AI-Driven Product Usage Analytics and 3D Intelligence — with inUse and SP3D | AI Across The Product Lifecycle | DemystifyingPLM

Episode Summary

The episode titled "AI-Driven Product Usage Analytics and 3D Intelligence — with inUse and SP3D" delves into how artificial intelligence (AI) is transforming product lifecycle management (PLM) through innovative solutions from two French startups: inUse and SP3D. The conversation, moderated by Michael Finocchiaro, features Laurent Couilland, founder and CEO of inUse, and Olivier Mathey, CEO of SP3D. InUse specializes in AI-driven field service analytics, enabling companies to optimize their maintenance operations through predictive insights derived from IoT data. SP3D focuses on 3D printing and digital twin technologies, specifically with its product Teya, which offers a full lifecycle management solution for spare parts.

The episode highlights the rapid adoption of GenAI by these startups, emphasizing how it has revolutionized both internal processes and field service approaches. Laurent Couilland underscores that AI on data provides unprecedented access to advanced analytics, significantly enhancing development capabilities and operational efficiency. Olivier Mathey shares insights into the transformative impact of AI on small companies, noting a substantial increase in their development teams' productivity due to enhanced coding abilities.

For PLM and engineering professionals, the key takeaway is the pivotal role AI plays in driving innovation and efficiency across product lifecycle management. The episode underscores that embracing AI can lead to significant advancements but also highlights the importance of strategic planning to ensure equitable access and mitigate potential social disruptions.

Full Transcript

Michael Finocchiaro

So, hello everybody. This is the earliest version of the ⁓ AI Across the Product Lifecycle podcast. My name is Michael Finocchiaro and I'm joined by Laurent Couilland founder and CEO of InUse and Olivier Mathey who's the CEO ⁓ SP3D. Two exciting French startups. So it's fun to have two French friends because I'm also a French citizen So what maybe we start with you long you can give you a little bio how we you and I met which is a long time ago and then a little on menus

Laurent

Yes, some time ago. Thanks, Michael, for this session. My name is Laurent. I am the founder of e-news. As you said, I'm an engineer. I've started in a company that is well known in the PLM ⁓ area. I spent 16 years in the system in various jobs, organizing the ⁓ wind of Airbus and then transforming the channel from IBM to resell the channel all across Europe. And lastly, I finished by leading ExaLead, which is all the NetVibes type of systems that exist in Indus system. 10 years ago, I decided to jump into the ⁓ new world of IoT and AI. And I think that's what we're going to discuss now.

Michael Finocchiaro

Yeah, exactly. Thank you, Olivier?

Olivier MATHEY

Great. Thanks, Michael, and thanks for inviting me. On my side, I'm married, I'm father of two, and I spend almost 20 years in a French group called Valoic. So in the industrial world, mostly focused on the energy, where we supply pipes for the major engineering project across the world. And four years ago, while I was in charge of the Geotorx transformation for Asia, I met Spare Part 3D and I... I decided to switch from being an entrepreneur to become an entrepreneur to see what was the difference. And now I can tell the difference. I joined them as a head of sales. And since one year, I'm now the CEO of the company. And we'll discuss a bit later. And we did that shift when we decided to accelerate to the launch of our second product, Teya, which is the full OVA.

Michael Finocchiaro

Awesome. Thank you for that. So we wanted to talk today about AI. And of course, this is the subject on everybody's tongues, as we say in French. And I wanted to understand, like when a JIN AI first came out in 2022, and it was kind of a new thing. Did you guys immediately see the potential? Were you sort of ⁓ skeptical? Because I think people had varying reactions when it first came out. Whoever wants to answer first, but I would like to get your opinion on that.

Laurent

If I take the view there, when Geni went on the market, at least when it was accessible by companies like us, ⁓ of course we jumped in right away. Why? When I started the company in 2015, we discussed a lot about IoT and AI on the data. What I call AI on the data is the AI that works from the Internet of Things, assembling a lot of data, making machine learnings, all this deep learning and all that. Adoption was quite long and working on that. Gen.AI makes a new paradigm there. It gives the access of a fabulous technology for ⁓ more people than ever. ⁓ And that changed the game. It changed the game for our developers, of course, when we develop code. And it changed the game in the sense where we are working with the news on the field service approach for the people on the field in order to capitalize the knowledge there.

Michael Finocchiaro

Fantastic. How about you, Olivier? Were you just as optimistic as Laurent?

Olivier MATHEY

When it started, we looked at it with two sides. The first, think, quite similar to what Laurent said is how can we use GNI to accelerate or improve our internal processes. ⁓ At the beginning, we were disappointed by the first results. Now we're using much more. And definitely our development team is basically empowered. And we are a small company, are 12 people, but our development team is basically two times or three times what it was a year ago. because their ability to develop code, the ability for our engineer that are working with our AI experts, but that don't have the knowledge of coding, the ability for them not to code and accelerate the process is quite phenomenal. And on the other hand, for our own proprietary AI that we develop, we look at it as a potential threat. So we keep an eye on the capabilities that are developed. Luckily, there is a good gap between what is available on the market and what we do. And so today we believe we still have some barriers in terms of technological breakthrough that we went through that this type of technology are performing. So we are using other type of technology of AI. is big families. But the GNI today is not a threat for us. It is a great tool for all internal efficiency and still something that we keep in our watch.

Michael Finocchiaro

Okay. Actually, might be good because I'm not sure that the audience is familiar with the two companies within use and SP3D. So maybe you can give us a very brief overview of what you guys do respectively to. You want to start Olivier with SP3D?

Laurent

Yeah, I would have you.

