YOHANN DOILLON: Yeah. Bringing that picture all together is what is required to have a, efficient AI in production. So

as said we can have agents running around. It's very easy to set up agents. If they don't have the right context, they cannot take the right decisions. They might access data that they might they actually shouldn't access. YOHANN DOILLON: They might take actions on that data, which, Also they should not. So all of this is let's say, yeah, allowed by this ontology. So once you have that coherent picture in place, you can imagine that you can build, very different use cases, very different things that you could not before, because now you have your logistics department that talks to your- Yep YOHANN DOILLON: 20:35

manufacturing department, that talks to sales, right? And then, you have those entities nowadays that, that are brought together into, of course, data is backing that, right? But you have e- entities that the, the, the business actually understand And you can build on top of it. And then once you've done that o- that original ontology building, so to say, of course- Yep YOHANN DOILLON: things starts, start to compound because that's what you're doing on a daily basis that is mapped. Yep. And the first use case might take you a bit of time, but then the second one would be shorter, and the third one would be shorter, et

cetera, et cetera, et cetera. Yeah. So that's why we see a lot of companies that are actually taking that, that step- Yeah YOHANN DOILLON: Being quite successful because the, the speed of advancement of AI now combined to this ontology- CUPPS: Yeah ... YOHANN DOILLON: layer is actually a big one. CUPPS: Yeah that to me is what's exciting about it. Because, I've worked in the insurance industry for a number of years and their data silos are massive, and they're important. CUPPS: 21:35

Yes. It's how p- people get reimbursed for their hospital stays, or they are reimbursed after a cata- catastrophe, a h- hurricane or something. But what was interesting is that to use a simple analogy, the left hand and the right hand didn't always know what was happening in those systems. CUPPS: And this ontology layer actually allows it to become this living thing, I think, where you can start seeing that something happened over here and it can trigger something over there which normally had to be a very difficult process. Is that... CUPPS: i'm trying to th- think of a layman's way of saying it, but I look at a kind of a caricature of a company. CUPPS: They had sales over here and marketing here and order processing over here. But now that... Now th- picture that

in an actionable way, right? Not just independent units. YOHANN DOILLON: Exactly, and think of all of the updates that one side have to give to one another, to, to one another just to move things around in a traditional setup. YOHANN DOILLON: It's, this should... It's not that it should not happen, but it should be automated because those let's say business areas should be connected by data- Yeah ... and make decisions on top of the same data. They should not be updating and trying to reconcile an Excel sheet left and right? YOHANN DOILLON: There's so much time that is lost that is lost there. Yeah. But to get to, let's say use cases, typically- CUPPS: 22:35

Yeah... YOHANN DOILLON: once you have that mapped up, right? So you have, let's say the, the, the insurance example. You have many systems nowadays that, you have pricing on one side, you have underwriting on the other side, you have reserving on the other side. YOHANN DOILLON: And traditionally there is, you try to bring people in the room so that there is a bit of a... Even just an agreement of, on, on- Yeah ... definitions, right? CUPPS: Yeah. YOHANN DOILLON: So once you have those kind of pain points that are resolved, then you can move on to actual, concrete use cases. YOHANN DOILLON: We see that typically when it comes to the resolution of claims. Yeah. It's,

it can be very complex to, to resolve some claims and to make sure that you're not, leaving money... Y- there is no money leaking from those claim resolutions. Yeah. Some processes can be 15, 20 steps long, and you have to go through quite some documentations. YOHANN DOILLON: You have to sometimes, you also call some other departments. A lot of those steps nowadays, if you have the right context again, and all of the data connected, can be automated and have- Yep ... let's say, at specific points where, like a human decision is needed. Again, a, a human check, a human control. YOHANN DOILLON: 23:35

