Claude Code Moment for Factories? Cognyx + Oplit

Episode Summary

Software development got its agentic moment years ago. The open question is what happens when the same shift reaches hardware engineering, supply chain planning, and the factory floor. In this episode of AI Across the Product Lifecycle, Michael Finocchiaro talks with two French industrial AI founders — Matthias Berahya-Lazarus, CEO and co-founder of Cognyx, and Thibaut Wilhelm, CEO and founder of Oplit — about where industrial AI is already delivering and where the scaffolding still does not exist.

Cognyx is building an AI engineering platform for hardware, bringing software-style acceleration, executable knowledge, and agentic workflows into product development. Oplit is building an AI supply chain platform for industrial companies, combining agentic planning with operational research and industrial knowledge models to optimize factory performance. Both founders describe the same structural change from different angles: AI shifted their bottleneck from writing code to deciding what is worth building. Wilhelm argues supply chain is an ideal AI playground — data-rich, repetitive, and complex, with one demonstrably better answer per problem. Berahya-Lazarus counters that agentic engineering needs a single source of truth, composable modules, and executable knowledge before the experience can exist at all.

The conversation lands on an uncomfortable diagnosis for PLM buyers: AI is not so much creating the digital thread as exposing how many organizations never had one. Both founders put customer digital maturity between levels two and three — a PLM holding part of the information, shadow engineering in Excel, and institutional knowledge stranded in PowerPoints and PDFs. They also push back on romantic reindustrialization, arguing the factory of the future will be software-heavy, heavily automated, and AI-native rather than a nostalgic return of manual work. On whether AI replaces engineers and planners, both land in the same place, for different reasons: the jobs get more strategic, not fewer.

What happens when the “Claude Code moment” hits engineering, supply chain, and manufacturing?

In this episode of AI Across the Product Lifecycle, Michael Finocchiaro speaks with two French industrial AI founders:

• Matthias Berahya-Lazarus, CEO & co-founder of Cognyx • Thibaut Wilhelm, CEO & founder of Oplit

Cognyx is building an AI engineering platform for hardware — bringing software-style acceleration, executable knowledge, and agentic workflows into product development.

Oplit is building an AI supply chain platform for industrial companies — using agentic planning, operational research, and industrial knowledge models to optimize factory performance.

We talk about why ChatGPT was only the beginning, how coding agents changed startup execution, why supply chain may be the perfect AI playground, and what it really takes to move from Excel-driven operations to autonomous industrial systems.

The core question: Will industrial AI replace engineers and planners — or make them dramatically more powerful?

We cover:

• Why AI shifted the bottleneck from coding to product thinking • Why Oplit thinks supply chain is ideal for agents • Why Cognyx believes hardware engineering needs executable knowledge • Why the digital thread still does not exist in many companies • What “supply chain as code” really means • Why reindustrialization will not look like old factories coming back • Why the factory of the future will be software-heavy, automated, and AI-native • Why digital maturity determines whether AI agents work or fail • Why Excel is still the hidden operating system of industry • What young engineers and supply chain professionals should learn now

Chapters:

00:00 Introduction: Cognyx, Oplit, and industrial AI 00:35 Cognyx: AI engineering platform for hardware 00:54 Oplit: agentic supply chains for industrial companies 01:24 The ChatGPT moment and the industrial AI wake-up call 01:58 From Copilot to hardware engineering AI 03:00 Cursor, coding agents, and the shift from tools to agents 03:32 How AI changed startup development velocity 03:57 3–5x faster shipping and the new product bottleneck 05:14 Solving 10x harder supply chain problems 07:11 How AI is embedded inside Cognyx and Oplit 07:40 Symbolic AI, optimization, and engineering outcomes 09:21 Oplit’s two foundational models for factory optimization 11:05 Supply chain as code 12:40 The OSS Ventures origin story 13:19 How Oplit found its wedge on factory floors 15:11 How Cognyx found the engineering data problem 16:45 Is AI exposing the missing digital thread? 17:51 Will AI eliminate junior jobs? 18:41 Why supply chain jobs may become more strategic 20:31 Why AI makes engineers more powerful, not obsolete 21:58 How to attract the next generation into industry 22:36 Manufacturing is cool, concrete, and sovereign 24:35 Reindustrialization will not be romantic 26:06 The AI factory of the future 27:04 When does supply chain get its OpenAI moment? 28:23 When does engineering get its agentic AI moment? 29:29 Deterministic industry vs probabilistic LLMs 29:58 Digital maturity from Excel to autonomous twins 31:10 Why some manufacturers are already at level four 33:27 Why engineering is still stuck between levels two and three 35:07 The customer “aha moment” 35:45 Engineering rules as executable test pipelines 36:34 When managers see the schedule update itself 37:57 Why startups may move faster than incumbents 38:47 Where to meet Oplit 39:53 Where to meet Cognyx 40:26 Closing thoughts

If you care about PLM, digital thread, supply chain planning, MES, factory automation, reindustrialization, or AI-native engineering software, this is one to watch.