Olivier MATHEY

SP3D, our mission is simple but quite ambitious. ⁓ We want to facilitate the adoption of digital process by the manufacturing. We are helping industry digitizing their inventory. So looking at their catalogs and identifying what could be digitized to be produced on demand, produced with reprinting, know, this distributed manufacturing concept. And as we work with more customers, we face a major bottleneck that was that transforming a 2D drawing into a 3D model was something that was manual and largely, even though it sounds like a technical detail, it's actually a huge pain point. And so four years ago, we started another R &D program, which ended last year and we have now a third one solutions on the market called Taya where we automatically reconstruct 3D model from a 2D drawing. So this is our two solutions. So we are really industrial DNA. and we've been the digital and try to simplify the adoption of digital tools.

Michael Finocchiaro

Probably a lot more exact than companies like Zoo that's just going from text, Since you're starting from a 2D plan, which is relatively exact.

Olivier MATHEY

Yeah, which is in fact, it is exact, but at the same time, there is a lot of ambiguity on the 2D drawing. If we, the three of us have to do the 2D drawings of the same component, we will end up with three different 2D drawings with the same concept. and this is the difficulty that we are raising thanks to AI. So, ⁓ deep neural network graph neural graph network to be able to basically compensate and being able to replace the human eye.

Michael Finocchiaro

course. Hmm. Excellent. I like that. How about you, Laurent? I think yours is more tied to the machine and the machines talking to each other,

Laurent

Exactly, what if machine could talk? That's our motto there. What we're doing is basically we are trying to mitigate what I call the industrial amnesia. It's about ⁓ a lot of people going on retirement in the next years, but we count about 7 million people going on retirement in the North America, Europe and Japan, which are our three regions where we are operating.

Michael Finocchiaro

Mm-hmm.

Laurent

And ⁓ if we look at those people living, there's no recruitment at the speed of replacing the people that are living. if you do not capitalize the knowledge, if you have no replacement of the people, you will face an industrial amnesia, which will basically cost up to $500,000 per year per factory or something like that. mean, it's a huge, huge gap there. So what we're doing there is at the news, we build a platform that helps to capitalize that knowledge on two fronts. Originally, we do it from IoT and AI on data. So basically, we interrogate the people from the technicians. And we'd say, how do you solve those problems? And by interviewing them, we transform what they said to us into algorithm. And we build that inside, faithfully, inside a machine. So the machine can talk and can identify what will be the problem. And the second coming since 2022 is now we are also increasing that with the generative AI, where we ask the people to capitalize on what they do by text, by interventions, and all what they are a little bit ashamed ⁓ or they don't want to do it because they don't know why they would do that. And I think the GenEI is giving them the why. They do the why because now they can reuse that very easily everywhere in the world. So we do it both sides to capitalize. And of course, the agent side, which is the merge of the two, it means that you can ask the machine, how did you produce better this week than last week? And what are the reasons and what should be the action to do? And that's the merge of the IoT world and GenEI world into one.

Michael Finocchiaro

So it's, you really did the whole shift from machine learning to generative AI to, that's really cool. Do the agentic stuff. It's a good story.

Laurent

Exactly. And agentic is the convergence of that. And as of today, we have more than 33,000 machines around the world that are on our platform, more than 5,000 users. So that makes quite a significant move on that, even though I consider it's still slow to get those adoption.

Michael Finocchiaro

Do you have to tell them to get off of Snapchat sometimes? They're not like kids, you know? So I'm imagining that, ⁓ actually both of you already talked about the fact that when you develop the product, your developers are using AI. I I do this on every podcast, there's still a lot of mystery around is AI really great for coding? Is vibe coding a thing?

Laurent

Exactly.

Michael Finocchiaro

⁓ And yet, Lovable, I think they just said they made 100 million. I think their fourth release hit 100 million within like four weeks or something, which means it may not be a thing, but it's something that's generating money, like real money, not just investment capital, right? So how does it work for your developers of Inusent for SB3D in terms of, it? Do you use it for the brainstorming? Do you use it for the actual scaffolding? Do you use it for creating the test bed? How are you using the AI in general? Or I suppose in Laurent's case, you have a lot of the regression analysis still, but it's done by AI rather than ML. And Olivier, you have a lot of ⁓ inference of the geometry that would be created from the 2D. So it's interesting to know how you're using AI to achieve these amazing things in the software.

Olivier MATHEY

Yeah, and if I start to answer, we have, I would say two, two, two usage that are pretty different. And the first one is the, have an AI expert team. have two PhD in AI on different families of, of, of technology. We have two additional AI engineer and, and there are them, these guys are building the core of TEA. So the two-day to three-day reconstruction. We don't aim to be the expert like, chat GPT can be or Google can be or Amazon could be in AI because of course their capabilities and manpower is way beyond our reach. However, we have developed the ability to combine the right technology at the right process. So we have a lot of algorithm that goes one and work one after the other up to 30 to make these 2D to 3D reconstructions automated. And the challenges have been to combine the AI technologies and technology with the S with the traditional computer vision. So this is read-off-first application. And this is, I would say, our own IP, our proprietary IP. And that requires dedicated skills, dedicated people. And to be honest, they are not allowed to use Gen.AI to develop their own code because it has to be our black boxes, IP. What's very interesting with the different tools that have been massively being simplified in the last couple of months is not only have these four guys, but the rest of the companies capable to support them. We have a team working on the first product and they are almost self-autonomous in developing the solution today. We worked before on AI, we have machine learning on the database. We work with a lot of data. We have a lot of additional tools and simulation and process which require. the IT team, the development team to step in. And now they are almost autonomous to develop the product. So we have basically a product that we are capable to develop with non-experts. And we have our experts, which have a red line in using external technology.