Yeah. And then what's important is that th- those decisions are encoded back into the ontology, and then- Yeah ... will be used for the next decision that will be made, and then the AI will be get will get smarter, and smarter, and smarter- Yeah ... because it will have more context, again, more intelligence from the business. CUPPS: Yeah. Abso- absolutely, and that self-learning is important. We do have a question from Killian, and I appreciate the question. We're talking about how it can improve business or Killian is, and then, but he says "Do you take environmental footprint into account when thinking about deployments on this scale?" YOHANN DOILLON: That's a great question. We

have, let's say, the luxury of working on the platform side, so more the how do you actually engineer, let's say, all of these all of those data foundations to then bui- build use cases. Of course, so sadly we don't have, let's say, live data on how much let's say AI is costing in terms of environmental, Yeah YOHANN DOILLON: 24:35

emissions live. But, th- at least the way we work, of course, in a cost effective manner, try to reduce the, the, the- Yeah ... social footprint. There is still the Jevons paradox. The the bigger the system gets, the more you want CUPPS: to use YOHANN DOILLON: it. So- YOHANN DOILLON: Yeah ... yeah. We are seeing that- Yeah ... work. CUPPS: I think that's, I think that's a big one too. I know of a firm in in Canada that is actually retiring systems because they've now got their data in one place and they can start to... their digital footprint is less because now they're able to do more workflow with less, those legacy systems. CUPPS: And so I don't know if that you can, like you said, I don't know if you can measure the CO2 of what that means to them by...

but they can start at least looking across that- Yeah ... that landscape and saying, "What can we reduce?" And I actually saw- ... this particular construction firm actually put out a note to their vendors and said, "Prove you need to, we need to continue to work with you," which is pretty dramatic, but it's a good thing- It is CUPPS: to think about, right? Yeah. Because do we really need five of those systems if we can do it with one? And so what does that do? So I, again, I don't, I think it's hard to measure the CO2 of that, but it's an interesting perspective. YOHANN DOILLON: Exactly. It's not, of course, AI drives a lot of that. You, I think- Yeah YOHANN DOILLON: 25:35

you've seen maybe the numbers that were, the emission numbers that were released by, the big tech of this world recently. Definitely there is an impact. Is it, is part of that offset by what you just mentioned? I would say that, yes. What's the delta? I'm not sure, honestly. YOHANN DOILLON: I- Yeah ... leave that to CUPPS: experts. Yeah. Yeah, exactly. Yeah. Thank you for the question- ... Killian. Now let's get back to the use cases Yohan. W- can you walk us through a... and you don't have to use the company name if it needs to be confidential, but just maybe a, how you, how it started and where it got to. CUPPS: An end-to-end project that 10x ran and the big results. YOHANN DOILLON: Yeah, that's a good

point. So I would say if we're thinking greenfield projects So think of... Yeah, I think that- that's- that's potentially a good one. We have been engaged, and that's- that's a rather recent one, right? So it's not... YOHANN DOILLON: We're not there yet. But typically we've been engaged to by... It's- it's an insurance firm that is based in in North America, where essentially we are getting into, their operations to- to try to understand how do we bring a lot of the, let's say, carrier state. They are... YOHANN DOILLON: 26:35

They are brokers, so a lot of the carrier state that they have that they have internally into one coherent picture, right? So typically you would have people who- who are, managing... They- they have like long Excel sheets with all of the- the offering of the different carriers. And then you have you have that at a carrier level, then you have that at a broker level, and then you have the actual broker that manages those brokers that have to- to reconcile a lot of a lot of sheets. YOHANN DOILLON: Yeah. And to get to an actual view where you can say,

"Okay, this broker is actually performing much better than this broker. Maybe I should do some marketing campaign to incentivize this one or this one, or incentivize this geography." Or- Yeah ... maybe this one broker is not, talking as well as as I would expect. CUPPS: Yeah. YOHANN DOILLON: 27:35

Then I need to, reconcile maybe 14 or 15 systems together to get to that picture, right? Yeah. So the- the target that we are ... That we- we're still working on it, right? But the- the target that we have is essentially to bring all of this, Again, I think we're talking 14 systems all together into one coherent picture so that the broker can make, Yeah YOHANN DOILLON: can have a proper view over their, all of their individual brokers and then take actual marketing actions, right? Yeah. So it's nothing that is completely new in terms of marketing initiatives. But let's say that once they have this holistic view- CUPPS: Yeah ... YOHANN DOILLON: let's say the people who are making those marketing initiatives are not... YOHANN DOILLON: they- they were not even able to- to see some of the, the underlying data because that was- Yeah ... essentially laying in with- with operators, right?