#IndustrialAI #EngineeringAI #SupplyChainAI #PLM #DigitalThread #AgenticAI #ManufacturingAI #FactoryOfTheFuture #Reindustrialization #HardwareEngineering #SupplyChainPlanning #MES #AIAcrossTheProductLifecycle #Cognyx #Oplit #ThreadMoat

---

Full Transcript

Michael Finocchiaro

And we're live, as I like to say. Um, hi everybody. This is Michael Finerero. Um, very happy to be joined today by two amazing French founders, uh, Matias, I won't even try to pronounce your last name, uh, and Tibo Willham, um, of Oplet. And, uh, just want to say welcome you guys. Thanks for joining. Um maybe uh Matias you can pronounce your name correctly so I don't mess it up and you can tell us what Cognix is doing.

Matthias Berahya-Lazarus

Yeah, Matias Ber Lazarus. So I'm the CEO and co-founder of Cognix which I define um as an AI engineering platform for hardware. So we're essentially building the code for industrial products if I'm putting things simply.

Michael Finocchiaro

Awesome. And how about you Tibo with Applet?

Thibaut Wilhelm

Hi Fino. Very uh very happy to be here. So yeah, I'm Cam. I'm the CEO and founder of of Oplit, which is an AI supply chain platform for industrial companies. Um, we maximize factory performance with agentic uh supply chains. So basically our conviction is uh supply chain is full of data, super complex and super repetitive and it's the perfect playground for AI uh to make your supply chain more performance uh with less people. I was just talking to one of the uh the big uh the big four today about that specific subject. So, we'll have plenty to say about that later.

Michael Finocchiaro

Um I wanted to ask you guys, I always start this podcast with a question. Um November 2022, the world changes with this crazy open AI thing, ChatGpt. Some people are skeptical, some people are a bit bullish on it. Were you guys What about you guys? Were you guys like super this is going to change everything? like, whoa, the the pink elephants when I ask for something that's really not what I want. How how how did you guys live that moment?

Matthias Berahya-Lazarus

Well, I I'll start if you want. For me, it was definitely a a defining moment of my entrepreneurial life. So, I thought, all right, this this is going to be extremely big. I I could sense this from the very beginning because I saw tech people around me who were extremely enthusiastic. But the real aha moment came I think a year later when when we started seeing the first use cases in in assisted coding

Matthias Berahya-Lazarus

so 23 early 24. So that was the early days of you know copilot helping you uh complete your code essentially and I thought well that that sounds like a really powerful use case and it seems that software engineering is greatly accelerating and that was the what actually inspired a vision for cognix. So we thought well if that works for software engineering well why wouldn't it work in some way for hardware engineering? [snorts] So let's let's start building in that direction.

Matthias Berahya-Lazarus

So that's what happened for me two to three years ago.

Thibaut Wilhelm

Align with you Matias as the aha moment clearly was you know the happening of cursor and Novabel you know few know a couple of years ago now where we were like okay so it can really replace some people it can really make some real work right. uh and that's when we were like okay we are maybe going to go from a tool for supply chain professionals to AI agents that will uh replace some part of the job

Michael Finocchiaro

and um when um since you guys oh so just talked about code so how did it change the way uh you thought about code and the way you manage your developers right because you're no longer managing just people you're managing agents and coding agents and so forth how how did that change the way as a manager, as a founder, how did that change the way you manage a development organization?

Matthias Berahya-Lazarus

Um, for us, I mean, we've we've witnessed a dramatic increase in productivity as everyone else. So, I think my rough estimate is that we ship probably 3 to 5x faster in code. Um, so that's uh, of course, a dramatic increase. I think the the impact of on the company and how we build is that the the bottleneck has shifted from pure tech meaning writing code which was essentially a manual job um and the bottleneck shifted to product thinking. So what do we build? How do we build it? Is it even a good idea to build it in the first place? Because once you have the power to build pretty much everything in a matter of minutes or hours, uh then you can pretty much get carried away quite quick and build a product that makes no sense from a business standpoint. So you need to be very

Michael Finocchiaro

especially when cloud says that's a great idea. [laughter]

Matthias Berahya-Lazarus

Exactly. Exactly.