Michael Finocchiaro

Very cool. Interesting. About the year long.

Laurent

Very interesting, Olivier. I'm encouraging AI ⁓ regarding my team. I will make the differentiation between what we do internally and what we do for the customers or at the customer side. Internally, I'm encouraging AI at every front because I consider AI in the mode of for the code, for test, for whatever here is ⁓ speeding up the development. So it's a question of over there. And I like my team to use cursor or other solution that help them to do better. I think the main approach is not to develop the code by itself, it's to accelerate what they do. And I think it's in that way that it should be taken. ⁓ It's an accelerator, it helps to do some tests. it has to do some review over there. Anyway, we have not changed the ⁓ review code processes that we were doing before. But what we can see, ⁓ since we put that in place, is that we accelerate that. And as we've got a release every two weeks, so we've got sprints of two weeks, we really want to make sure that in those two weeks, we increase the level of delivery and the level of quality during that. And for me, it's speed over there. Gen.AI is about speed and all that. We use that also in the marketing side. ⁓ We should not forget that there's the development, but there's a commercial development to make also. And it does change a lot in the marketing side. How we can generate content, how we can communicate, how we can define that, how we can do market surveys with all the show for Mistral. We're using a lot Mistral on our side as being part of the French web.

Michael Finocchiaro

Yeah. Big time.

Laurent

⁓ being able to make some survey into Le Chat and analysis that was, again, taking a lot of time before. So I think internally, I'm really encouraging what we can do there in order to increase. Then on the product side, my philosophy is a little bit different. I tried to make very precise and useful AI. I'm not doing any toolbox approach there. I'm really doing an applicative side. in order that we target a persona, which is the field service guy. And we want to deliver to him the right AI for his job, and not making him the ability to build his own AI. But we want to deliver for him the AI. And I think it's a population that is likely to use AI, but not likely to design AI or design their own AI. And that's where he uses fitting, is we really want to make agentic AI. that fits the world of ⁓ field engineers, field technicians that accelerate the repair of machines. So you talk about what if machine could talk. The effect is, again, speed is how much the machine can restart faster, how you shorten the stop of the machine, which is huge euro or dollar, or whatever you count on that. And at the end, the impact on the top line of those companies is huge.

Michael Finocchiaro

Hmm.

Laurent

of course, with a level of adoption that Gen.AI is giving, which is higher than ever compared to all the machine learning and all those algorithms which were expert for expert. Here, can tell you we've got examples where we've got people, operators next to the machine that are using that as a natural way to help them doing their business and they like it.

Olivier MATHEY

Yeah, I think in what you mentioned, you mentioned speed. There is another aspect of time, which is very complimentary. What you said is the, and you mentioned it at the end is the ability to focus on what matters. The operators should not develop AI. It doesn't have the skill and it's not the job. So, but however, being capable to use these tools to give him the simplification and giving a tool that works for him, it's simple to use. I think this is what has been accelerated.

Michael Finocchiaro

Olivier MATHEY

by the general TBI of the last few months, because now everybody in the company is capable to use that tool by asking questions, doing the no code, doing the QC review of the code. And the way I encourage my team is that the time is saved, just make sure you focus on the pain of the customer. Because at the end of the day, and we launched the day on the markets, we had these 2D, 3Ds automated, it's a great technology. And when you talk to your customers, sometimes it works, sometimes it fails. Say, hey, fail, can you do it? Because I need it and I love your interface. I don't care if it's permitted. Is it the same price? Yes. Okay, go for it. I don't care. Just make it, I need it in two hours. And these are tools. What matters is I have a problem. I'm on my city, I'm losing what you said. People are retiring, how do you manage? They don't know how to capitalize. This is the key point. And then how could you simplify? and that's an amplifier in

Michael Finocchiaro

you

Olivier MATHEY

fact. This is what matters and I think they are great tools for that.

Laurent

And you're right. And I think in the context of what you said just before and what it brings is also the level of quality. Because you said that if we would have generated 3D from 2D altogether or vice versa, 2D from 3D, we would have done different output on that. And I think one of the ways to control also the quality together is the ability to ⁓ organize that through ⁓

Michael Finocchiaro

Laurent

AI, which not only speed up, but also control the quality. We see that in our side. When you've got technicians, he has done his daily job. He has repaired the machine. It took him about eight hours ⁓ on site with the stress and the pressure from the customer to restart faster the machine and so on. So the guy is stressed. And what the boss is asking is now you need to take two hours to make a report. And the guy, what he doesn't like. is he doesn't like to write, doesn't like to make the report, he doesn't like to copy paste the pictures he has taken from his phone to the Word document to read, to write something in the language and so on.

Michael Finocchiaro

No.

Laurent

So what he has to do now is press a button and what I mean he may not be a post or whoever as a ⁓ writer, okay, but at the end it will be the correct quality of the report generated in few seconds and what he has to do is to review it's ⁓

Michael Finocchiaro

haha

Laurent

We have to train people on what it is. It is still an assistant. It's not generated. And then you can send that directly to the customer. You have to read it back. You have to correct it if needed. But at the end, it's the quality. And we see that everyone can write report at the right level of quality expected by the company on the easiest way. And I think, I guess, for your case, 2D to 3D, it's the same way. You can control the quality of what's generated.

Olivier MATHEY

Yeah, I fully agree. We are working in an industrial environment. Safety first, then quality. And the quality of the product or the process is the safety of the people using it. When you talk about AI, are, know, it's a black box. We're not too sure what's going on. Sometimes it's statistics. You can't choose. know, when Google makes mistakes, we don't care. You receive the wrong advertising. Okay, fair enough. That's already paid five cents extra. If you send the wrong equipment, you brought the machine, you can kill somebody that's different story. I think the challenges of companies developing tools or software for industrial application like we do, quality is the most. It's a right now. You don't deliver quality, out of the market. And I think this is one of the great challenges for all development team and for on our side, we do a of collaborative work where we have AI or algorithm, but also engineers.