Yeah. Typically. Yeah. So- CUPPS: And what that... And what that YOHANN DOILLON: means to the- Breaking the silos. CUPPS: Yeah. Yeah, I'm sorry, go ahead. I didn't mean to interrupt you. YOHANN DOILLON: The whole point is breaking silos again. Yeah. Yeah. So we've seen that in the... In the insurance space, as I was describing before. We've seen that in the man- manufacturing. Manufacturing is probably the- the best example. Yeah ... if you can look online at the public sources, like the cases of of Stellantis and the likes, it's really impressive how they really broke down those silos. YOHANN DOILLON: But now you have like smaller firms that are definitely also interested in- in similar similar setups. CUPPS: 28:35

Yeah. And what- what I was... When I rudely interrupted you, I do apologize. But w- what I was gonna say is what that means to the customer is phenomenal. Because how many times- ... have we tried to get an answer to sol- resolve something and it goes through multiple channels and you don't... CUPPS: you- you don't get there as quickly as you want to as a consumer? But what this is doing when you're s- when you're talking about breaking down silos, it actually allows the company to be more responsive to you because no one wants to wait on an answer. But the reality is in some of these companies, they have to wait on an answer because that data is somewhere else. CUPPS: Exactly. And now what you're saying is it's gonna be at their fingertips so they can take care of that customer

more quickly, right? YOHANN DOILLON: At the end of the day, we think about value, and I don't think that bringing information from A to B is a very valuable thing to do. I think the value really lies in the intelligence of the human that the decisions that an- anyone can make based on their judgment, on their experience- Yep YOHANN DOILLON: 29:35

once they have all of the data available, right? So this, this last mile that I was describing is where the value is. It's not, sending an Excel sheet from A to B that will essentially drive a lot of business value from my perspective. So we try to break those silos as much as possible. I know it's quite conceptual. YOHANN DOILLON: In reality- Yeah ... it's much more complex than that. CUPPS: Yeah. YOHANN DOILLON: But, Yeah, of course ... yeah, it's it's at least directionally what we work for. CUPPS: Yeah, absolutely. And I thank you for painting that picture of your, of that customer and the insurance example. I think everybody... I don't know if there's a human on the planet, maybe some younger kids don't know they have it- CUPPS: but everybody has insurance in some way. There's insurance in, in your life somewhere. And, Yeah ... making that more efficient and hopefully more affordable or at least price competitive is really a, a benefit YOHANN DOILLON: for them. Yeah, because

th- think of it, Michael, like o- once... H- how many, let's say, it's not that we have to reduce head count necessarily. YOHANN DOILLON: I think like actually you can do much more once you have all of the systems in place. But all of the positions that we have to fill just to, again, pass information from A to B, check an Excel file because the mapping is not existent or something like this, it's, it adds up. It adds up. Yeah. It adds up. YOHANN DOILLON: 30:35

And I think those... th- this kind of work will get, more and more done by AI, by agents so that people with actual judgment can move up the ladder in a way and actually spend not 80% of their time on, on, on small processes- Yeah ... and 20% on judgment, but more the opposite, right? Yeah. 80% on judgment and 20% on, on internal topics. CUPPS: Yeah, that's a good way of putting it. Yeah, absolutely. Tell us a little bit more about 10x just just And how people can find you and YOHANN DOILLON: Sure. Yeah, how people can find us. So we are on LinkedIn . Yeah. On LinkedIn, on, on website of course. People can contact any of the partners. We come from from Palantir. YOHANN DOILLON: My CEO, Antoine is ex-Palantir. The three other

co-founders we come from from management consulting, but always in the data and AI space. We've been building on Palantir, but also on other data platforms focusing on ontology sort of setups for the past, I would say, five, six years, if not more. CUPPS: Yeah. YOHANN DOILLON: 31:35