Matthias Berahya-Lazarus

So you need to be extremely well thought around uh how the product is actually making an impact for customers. Is that a good idea? Does that move the needle for them? Um and I think the the bottleneck has shifted to this area.

Michael Finocchiaro

Yeah.

Michael Finocchiaro

How about for you Tibbo?

Thibaut Wilhelm

Yeah. On our side, um I'll align with with Matias and also what's really changed for us, it's two things. Uh the first thing is what we are able to solve. So no in terms of supply chain and scheduling you know it's tough mathematical problems uh that you you need to solve um and in 2022 we're able to make with operational research we were able to solve simple problems I'll say and now we can solve 10x uh more difficult uh problems and we can optimize what we could not optimize before uh and so yeah as a bottleneck to change to understanding the complexity of the problem we're trying to solve and then delivering it uh is uh now quite easy and we you know we can deliver in two months what we could deliver in two years uh you know a couple of years ago so that's a dramatic change and internally also so of course your velocity increases so you know developer go faster and so on uh and we're I think what's we're still working on everybody in in the tech scene is working on is what is the right organization, right? Uh so who are the good developers? Should I have only one very senior guy uh handling uh 25 agents or uh can we still have some juniors? Uh what about product designers and product managers? Should we have still two categories or should we have only one job that both that do does both product design and product management? Um so we've done uh already some uh you know organizational organizational changes uh in our team but uh it's a working process uh and uh I think uh yeah it's a fun fun thing to to to see evolve every day.

Michael Finocchiaro

So um in terms of the implication of AI into the way you guys are building outlet and cognix is it something I mean there was the original thing we already discussed where people doing co-pilots where Matis you talked about how the at the beginning people doing co-pilots and then I think AI is seeped all the way through the stack bound to even some vendors writing foundational models. I think Eibo you were talking about having two of those. So h how is AI actually integrated into OPIT and Cognix respectively? Yeah.

Matthias Berahya-Lazarus

Um so yeah in our case we use both kind of AI. So as TBO outline we we we solve some tough mathematical problems. Some of them are not solved with statistical AI. So you need good old school symbolic AI to solve some of these things. So optimization and constraints and so on. So

Matthias Berahya-Lazarus

we use both. Um and I would say I'm I'm less interested in in uh the technology that the outcome. So the outcome is shortening the the time to market how fast you're able to design product and how fast you're able to optimize bills of materials cost design uh under under constraints and in order to do this you need to bake AI into certain areas of the product. So in our case uh it's it's it's it's a lot about how do we extract knowledge from um uh engineers and how do we turn this into code or structures that that the the machine can understand and leverage. So that's where I think AI makes a lot a lot of the difference in cognics. Um and it's less about you know just having a blank screen and a chat that you need to talk to. I think the cognix interface from many aspects is actually um quite traditional if I would say so. So you you you see your products, you see your bombs um so you're able to interact with it. I mean at the end of the day engineers need to see what they're doing. So it needs a real interface. Uh but under the hood there's a lot of AI. So that's how we think about it.

Michael Finocchiaro

And um Tibo when we talked you were saying you actually were using two foundational models.

Thibaut Wilhelm

Yeah. So actually on our side um I I feel we want to replace at the end of the day the guy uh who is using our our tool. So it's AI can be interesting you know like for him to retrive some data build some dashboards and so on but we did not push so much on this topic. We have some you know basic uh retrive information uh tell me this or this um kind of feeders but at the end we really want to optimize you know how the you know the world production is done at the end of the day we want to optimize resources we want to optimize the schedule um and so for us it's really uh deep tech how to uh put everything the scheduleulers have in their brain to optimize a factory and so yeah we have two foundational model models. The first one is you know optimization so how to combine AI with operational research which is you know a longlasting mathematical uh branch of mathematics um and also regarding uh so the industry uh knowledge model so how we capture uh rules that are in planner's head in order to automate and make the agents uh do the schedule by itself. So for me at the end uh my ideal product is no product at all is only back end and just the factory running by itself. It's going to take some years uh of course uh but at the end it's really how you capture rules how you capture knowledge uh to to run the factory.

Michael Finocchiaro

So so if I rephrase what you said you really want to have supply chain is code.