Michael Finocchiaro

you ⁓

Olivier MATHEY

and talking to each other to make sure that the quality is there. Then we talk about the lead time and of course at the end of the day we talk about cost.

Michael Finocchiaro

I think that in fact when we talk about Copilot, we should always be talking the plural. There should always be a whole committee of them to make sure. And I had this idea of having a Mamminities agent who's always doubting everything. No, no, no, you're not the most beautiful guy. No, no, no, your ideas aren't all great. In fact, I really don't agree with this one, you know, rather than just telling you how great you are. There's actually a question in the chat from a guy that listens a lot to this podcast, Steve Klein. It was a question for Laurent. It says, is in use supporting servitization, which is one of the most important values to create network effects as described by Ben Alstine. I never read that. Connecting intelligent equipment to not only the customer, but also customer to customer, supplier and others, which is described in another book he talks about. This plays an important role. John Deere was one of the first companies to realize this.

Laurent

Yeah, that's the aim at the end of E-news is the servitization. I like to talk about servitization and ⁓ the derivative. It's how we make the commerce and the interaction of the know-how. If you've got know-how along all the supply chain, you can change the business and you can change the model everywhere. And definitely, that's the... That's the older servitization wave on that, that we see and we talk since the wine is how on top of a product, if you are a machine maker, how on top of a product you can build services and change your business model. We see that for the common way for cars and for all the leads and how that we can have what not in the industry, which has an effect is because we lack of expertise everywhere. If you bring machines plus the expertise to be run ⁓ at the factory, then you have a kind of full package there of servitization. ⁓ Same way machine to machine and when you want to build the line and so on. So we work with people who really develop their own application in order to work and sell additional environment. We've got a... that in recycling, we've got that in the poultry industry, we've got that in the food and beverage, we've got that in the packaging. There's really that wave of using that knowledge now that you can capitalize through AI, under data algorithm optimizers, or through the Gen.AI. This is really becoming a product itself and then delivered as a service. And that's where civilization is growing up. And I think this is a mainstream approach. It takes long in the industry because it has an impact on the economical environment, moving from capex to opex for factories and moving on that. But I am pretty sure that because of the lack of expertise and what I said, the silver tsunami and other people going to retirement, there will be the need to commerce on that, that civilization will grow up.

Michael Finocchiaro

I think this is the main thing, course, is that long-term, because it has an impact on the political environment. I ⁓ was curious too because we were talking about how ⁓ AI was being used in your development process. Are you guys developing your own ⁓ models like your own LLM or as Leo AI's founder, Mauer talked about an LMM. He created a large mechanical model, which I find fascinating. ⁓ Or are you guys using Mistral and your case long? And I'm not sure which one your favor of ⁓ Olivier and then just using rag in order to augment the for a question or something like that. Maybe Olivier, what do you?

Olivier MATHEY

Yeah, on our side, we have two solutions to product. Tegard, the automated 2D to 3D, we are developing our own AI. We are using the latest research paper on the computer vision on several families of technology. So neural network, graph network, some specialized language model. We know the LLM, Mistral and others.

Michael Finocchiaro

Great.

Olivier MATHEY

but there is a similar specialized one. And so we are developing this one. And so this is really our IP and this is what makes us efficient. This being said, customers don't care how it's done. They want the service. And we are facing the same pain points for all customers, the same pain point that Laurent mentioned. We have a big chunk of the engineer that's going to retire. And we have a partner in Japan that's telling us... The young kids coming out of university don't know how to read the 2D, so they got reconstructed. So it's not a question of cost, it's not a question of spin, it's a question of, I need to train a guy during two years before he can take over the engineering that's going to retire. And this is a huge, and this is mentioning it's coming in Europe, it's coming in the, North America, it's just a matter of time. So this is a huge challenge. ⁓ And then, yeah.

Michael Finocchiaro

Hmm.

Olivier MATHEY

We use a lot the external and I forgot to mention the marketing and sales. And I think it's a great one by the whole. So we have really, I would say, increase the impact of our sales team, which include myself a little bit. And I'm very happy, not spend a lot of my time with the lead generation because I'm the CEO of a couple of other stuff to do. But yeah, we are using it lot. So I would say everything that is process oriented. We use them to see how we could simplify, save time and develop new tools that will be in our solutions. And everything that is a core, which is when we develop more AI, that's when we develop more IP. So I hope it answered your question. So really two levels.

Michael Finocchiaro

and everything that is important. Yeah, that's sort of I was expecting, because I think there's also like the, I'll let Lauren answer first.

Laurent

I think it's...

Olivier MATHEY

So we spent four years of R &D developing our own model. So this is really a value in which we invested and we invested several millions, which for a company like us is a huge and then matter of efficiency. I think I agree with what Laurent said, where you can save time to do it. I think you have to be pragmatic, remain pragmatic as a leader.

Michael Finocchiaro

Mm.

Laurent

I differentiate again to AI. I differentiate the AI that we do for all what are the technologies on machine learning and all that, which is based on the data on the IoT side. And on those ones, ⁓ we've got our internal models, like prediction and vibration to predict for auditing machines. ⁓ If you've got a problem of alignment or a ⁓ misalignment and so on. So this is algorithm that we do on our side because we train them and then we are able to develop and deploy that to our customers. Same for ML. We've got some ⁓ proprietary model and then we also enable our customers to develop their own. So that's the value of our platform there. Regarding Gen.AI, of course, we're not rebuilding our LLM. Definitely not. I mean, we are not the size that but we

Michael Finocchiaro

That's more of an ML. That's more of an ML. Yeah.