So we've been there, done that. We ha- we have our scars. We know- Yeah ... what works, what doesn't work as well. We ha- we had like a, of course, a few failed deployments that we learned from. And yeah- Yeah ... always a pleasure to engage with with anyone who's interested in bringing AI into production, es- seeing actual value from from artificial intelligence, going beyond the hype- Yeah YOHANN DOILLON: Being very grounded, really. That's what we focus on. And yeah, I would say anyone who's interested in having a chat feel free to- Yeah ... to reach out to myself or to my partners. CUPPS: Yeah, that's fantastic. And thank you for that. And I would say that, what we talked about today you and I have talked a lot about, is these are set, a sim- not I don't wanna say simple, but they're, they don't have to be six-month, nine-month engagements to assess where you need to go. CUPPS: We can do it pretty quickly. No. Find you an action plan, figure out what that data story looks like, what the competitive story looks like- Exactly ... et cetera. Yeah ... really encourage that,

that view as opposed to the alternative, which is really expensive and slows you down. Y- your competitors can move faster if you're not careful. CUPPS: So that's fantastic. I really appreciate the context. I do ask every guest I used to ask what's the one habit you can't do without every day, but I'm, but now I've moved it to what is one thing that Yohann's doing today that you are gonna replace with an agent or AI or something in the very near future? YOHANN DOILLON: And I can use something that I've replaced recently. Would that answer the question? Yes. Yeah, CUPPS: that, that works too. Okay. Yeah. YOHANN DOILLON: 32:35

All right. So my partners who are listening to me right now are gonna say that I speak about it way too much, but essentially I managed to recruit- to create what you call like a second brain. YOHANN DOILLON: So- Wow ... essentially all of the interactions that I'm having every day, all of the, let's say, all the noise that happens in my life, I've now managed to capture it in a structured manner so that- Wow ... I can retrieve thoughts, I can retrieve context, things that happened many months ago. It's also structured in a way that, you can create relationships. YOHANN DOILLON: AI can find relationships that I was not even thinking of, Wow ... between

different elements and everything that's happening on a daily basis, right? So think of it as a system that scans parts of my Slack, parts of my emails parts of my meetings as well, not all of them. Various different sources- Wow. YOHANN DOILLON: Yeah ... and then bring that into a structured manner so that I don't have to remember, what happened on the 20th of of, I don't know, January, because I can retrieve that information very nicely. And everything that, that, got connected to that decision I made on that day is also traced, right? YOHANN DOILLON: Yeah. So that's really how, CUPPS: That's fascinating ... YOHANN DOILLON: 33:35

it's not really a habit. It became like a, a habit of like- No, no. Yeah. Yeah... or my agent, actually. My agent is doing that on a daily basis. Yeah. CUPPS: Yeah. No, that- But it's, ... that's exactly... YOHANN DOILLON: Yeah. CUPPS: That's a that's a fascinating example. CUPPS: I, I don't do that, but I've started experiencing it being done to me because I use Granola. I don't know if you're familiar with it. Yeah. It's a notetaker kind of t- app. And I noticed this week it was pulling in some emails and some other meetings and putting a snapshot together for me, but I'd love to talk. CUPPS: We should have you back just to talk about your, your- YOHANN DOILLON: Definitely ... CUPPS: your second brain. So YOHANN DOILLON: that's fascinating. It gets like context. It get b- it get

backs to context building Michael? Yeah. CUPPS: Yeah. YOHANN DOILLON: It's- it's nothing else than that, but f- on, at my personal level, right? Yeah. So we do that for companies. YOHANN DOILLON: We also do that for - CUPPS: For you ... YOHANN DOILLON: individuals. CUPPS: Yeah. Excellent. Thank you so much for for joining us today. I think it's fascinating. I would like to have you back to talk about more of this as, as we can. Of course. YOHANN DOILLON: Happy to have a follow-up. CUPPS: And the, I know you had to... That painful... You had to meet Florencia, although that's not painful, but the fact that she- CUPPS: 34:35

is Argentina and you're Switzerland, and at le- i- if it can s- spur a r- rivalry there, YOHANN DOILLON: Absolutely. I'm a French person living in Switzerland, so yeah, not not the best results- ... against the Argentinians. CUPPS: That's right. That's right. Yeah. Thanks again for joining us Johann. Really helpful. CUPPS: And find Johann on LinkedIn, and I know he t- he'll take questions and talk about your business, et cetera. So please do reach out to him. And thanks again for joining The Habit Architect. We appreciate your input and your... And again, if you could share it and like it, we thank TimeBen and Nev- Enterprise Diagnostics for sponsoring us. CUPPS: And do forward this on to your friends so we can get more viewers. Thanks a lot, and have a great

week. YOHANN DOILLON: 35:35

Thank you very much, Michael. Have a good one. CUPPS: Thank you. YOHANN DOILLON: Bye.