Thibaut Wilhelm

Yeah. So yeah actually [snorts] like the planners uh we want them to write code instead of applying the same rules every day. We want them to define the system how it should operate what are the objectives uh you know should I prefer uh on time delivery or should I prefer the efficiency of my factory. So we want them to define the system and then at the end the system to run by itself. Um so it's a different uh uh so it's a shift in terms of how you you you think product uh production organization

Thibaut Wilhelm

and by the way pretty similar to what we discussed earlier. So the the bottleneck is shifting to what is it that you human want to do. you want to optimize for your factory efficiency or your inventory level or you need to make a decision and your ability to to define it properly because like if I look at schedulers uh at our clients so schedulers are usually very smart people but some will be able to code the system and and define and structure the knowledge model that will run the factory uh and so they are the ones who will control the agents and some others will not have the the skills to do it. So they will probably uh do other jobs within the within the company because they are always people that very smart people with a lot of knowledge uh on how the factory works uh that you want to keep um but they're just doing all their jobs afterwards.

Michael Finocchiaro

Um that's uh really cool. Actually, it wasn't on the script, but I'm just be interesting because I find that um both you guys were created with um a French venture company called OSS Ventures, and I find the thesis behind OSS really interesting that they look after problems to solve and then they find brilliant people like you guys to go out and solve them. It would be I would just like to know the story like did John Phipe find you uh sitting in a bar or he already knew you guys? How how did that actually work to because obviously he found a problem in terms of engineering data and another problem in terms of supply chain and manufacturing. So how did how did that actually happen for you two guys? [snorts]

Thibaut Wilhelm

Um on on my side so yeah it's a fun story. So um with Sufan so my co-founder we both knew that we we wanted to create a company on my side I had like some few iterations and I was discussing with with Rena and he told me okay next week we work with Sufan we are going to travel around in the factories do you want to come with us and I was ah yeah maybe do you have something better to do I was like no you know I just resigned from my job so maybe the next two weeks I can uh you know just travel around. uh and I felt in love with uh the passion people have in manufacturing. Uh they are really people that uh love their job that are very uh skilled um and uh and they are like really pragmatic. So I was like okay I want to I am really happy to work with those people. I knew manufacturing quite a lot because I was working in this area but I was not sure I wanted to create a company here. But uh that's how I fell in love with manufacturing and regarding production organization at the at the beginning we really wanted to to tackle something different which was um middleware. So how we degraate integrate data from ERPs and uh by chance I'll say some people told us yeah we want to integrate this data and this that data to make this super complex scheduling uh product and the fun thing is that when I was working for Mckenzie uh I did several projects where I was the one building some crazy Excel files to optimize the scheduling of factories. So I was like oh this rings a bell. Uh so so this is why we uh we're like okay let's go uh all in on uh how to organize factories and and the world's production.

Michael Finocchiaro

Awesome story. Thank you. So that was like walking the factory floor. Matias, same thing for you or you were watching to an engineering organization and seeing the Excel spreadsheets rather than the the PLM systems.

Matthias Berahya-Lazarus

Yeah. So my uh my story with OS venture started um when I met Ren. So I actually uh reached out to him when listening to a podcast where I found that his thesis around manufacturing in the west was [snorts] wrong really really true. uh he was already on a mission to you know um help the west rebuild an industrial capacity and I thought this was extremely important for um u France and the west and the US in particular. So um I think the mission rung uh really true to me. Then I met with we bounced a few ideas um and and found that the biggest problem we could think of that um were surfaced by manufacturers was around their engineering design and development processes. how slow this was and how this was true completely across the board in automotive, aerospace, uh consumer goods, um mobility, you name it. They all were complaining about how slow it is to reconcile tons of documents that make no sense. Um so that's how we started. We thought this was a really tough mission, but the problem was huge. And as an entrepreneur, you need to fall in love with the problem, not so much the solution.

Michael Finocchiaro

Uh, good one.

Matthias Berahya-Lazarus

That's how we started building in that direction.

Michael Finocchiaro

It's actually uh my friend Jason Casper in the chat is asking a good question. He says um is AI making the digital thread more important or is exposing many to how many organizations never had one in the first place?

Matthias Berahya-Lazarus

I think many organization don't have one. Um and now AI and new technologies especially around databases uh knowledge graphs and so on are indeed the capacity that will unlock this. So there's a great and very strong opportunity to be building that now.