Laurent

We really fine-tuned that to what is the field service approach there. And the problem of the industry with the LLM is that the vocabulary, ⁓ all the typology of industry is not really present in LLM. It's not really well organized. And specifically in a given value, which is the translation, if you've got documents on English and you want to have them in German, the words are not exactly industry related. So of course, what we do is we fine tune that. we have some glossary that we put as industrial glossary that we enrich so that really the system can speak the world of the customer there. And also we use another system, which I mean, we have a constraint in the industry is that if we suggest an action to an operator next to the machine, we should not be wrong. I mean, we cannot hallucinate.

Michael Finocchiaro

Okay.

Laurent

on that. We cannot give any answer coming from the web. So we are very restricting the content, of course, because we protect the IP of our customers. But also, the answer is strictly the one that exists in the corpus. If not, the system will say that it doesn't know, and it cannot help, and you need to call an expert. It may happen. mean, the system is not ⁓ getting 100 % of the solution over there.

Michael Finocchiaro

Exactly. You

Laurent

Regarding AI, I would say ML type, yes, we've got some algorithm and we deploy that. ⁓ Regarding Gen.AI, we do not develop our LLM. We fine tune that to the industry. put glossaries and we put the temperature at zero so that there's no hallucination in the corpus.

Michael Finocchiaro

⁓ And so when customers are using ⁓ SP3D in any use, I suppose that there are several touch points where they have AI. There might be a chat bot for the field service guy or maybe the designer to make sure that the design intent of the 2D is translated to 3D. Probably there's also some underpinning technology as well of AI, graph knowledge base or whatever. how is it, ⁓ what is the interface, where's the AI being used in your respective products?

Laurent

⁓ Just as a fun stuff, Michele, I want to introduce ⁓ there. What we did is we put AI at the beginning in the reformulation of cases, intervention cases and so on. When the guy is taking the report of his work and of course he's using a mobile phone now on the news and then he's dictating or writing or typing, but they're not very good in... ⁓ in the way they type and making a lot of mistakes and so on. So they can reformulate. It's a little bit of a very simple, accessible AI. But at the end, ⁓ what does it make? It makes the technician not afraid of writing something. And the most important thing right now is to capitalize what they know and is to incent them to really write something into the system because they have the knowledge in their head. So they have... They came for problem. They solved the problem. If you help them to put that into the system, and then our AI is used to take the content, not any work from the technician anymore, take the content and send that to the general AI for everyone to ask like a chat bot, ask a question. Anyone, anywhere can find, again, the solution. So that loop on that.

Olivier MATHEY

See.

Laurent

has been engineered to facilitate the life of the technician and the retrieval behind that. And I think it's quite interesting to see the impact of the reformulation, where people were shy to write because of that and not finding any reason for doing it. And now they find reason to do that because it's easy for them and they can get all the knowledge of their friends easily on their fingerpins.

Michael Finocchiaro

Cool, very cool. And how about you, Olivia?

Olivier MATHEY

We tried them to have the most simple interface as possible. the answers is different depending on the true product. On Xeia, they don't interact with AI. We have a very simple interface. You drag and drop. You grab a coffee when you come back. You have your 3D model. And then you have a chat box to share some, so you have discrepancy or the missing dimension, et cetera. But we try to simplify that with checklists, checkbox so that they don't have access to the complexity. So we do as much as possible simplifying the interface so that all the complexities is not a burden for them. On our first solution, which is really impacting the supply chain of major companies, we're talking about companies that have

Michael Finocchiaro

Hmm.

Olivier MATHEY

few hundred millions of dollars of parts on stock. In this case, it is having a human interaction to help them challenging clarifying what is their pinpoint, what is their difficulties. Probably one day we could simplify that by implementing what Laurent said, but we are at a moment where people need to have a few hours of meeting to express properly what is their pinpoint. And then we take that and we put that in the algorithm to basically check the database and extract. information. So we use it to transform the data into information to answer the questions. At the moment, we still use the human interaction because we are on very complex pain point and people are more comfortable to discuss. And we even travel to see them. We have been to Africa, been to Middle East, which is not expected from a software company. of our first customers in the line, we went there and we had this opening meeting and say, we did not believe you will come. Until you came yesterday in the...

Michael Finocchiaro

Olivier MATHEY

We are not believing you will do it. So thanks for doing what you said.

Laurent

And, Miguel, may I add something on that approach? I think one of the criteria also that we take, we need to take into account is the legacy. It's not because we've got a new technology that everything starts from zero right now. I mean, you mentioned 2022. This is not the birth of something. I mean, there was a huge industrial world before. So what we have done also is we have engineered the system that we can connect to every legacy system. Specifically, we are not asking our customers to reformat their documentation or their past ticket intervention and so on, which is a huge world. So it means that at different level of the chain, we use the AI because we take the AI, we OCR, we put character recognition, then we put the transformation, we do special chunking with some algorithm that we have invented.

Michael Finocchiaro

Right.

Laurent

I think one of the issues that we have, and I guess Olivier, that's the same for you, I mean, you're not asking the new 2D to be generated, you take the legacy. And we are working in a world where it's not a brand new world, I mean, a total white space. We have to take the brown thing. And we have engineered the system so that easily we can recover from what they have made in the past and get the benefit right now. At the end, if you take that, the return of investment on the solution is very fast because you take all that and after a few days, you have the system that has been adjusted to your own knowledge available for everyone. And that's where the benefit comes from.

Michael Finocchiaro

So it's about mitigating the risk of technical debt basically. you're making sure the Tegra doesn't become an albatross holding you back from progress basically. Very, very cool.

Laurent

Yeah. Exactly. mean, you have to take it into account. Everything that we do, you can't say that everything will start from scratch right now. I mean, you have to take the past. And we see on our customers documents of many kinds of different way of building them and so on. And it's a huge set of documentation in the industry that, by the way, was not used before. Just a mandatory, that's just an obligation.

Olivier MATHEY

If you may make a choice to rebounds on what Laurent just said, it depends on the complexity of the problem of the paint. When we do the 2D to 3D, the paint is very simple. I have a 2D drawing, I need a 3D model. And it's basically for us, like you work with Excel and your customers and you work. Because when you do the 2D to 3D, you bring nobody, you just transform the format so that you can work. In this case, the complexity is the building the answers, but the main point is simple. So in this case, people don't want to talk to somebody. want to do drag and drop. I do my stuff and when it's ready, I have it and I do it and it has to be simple. And what Laurent mentioned, which is quite similar to our first product, the pain point is complex because there is massive data and then it's how do I make the data speak? How do I make this data the legacy one, the collaborative information, the collaborative intelligence bring me something. And then there is an interaction and then this interaction being with

Michael Finocchiaro

Hm.

Olivier MATHEY

algorithm, interface, being with people. And in this case, the complexity is in the description of the, the, of the painful, but she said, using AI to clarify the problem, I think is super smart.

Michael Finocchiaro

Makes me wonder too, and this wasn't one of the planned questions, but I feel like it's an interesting one. Of course, you guys have a competition, maybe less SP3D because it is sort of a niche solution. Certainly in use, you're probably up against a service max of servagistics, right? PTC is pretty strong. ⁓ How do you guys, and yet those companies are slower to adopt AI than you. mean, PTC has made a bit of progress. Daso is pretty slow so far, you they haven't. And I suppose for you, Olivier, it'd be more like Fusion, the CAD products that are supposed to be able to do this to some degree. So how does AI become a competitive advantage for in use and SP3D against the incumbents, the billion dollar incumbents?

Laurent

I think this is the magic also of AI. It has not been restricted to the big ones that have big capital funds to access the AI. So what makes us competitive is that we access the same mistral ⁓ environment and the same technology is up to invent what we do with it. And ⁓ what I like here is that you don't need to have a this size, what you mentioned, ⁓ DS or PTC or whoever in that PLM space, is immense to have such kind of capital in human and in cash. OK, both also that. ⁓ So legitimately, we can really be competitive at the size of e-news in that field. And the only thing is we need, I think, to be clear on the value prop that we are delivering, which is really

Michael Finocchiaro

A seam hens.

Laurent

targeted a given part of the space, which is the field service and the proactive field service. Service max is part of that. I think we see more effect from service max or people coming from the service desk side ⁓ than the PLM side, which is for me a misunderstanding because I think you talk about PLM and all that life cycle. Lifecycle is not only virtual. goes up to the machine in use, in usage. That's the reason of the name, ⁓ machine in usage. And that's the prolongation as operated. And like Steve Klein said, servitization is the prolongation of the virtual product to the service over there. And I think this is life there. And that's where the link there. But definitely, AI is accessible for company like us. We are able to run AI. We are able to have a

Michael Finocchiaro

Of course. As operated, right? As operated.

Laurent

protect what we put in the AI for our customers, to protect the IP over there, as if we were a big company. And I think that agility is our best level of competitiveness.

Michael Finocchiaro

And how about for you, Olivier?

Olivier MATHEY

⁓ This conclusion is the way I will differentiate myself in a few years when competition will come. Today, we've been the first one doing the breakthrough. This type of topic has been in R &D since 72. ⁓ And these big names, Dassault, PTC, AutoCal, we know they've been working on it. It's probably not a priority for them because today they do a design for additive manufacturing, they do design optimization and they are competing themselves with how best they can improve their customer experience. And they have invested massively on that. 2D to 3D is little value. So it's complex to solve, but brings little value to the customer. At my level, it's huge because the market from my standpoint is huge. From their standpoint is a niche. So ⁓ we don't see them competing against us. We have been in talk. They know we exist now.

Michael Finocchiaro

Interesting.

Olivier MATHEY

because we have been, of course, communicating. have people using our solutions, so they started reading. So we see them just having a look and see. We also look at the big LLM because there is a lot of 2D to 3D from pictures. So we have a lot of time to question, but people can do it from a picture. Yeah, from a picture, you don't need to perpetuate. You go, you use Sora, you use Google, everybody does that. But you won't get a 3D drawing that you can't produce with. get it made. Exactly.

Michael Finocchiaro

Hahaha. The, a manufacturable one.

Olivier MATHEY

For the delivery, we can have to get a 3D model for quotation. In this case, we have this type of solution that could be a competitor. But for the 3D model for production, today the technological challenges have not been solved. It will be at some point. If tomorrow Daso decide to put 50 million of R &D and 50 guys, I'm sure they will catch up. Will they decide to do it? Not sure, because they also have other priorities. This is for us the space and I'm working and we are using DAI to amplify our impact so that we can grab as much market share as possible while they catch up.

Michael Finocchiaro

Of course. So in like one minute or less, because I want to get to the real world piece of the conversation. Over the time you've worked on the latest release, we talked about how you thought about AI when you started working and now we're in 2025 or almost the end of 2025, agentic, MCP and all the crazy stuff around sovereignty that's come up, right? because we've also had this with Anthropic saying, oh, actually we lied. We're gonna steal all your prompts even if we'd say we weren't. Where are you sitting on AI today? Are you still very bullish about it? Are you more reserved? Are you more tactical perhaps? mean, how has your position changed from 22, 23 when you started?