Thibaut Wilhelm

Yeah. Uh for for us um AI is a big opportunity to make industry more competitive. That's as simple as that. Uh so it's just for me it's just uh continuous improvement journey of digital and and the impact on on manufacturing. Um there's a real game changer is also like the ability to capture knowledge and navigate through it. Uh so really replace a human and not uh give tools to human.

Michael Finocchiaro

So let's stay there just for a second. Um, it's probably worrying for a lot of the younger people listening that don't have years of experience. They're like, "Okay, so these guys are after killing my job." So, what kinds of things should the younger generation because the demographics of the people watch my podcast, there's about 23 to 32% depending on the podcast of entry level people or or maybe some of them are actually, you know, just graduating or grad school. And they must feel a little scared when you say, "Well, our job is basically to eliminate your job." So, how uh what should they be doing so they're not on that list of easy to eliminate uh don't bother hiring like how what what are the skills they need and why should they come and work for Cognics and Outlet instead of going to somewhere boring like Accenture or Microsoft?

Thibaut Wilhelm

So, so those are two two two different questions, but for for the guys for for the guys in supply chain, I think I I will make I mean OP and and the AI wave will make their job more interesting because they will really be able to extend the scope and the impact they can have on on a company and on a factory. You know before before you know the guys were just saying like okay after apple pie should I do strawberry pie and then banana pie and so on. So every day the same questions and so on. No they can really control the factory and say okay how do I increase the throughput of my factory? How do I deliver on time my clients? So you really are focused on uh you know the the output rather than really doing some constraints every day. So in that in a sense the job is like really like 10 times more interesting. So you just need to really understand the basics of supply chain, the basics of production and then be someone super structured. Uh and it will help you you know master your job for the next 20 30 years because at the end you will always have some strategic decisions to make. uh you know it's always trade-offs that you do some arbitrage between uh you know between your clients basically and the competitivity and the productivity of your of your factory. Um and why should people join us? uh you should join us if you like to optimize stuff. Uh for example, you know, the morning I I I go to work uh running and I listen to a podcast in the same time or I do a call uh every time I'm thinking in my head, should I, you know, start my coffee now and do something else in the same time. So if you're crazy like that and like to optimize stuff, you should you should join.

Michael Finocchiaro

I love it. for you Mats.

Matthias Berahya-Lazarus

Yeah, I I would have maybe a slightly different opinion on on whether Cognix is going to replace engineers. I I don't think this will be the case. Um I think engineers will definitely be a lot more productive in the sense that they'll process probably a lot faster and a lot more information probably for a lot more products. products might be a lot more complex so they can handle more complexity with the right software tooling using of course a lot of AI. I think it's pretty similar to what happened maybe 30 40 years ago for accountancy. So if you introduced Excel to accountants 30 or 40 years ago they might say okay this is the end of my job. The reality is 30 40 years after there have never been you know there still a lot of of accountants and probably a lot more than 20 30 years ago because now finance has moved not to accounting but to controlling and and financial projections and so you can handle a lot more work with the right tooling. So in that case Excel and of course many of the SAS and AI that came after this. So I think that's that's more the direction that I'm seeing. uh using Cognics uh manufacturers will be able to handle a lot more complex products uh ship them a lot faster with a lot more consistency and and a lot more reliably and that's what I hope will make engineers life a lot more interesting

Michael Finocchiaro

and just as we're talking about students there's a student from HC that just wrote to me okay

Michael Finocchiaro

chat uh Lewis Kola Michelle and he says um a question about re-industrialization in the west from Matia Centibo how Can we attract the next generation of workers to this sector, especially in France? In other words, not all moving to California or Boston, right? Um, industrial jobs have been devalued because they're seen as dirty old jobs is a question that would probably require 10 episodes, but what advice do you, Matias and have to policy makers um in order to keep the jobs here?

Thibaut Wilhelm

Um yeah so for for me it was it's one of the reason why we created optit uh so opit it's called it's the contraction of operations lit so the the story behind that is to say you know manufacturing industrial uh uh sector it's cool it's lit you know uh and we want to showcase to everybody that there are some fun problems to solve there are some really concrete uh things that are being built uh and when you go to a chocolate factory or when even when you go to a fasteners factory, it's really fun to uh to see. Uh so this is really embodied in the in our uh company to say to to people look manufacturing is very cool. Uh you are going to learn a lot of stuff uh and you can really be passionate about it for the next 30 or 40 years of your uh of your life. So this is what we we we try to to to scream to the world but yeah it's tough. Uh so for example I remember so I did an engineering school school called central in in France. Uh and uh I think we were like 300 in my um in my division and maybe like five to 10 people went working for industrial companies. Uh so we need to increase that number. Uh for sure maybe less consultants and more guys working for industrial companies uh because at the end uh that's uh that's what is uh important for your country. Uh you know you need to to have some sovereignty and some ability to uh to build your own stuff right uh otherwise you just consume what are done by others and uh and you're stuck.

Michael Finocchiaro

So Matias, what advice do you have to the policy makers to keep that jobs [snorts] here?

Matthias Berahya-Lazarus

That's tough.

Matthias Berahya-Lazarus

Yeah. So besides lowering taxes, of course, that's like number one. [laughter]

Michael Finocchiaro

Increase taxes. Yeah. [snorts]

Matthias Berahya-Lazarus

Especially in France. Yes. Uh well, I I would say that we have no choice but to re-industrialize. I mean, if you're looking at the geopolitics today, um we have no choice. Uh we need to ramp up industrial uh industrial jobs and and industrial capacity in the west being able to make our own things, our own cars, our own weapons, our own everything because the reliability of the supply chain is being questioned every day as the world is getting more crazy. So being able to to to make industrial and design our own um products uh becomes pivotal for the company. Now I would advise uh probably policy makers to stop having a I would say a romantic idea of reindusterizing the the west. So you might be thinking oh we'll have the small factories in our countryides and this will be nice. We have workers you know doing everything by hand. No you have to get rid of that idea. This is not going to happen. uh the the factory of the future will be extremely technical with heavy investment, heavy automation. Probably 10 15 20% of the cost of a factory will be into software to make sure that it's automatized enough and that you have a lot of leverage from software on the hardware being made. So I would definitely view um manufacturing as a heavy investment area. It's not coming back like 50 years ago.

Michael Finocchiaro

Um, which brings me to the last question for this section. And it's interesting because you mentioned the AI factor of the future. I remember last year all of Jensen's GTC GTC talks were about this AI factor of the future that every physical factory will have an AI factory and every AI factory will have a nuclear power plant, preferably fusion rather than fision. And I think that you guys, the startups are all building like one brick at a time. You're building that AI factor of the future, right? So at the same time, we saw an open AI moment for programming, right, with cloud code and and cursor. We saw the open eye moment in 2022. I'm not sure we've seen the open eye moment for supply chain and engineering. Or maybe you I you guys made it and I didn't see it. But how long do you think before we have that there's a before and after? Because clearly with open AI and with cloud code, there was a before and after and nothing will ever go backwards again, right? So when do you think we're going to hit that in your respective areas?

Thibaut Wilhelm

I I think on on our side I think the it's going to come in in the next months really uh at this moment uh you know for for a supply chain really why because it's so it's a problems that are full of data super repetitive quite complex and for one problem you have one solution that is obviously better than the others. So it's really good AI playground. Uh and uh we see that a lot of startups are trying to solve this. We're trying to solve this. Um also the you know the tech giants like uh SAP uh and similar companies uh are working on agency planning. So a lot of uh efforts are are put into uh into this. Many people think uh it's going to happen uh by 2030. Uh so yeah, this is 100% sure that by 2030 nobody uh is going to uh plan using an Excel file uh in you know in advanced factories. So

Michael Finocchiaro

take you as as a bullish very very bullish. Okay. And Matias

Michael Finocchiaro

as bullish in terms of engineering or less so [laughter]

Matthias Berahya-Lazarus

exactly super bullish. I mean what I think what you're seeing is that AI has not been productized yet for the engineering use case. So we of course building very aggressively in that direction. And the question is uh how soon uh can we actually have the whole experience working but in order to do the like I said the clo code for industrial products you need a lot of of framework before you can do this. So you need of course for all engineers to work on a single source of truth uh which is not the case today because they're working in different documents. You need um engineering to be composible so they need to build from existing modules so that humans and AI can understand what they're doing. You need knowledge to be executable so it needs to be understandable by again human and a machine. And once you've done this then you're able to do full agentic engineering. Um, but you need to have a lot of scaffolding before you're able to actually deliver that experience. Similar to clo code,