Laurent

Hmm.

Olivier MATHEY

If I answer, think for me, the maturity is why are you using LFR? What is the problem you want to solve? I think that's the answer, at least for us. And we have a clear answer, both on the internal use, which is basically make sure we spend time with CASR and simplify everything else from marketing, lead generation, and of course the process, the workflows we can develop for our tools. the deep tech that we are developing ourselves. We are not in the same competition as Google, Amazon, all the GAF and Chinese equivalent because we forget being in Asia, we see them a bit more where there is a possibility of the winner takes all. If you have the data, you are so this valuation is also the race on you're not sure who will won the race and you bet a bit on everyone. We are not in that type of game. I don't think the winner takes all is what industry is looking like.

Michael Finocchiaro

Hmm

Olivier MATHEY

As an issue, people like to have a bit of competition so you can challenge. So I think the game more efficiently can solve the problem of your initial custom. And I think as we focus on that, we'll be successful.

Michael Finocchiaro

So your opinion of AI has become more tactical, more or less, right? Pragmatic and tactical, if we resume.

Olivier MATHEY

The of whether you want to use it or not is done. You have to use it. The question is which AI for what purpose? I think that's the rule.

Michael Finocchiaro

Yeah. Lohan, do you agree?

Laurent

⁓ I think, I would say it's not yet the time for being tactical. Sorry Olivier to say that. think it's still, there's still an horizon for a huge transformation to come and we are not yet at that level. I think for me it's a very strong revolution that is going to ⁓ evolve fast. ⁓ Finding ⁓ new way of applicative approach, delivering new values there.

Michael Finocchiaro

Hahaha

Laurent

Agentic is definitely one when you can chain AI solution and you can do tremendous stuff. Just think that when we have the IoT data on one side and the generic on the other side, you just remove the dashboards. We are spending hours, millions of dollars on dashboard business intelligence that we know since long. Just ask the question to the system and it will generate for you the dashboard that you need.

Michael Finocchiaro

Mm. Exactly.

Laurent

I mean, it will create a system for you or not. That's just the next revolution coming. I'm not talking about 10 years. I mean, it's coming. It's coming 2026 at the news. will be there. OK, it's just you can. Yeah, because that's the next phase. And what do you transform there? You transform all the shift team when you transfer the shift from time to time. ⁓ You have a new way to approach what did happen in the last eight hours and the next eight hours and so on.

Michael Finocchiaro

you guys did...

Laurent

You can change the life there. So I don't think we have yet estimated that the impact is quite huge. We still need to be, I would say, ⁓ on that. We need to push on the AI. our job is not on the technology standpoint. It's on the applicative side. We need to find real value-added applicative solution that helps people. And if we do that, we will transform every domain of the world. and not only the industry, lawyers, doctors, and many, many professions are impacted there. So I'm very, very positive that there will be new way there.

Olivier MATHEY

Yeah, I agree with Laurent, the untapped potential. And I think there is, and the larger the corporation, the more difficult is the pace of adoption. Just us internally, we are doing things now that six months ago were not possible. I was chatting with a red-out operation this morning and say, but now you're doing this, what changes? They say, just open the page, I just ask it, it's doing my inner code, I go to the other soft. I ask you to do the QC and then go to the other stuff and then we connect. We were not able to do that. We are small, we are a giant. I think we have a difficulty.

Michael Finocchiaro

Thank you.

Laurent

And then you're right, Olivier and Mickaël, of course I'm very positive. There is also the risk and the difficulty over there, which is how people are going to transform the way they work with AI. Are they going to lazy? Are they going to trust AI first before their own knowledge? we still need to learn maths, physics, ⁓ philosophy and the rest and make sure that all the AI is not changing our mind on that. So there is a big transformation also in the way you criticize the AI and the result over there, which is something that we need to work on and educational environment and the fact that ⁓ I've got the result, is it the good result? I think this is the right way of... ⁓

Michael Finocchiaro

Thank

Laurent

mitigating the risk there. There's an education to make them.

Michael Finocchiaro

I think you'll like, I'm going to do an interview in the next couple of weeks with TD Engine, with Jeff Tao. And he has these dashboards, completely automatic. He calls it Snapchat for dashboards. I love it. It's so funny. So the last section, because we're coming up on the end, I wanted to ask, ⁓ when you go into your customers,

Laurent

Yeah, yeah, sure.

Michael Finocchiaro

⁓ I usually think of a digital maturity or around basically matured around data on a spectrum from one to five, right? And one is Excel and email. Five is a fully adaptive agentic digital twin. Like nobody is at five, barely anybody's at three, almost everybody's somewhere between one and three, right? ⁓ So when you go in and you put in in use and you put in SP3D, you must have a feeling for where the customer sits. You must have an idea of the level of their maturity and then you put in this awesome solution, which is AI powered and it's so different and so new. Is there a ripple effect? Does a customer say, my God, if I was more mature, I'd get more out of the solution. maybe, I mean, does it create sort of a ripple effect or even a tidal wave of change where the customer's like, my God, yeah, I've got a digital maturity. I've got to have a chief data officer and I've got to get my stuff together because otherwise I'm not going to be competitive tomorrow. So as my last question for you. Second to last question, I do have a final question.