Michael Finocchiaro

I was about to say I think that the the real the bottleneck is the fact that in both supply chain and engineering manufacturing, we're extremely deterministic. I mean, obsessively deterministic. And LLMs are unfortunately obsessively uh uh probabilistic. And so that's the rub, right? Is how how close can we get to determinism with the guard rails with all that kind of stuff. Um, so I wanted to finish because I know uh uh uh Tibo has a a deadline. Um I I I'd like to talk at the end about digital maturity in general. So I um you know we we we've been talking about Excel a lot on this call and that's sort of the basis of my question where I think about when you go when you go to your customers when you go to the customers of Cognix and the questions of Oplet I think of digital maturity sort of on a spectrum from one to five like one they're still using teams to communicate and they're still using Excel for almost everything else. Okay, they they bought Windshill. It's 3 3x but actually nobody uses it because well Excel and then on the five, you know, you're in fully agentic adaptive uh uh autonomous digital twins and basically nobody's at five, right? I mean there might be one division of SpaceX that might be there but it's pretty much La La Land still. So what is your perception when you meet the customers? I I I bet because I can't ask a customer that a customer is going to say hey I can't say that look like an idiot. You guys without naming the customer can say well I feel that they're closer to one closer to what is your impression of where they sit on that spectrum.

Thibaut Wilhelm

I I think there's a lot of uh heterogenity. I'm saying something that is very obvious but uh clearly some companies are at two and some companies are at four. Uh

Michael Finocchiaro

four really?

Thibaut Wilhelm

Yeah. some pretty I mean if if we want to say for me four is

Thibaut Wilhelm

uh I have structured all my data my data makes proper sense uh I

Thibaut Wilhelm

have a co I have a a data organization with a co on top okay

Thibaut Wilhelm

yeah and I can extract and use any any data and sh and share it right and I'm I'm starting to play with AI agents that uh that's have value right so that will be for for me um and we see some companies that's uh are able to to to make it happen. Uh usually it's more like big companies that a few years ago uh you know like took a bet and said okay I'm going to have a proper data lake whether it's a data lake or it's structured within my ERP but I will have this one single source of truth and I will have internal people that will be developers that will help some suppliers. I'm going to have a lot of uh tech suppliers. Uh and so those some companies took this bet and know I think they're quite advanced and some companies you know it's a complete a complete mess. But to work with us it's easier and you will capture way more gains uh if you start from three or four rather than if you start from one or two. So we have some other products in the company for you know like if you have a lower digital maturity and you just want to execute uh easy stuff but if you really want to have AI agents an AI brain that you know like schedules your factory that pilots your factory you need some proper structure uh to capture those gains uh and you need the data and you need the knowledge uh the technology and the supply chain and industrial knowledge.

Michael Finocchiaro

Thank you, TA. Wow, it's the first time Nvidia said that there was so many fors. That's really Matias. I'm just guessing here, but I'm I'm going to guess that there's not as many force in in our area of PLM and engineering, right? [snorts]

Matthias Berahya-Lazarus

Yeah. I mean, when we walk in, there's usually the the same kind of maturity everywhere. So, sitting between two and three, I would say. So, they usually have a PLM with part of the information. and they usually have a lot of shadow engineering on Excel. Um, and they they usually have many standard operating procedure and many knowledge uh in powerpoints and PDFs documenting how things should be met. So that's the status when we come in and what I can tell you is that every board right now has this conversation. So, how how fast can we embed AI into our most critical workflow because we need it for competitive reasons or for uh efficiency reasons or or or for bottom line reasons. Um, so they're all having this conversation. I think at this moment they've tried some things. So, they've they've tried some maybe proof of concept uh maybe with a few copilot here and there and realizing that this is not making the cut. So they need they need to be uh leaning in a lot more heavily and probably that means investing into the right tech but also investing in in human transformation. Uh so how do you change the people that are going to work with that technology because once you introduce such a powerful and transformative technology as AI you need to rewire the processes and essentially the human brains around it and that's a lot of work. So uh in our case, we're here to assist that. So we deploy specialists along with our technology to make sure that this transformation can happen. Uh but that's not to be underestimated.

Michael Finocchiaro

I appreciate that answer. In fact, that so the the next to last question, last question is just a goodbye question. The next question is um have you seen when people have t have uh we're at three or four well more like two or three and they use cognix or they use opletit is there sort of an has there ever been an aha moment like oh my god if I you know if I broke the data silos if I got the data out of the old files if I got different departments talk to each other because they don't talk to each other my god what I can do all this great stuff and I can have better ROI and get product out there have you have you seen that aha moment uh in your customers eyes.

Matthias Berahya-Lazarus

I love this question. I definitely did. Uh and especially the customers that have this aha moment are usually hardware engineers that can also somehow code software. So they're software developers maybe on weekends because they like do code things and so on. And so when we show them for example how we translate an engineering or manufacturing rule into code so Python or CL in our case and how this can be applied to their model subsequently they're like oh so this is like a test pipeline a CI/CD was like yeah

Matthias Berahya-Lazarus

oh that's amazing so then I can accelerate because I know it will test. Yeah. Uh and so that's the haha moment for them uh because they get the system thinking that is so powerful in in software engineering that they can transpose to hardware

Matthias Berahya-Lazarus

and

Michael Finocchiaro

awesome. How about have you seen that too?

Thibaut Wilhelm

Yeah, for the aha moment I feel is more for the managers maybe or persona so schedulers are they're geeky but not as geeky maybe as you know like engineers cognix are are working with so they they more see like okay it's going to be automated they challenge oh do you take this rule into account and so on so they really more into like making the thing work but uh uh they do But this I do not feel this big aha moment but once the manager uh you know like the supply chain director the production director says all right but like the schedule has just updated by itself.

Thibaut Wilhelm

Yeah. Yeah. That's that's the whole point right that's how it works. uh then it's a big uh big aha moment because like it's like a senior developer uh who understand okay now I can control uh 10 junior developers and they are going to do exactly what I told them to do. Yes. Uh oh yeah this is cool. Uh so now I'm sure that my scheduleuler is going to the schedule is going to be aligned with my strategic goals. Yes. Because you're going to be the one who will define the rules and the rules are going to be respected. Uh so yeah, the aha moment from for us comes more from the middle managers. I'll say

Michael Finocchiaro

that's a fantastic answer. Thank you. And I I think both of those examples really point to the my central thesis of some of these calls which is that if you want to move the dial towards that five waiting or for SAP or or the other big three to do it is probably not the right call. It's probably better to go to Matias or or Artibo and and uh use their their much uh faster software, right? Um I appreciate you guys being on here. Just before we go, I just wanted to like uh we're going to go on our summer break, but maybe in the fall, are you guys going to be at some trade shows? Maybe I just saw the first invite from Adopt AI. I don't know. I think I you guys were there last year. Um what's up? So, how folks that are coming to Paris or maybe you guys are traveling internationally, where can they meet you and and see uh see your solutions and demo

Michael Finocchiaro

besides online of course

Thibaut Wilhelm

they can come to uh to to office so which is based in Paris. We're going to open office in the US very soon. So, I'll be happy to to to meet anyone physically. Uh and yeah, we we cover a lot of trade shows uh especially aerospace luxury and and batch process. So uh if you're on a trade show for this type of of subject uh you just can can join us and we organize a trade show for aerospace uh in uh in September in Bordeaux in France. Uh so we are uh

Michael Finocchiaro

do you need guest speakers?

Thibaut Wilhelm

Um uh yeah but it's really focused on aerospace uh and so the so we're working with the JFAS which is a Korean industry

Thibaut Wilhelm

so consortium uh and so it's optit and the jifas working together on digitalization for aerospace so it's 22nd of se September.

Michael Finocchiaro

Awesome. How about you Matias? Where can we meet you and uh Cognix?

Matthias Berahya-Lazarus

Yeah, so far I haven't been a great party guest in in those conferences. I must admit I've been very busy at clients, you know, on factory floors as as TBO said. Uh making sure that whatever we build is actually used on the ground. So that that was most of my time for the last months and years. Uh but I I will be at Adopt AI and uh we will also open in the US uh I would say in the coming months. Um, so you will definitely see us um in in Paris or probably East Coast.

Michael Finocchiaro

Awesome. Well, um, in the interest of finishing exactly on time, I wanted to say thank you very much to both of you. Um, of course, there'll be another credit conference and you guys will be obviously privileged guests if you if you wish to attend. Uh, to everybody else, there'll actually be another podcast with Texaw 3D in about an hour from now. Uh, thank you very much. We'll we'll be back. I think Not sure going to get any more podcast in before September. We'll probably be back in in um in the fall. But uh thanks everybody and and a big thanks to Tibo and Matias for taking your time today. Thank you very much guys.

Matthias Berahya-Lazarus

Thanks a lot. I had a blast.

Thibaut Wilhelm

Yeah, it was a pleasure. Thanks a lot, Fino. And have a great holidays.

Michael Finocchiaro

Cheers.

Matthias Berahya-Lazarus

Cheers. Bye. Bye.

Thibaut Wilhelm

Bye.