Laurent

I think you're right on the maturity. One, two, three is the maturity. At least in the industry that I face, one, two, three is the maturity. You've got the two schemas. You've got the one that feel that, OK, it has been done like that, and they still have the knowledge, and they don't see in the near future the risk of losing that knowledge. And there's the one that are feeling the retirement wave, and they are in an urgent matter. Do they need to have a chief data officer and so on? I think if we properly present what we are doing, and that's what we do with E-news, I think we can overcome that. Because we can deliver the platform, making it very efficient right away, and evaluative in the way they're going to evolve. I mean, we don't ask them to be five right away just because we're going to give them an agentic solution there. OK, so the purpose is to bring them to the level. of usage ⁓ internally and then ⁓ with the ecosystem, suppliers and even their customers over there. So I think it's a question of evolution there. again, the good stuff is it's a big bang in terms of LLM. But if you want to deploy that, you can be very much adapted to your audience and your people. And it's really people-oriented. The Gen.AI is really people-oriented for me.

Michael Finocchiaro

Olivier? Olivier. course. Okay. Mm-hmm. Ha. .

Laurent

Hmm.

Michael Finocchiaro

you ⁓ Mm. So my final question before we say goodbye ⁓ precisely we've got this anxiety now we've got a real AI anxiety in the workforce because we're seeing all these people being laid off. ⁓ So how do you guys see that happening? Do you guys see that AI is actually an opportunity because we can give the mundane boring stuff to the AI and let engineers actually do engineering rather than plumbing? ⁓ Or is there a real fear that we we are kind of getting towards the age of post-scarcity, right? Because we will get to the point of a HDI and everything and it'll be the post-scarcity universe of Ian Banks. ⁓ How do you guys feel that we're gonna see and from a personal point of view for the engineers, is it a good thing or is it sort of a scary thing or maybe it's just a bump in the road? because, know, in the dust revolution, we still didn't have engineers even though we had machines, right? We just had less horses as all. Ha

Laurent

You

Michael Finocchiaro

think that growth is getting a bit scary for people that are looking at that industry. Looking from behind, what companies that are going to be the next people be year. It's not really a century. It's an up-and-coming The thing is that people are going to be in control of their nationals, so it's called globalization and all these changes are going deliver a great conference.

Laurent

⁓ I think there's a social challenge on that and there will be the one that ⁓ will take benefits because they will understand and see how they reorganize fast to use the AI and the one that will be slow or not accessible to that. ⁓

Michael Finocchiaro

Thank you. So, will the one that will take medicine, so that they will understand and see how they are really doing.

Laurent

And if you look at the size of the planet, you will have ⁓ some that are going to accelerate and some that are going to be slow. And that will create distance between those populations. I think this is the highest risk of the AI. It's not an individual risk at the level of one employee or something like that. It's the level of the people who are going to access the AI and the company that will give access to grow up and the one that will be slow on that. And that's the reason why I'm also pushing Europe with this trial and so on, because we need to have the technology. And you talk about sovereign approach over there. I think the one who will get that and the one who will understand how to organize, yes, there will be some population, not only in the industry, in the engineer field, but in the, as I said, in the doctor, there will be some help. The one that will benefit from that, cool. They will probably get a huge advance and the one that will miss that and there will be some some region of the world or some population that will not have yet and that will create a highest disruption and that will be the highest risk for the social part of the world and I think this is the main approach that AI is so much accelerating. We talk about speed, we talk about all that it brings. Yes, to learn how to use it is the biggest way.

Michael Finocchiaro

That should be the highest risk for the And I think this is the main approach that AI is to focus on the theory of things, to talk about speed, to talk about all that is great. Yes. you

Laurent

then they will be there. And then it's another way to fight because then the people who don't have access or fully access, they will probably use the AI to slow down by putting some wrong thing inside. And you know, there will be a new environment. So this is the highest risk there for social point of view. And it's not a question of laying down people and firing people on that. It's temporary. mean, they will find Michael Finocchiaro (1:00:01) you Laurent (1:00:08) that they can become ⁓ technicians. There's a lot of job to do there. But I'm not ⁓ thinking this is the highest risk. The highest risk is it's going to reconfigure the social approach of the world, and that will create new tension. Michael Finocchiaro (1:00:27) Thank you for that. That's ⁓ bit disquieting. It's true. There's some uncertainty. Laurent (1:00:30) Sorry to finish on that. if you ask for the risk, it's so much a transformation. It's so much high transformation that it has an impact on that. And that's why we need to fight every country, every need to fight for AI. And I do believe that they need to fight for AI. And look at the... Michael Finocchiaro (1:00:43) It's such a short period of time, Laurent (1:00:56) the last round of Mistral and so on. It's important to fight for that. Yeah. Yeah. Michael Finocchiaro (1:01:00) ASMR saving the day. How about Singapore? I know we're almost out of time, but Singapore, do you see them really fighting for it? Because I mean, they're essentially not in either camp, West or East, they're sort of in their own camp. So what are they doing in Singapore to keep up? Of course. Well, and all the networking cables that are pulled through Singapore to get from one side of the world to the other. Okay, very good. I'm not sure we're at that point in France yet. ⁓ Well, that's awesome. I had a really great time talking to you guys. That was super interesting. I think I learned a lot and your company sounds so exciting. It's awesome to see you guys succeeding and to get to know you better. Thank you very much for taking your time to tell us about AI and your two companies. I hope you guys had a good time too. Laurent (1:02:28) Hm. Yeah. Thanks. Thanks a lot. Yes. Michael Finocchiaro (1:02:54) Okay. Well, see you on the... Laurent (1:03:01) Yeah, thanks. Talk to you soon. Michael Finocchiaro (1:03:03) Okay, well, we'll talk to you soon. Thank you very much, everybody, and we'll see you on the next podcast. I think the next one's actually Thursday, where we're gonna talk about whether PLM should be dead or not. good Halloween topic, right? Thank you, bye-bye. Laurent (1:03:08) Yep. Okay, absolutely good. Thanks a lot.