Pod Insights Eye On A.I. #278 Julia Peyre: How Schneider Electric is Pioneering Enterprise AI at Scale About this episode AGNTCY - Unlock agents at scale with an open Internet of Agents. Visit https://agntcy.org/ and add your support.
How does a 150,000-employee global leader make AI work at scale?
In this episode of Eye on AI, host Craig Smith sits down with Julia Peyre, Head of AI Strategy & Innovation at Schneider Electric, to explore how the company is pioneering enterprise AI adoption through its AI Hub, hybrid AI systems, and real-world digital twin applications.
From breaking data silos and embedding AI into hardware, to partnering with startups and building predictive maintenance solutions, Julia shares a blueprint for bringing AI from pilot programs to full-scale deployment, across both internal processes and customer-facing products.
Stay Updated:
Craig Smith on X:https://x.com/craigssEye on A. I. on X: https://x.com/EyeOn_AI(00:00) Preview and Intro
(01:58) Meet Julia Peyre & Her Role
(03:19) Inside Schneider's AI Hub
(05:51) AI in Industrial Automation & Robotics
(08:43) Internal vs External AI Applications
(13:54) Why Schneider Is Ahead in AI Adoption
(15:57) Centralized AI Hub Model
(19:38) Hybrid AI: Combining Physics & Data
(25:44) Early Steps in Multi-Agent Systems
(29:54) Breaking Data Silos for AI at Scale
(32:02) Predictive Maintenance with Hybrid AI
(37:03) Long-Term View on AI Automation
(42:37) Advice for Young Professionals in AI
(44:21) Framework for Evaluating AI Solutions
(50:19) Involving End Users in AI Testing
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:00 Speaker 00 We developed products, connected products, software and services for everything related to energy management and industrial automation. And so it's always a bit hard to explain what Schneider Electric is doing because there is a lot of end markets, so lots of different buildings, residential, commercial, critical building like hospital, industry as well, where we have all of this automated solution, as well as infrastructure or the transportation of energy, for example. We are a manufacturer, so we have a global supply chain that we are trying to bring to higher standards.
0:40 Speaker 00 So we also use our products in our global supply chain, but also we sell to all our industrial automation clients.
0:47 Craig S. Smith? Build the future of multi-agent software with AGNTCY. The AGNTCY is an open source collective building the Internet of Agents. It's a collaborative layer where agents can discover, connect and work across frameworks. For developers, this means standardized agent discovery tools, seamless protocols for inter-agent communication and modular components to compose and scale multi-agent workflows.
1:25 Craig S. Smith? Join Crew AI, Langchain, Llama Index, Browserbase, Cisco and dozens more. The AGNTCY is dropping code, specs and services, no strings attached. Build with other engineers who care about high-quality multi-agent software. Visit AGNTCY.org and add your support. That's A-G-N-T-C-Y dot O-R-G.
1:57 Speaker 00 Thank you, Craig. Thank you for inviting me to your podcast. I'm very glad to be here. So I'm Julia Per. I'm leading the AI Strategy Innovation Team at Schneider Electric in the AI Hub, which is our center of excellence in AI. So a bit of background about myself, I have a technical background in AI. So I started my professional career in AI around ten years ago. As a PhD in computer vision at INRIE in France, so INRIE is a leading French institute. And I've joined Schneider Electric around three years ago. Okay, a short after, shortly after the creation of the AI.
2:42 Speaker 00 And so since then, I have participated to the development of different AI solution. In particular, taking, you know, a speed at the, at the, at the release of ChatGPT. To understand what this type of technology could mean for us at Schneider Electric. And also more recently, evaluating innovative startups and partnering with them.
3:11 Craig S. Smith? Yeah, I'm sorry, just tell us about Schneider Electric.
3:17 Speaker 00 Yes, I'm coming to this. Schneider Electric, it's a leader in efficiency and sustainability. And so basically, we developed products, connected products, software and services for everything related to energy management and industrial automation. And so it's always a bit hard to explain what Schneider Electric is doing because there is a lot of end markets, so lots of different building, residential, commercial, critical building like hospital, industry as well, where we have all of this automated
3:59 Speaker 00 solution, as well as infrastructure or the transportation of energy, for example. And so Schneider Electric, three years, a bit more than three years ago now, created an AI, so a center of excellence in AI to accelerate on the adoption of AI solution for our external offer, all of our clients, but also for internal efficiency. So for our day to day, for the day to day activities of our employees.
4:34 Craig S. Smith? And the products are software, is it a SaaS service? Or are there hardware products?
4:44 Speaker 00 So we have everything, in fact, so we manufacture products. It can be, maybe if you see in your electrical panel in your, at your home, you will see, for example, a circuit breaker that we manufacture, okay? But it goes up to software. So in fact, what we are selling is, we have an IoT platform called Ecostructure. And so that connects products to edge controllers, up to softwares that help our clients control all the different electrical appliances and also bring services to our clients.
5:25 Craig S. Smith? Yeah, because as I recall with Covarian, I think they were working in with, I'm gonna forget, I'll cut this out, but with a ABB, I think. Is that right? On robotic solutions in your warehouses, am I wrong on that?
5:51 Speaker 00 So we have, you're right, it's industrial automation part. Okay, so we are in energy management domain, but also in industrial automation domain. And we are also building, for example, robotic arms that can pick products and place them in different locations. We have also some conveyor belts, for example, and software to pilot these belts.
6:16 Craig S. Smith? Are those products that you're selling? Or those are in your warehouses?
6:22 Speaker 00 It's both, in fact, it's a good question. We have, we are manufacturer, so we have a global supply chain that we are trying to bring to higher standards. So we also use our products in our global supply chain, but also we sell to all our industrial automation clients.
6:41 Craig S. Smith? Yeah, I mean, that, what I find fascinating is that you're really applying AI end to end, right? Not only in business processes, but in warehouse management and then.
6:55 Craig S. Smith? Selling those services and so yeah, maybe take us through um how you well
7:05 Craig S. Smith? what are you working on right now? Let's start with that.
7:08 Speaker 00 Okay, so it's it's a good question. So I I I'm leading the AI strategy innovation team in the AI. So the AI is the center of excellence in AI and my team is kind of the R and D of the AI, if you will. So what I'm trying to do is anticipate um anticipate the needs for tomorrow. Okay, and tomorrow is maybe 12 months, 18 months, 24 months, maybe, but not too far neither. And so I'm trying to understand what are the key strategic topic we need to to to solve problems for our business. And I'm trying to anticipate what are um the new solution, innovative solution that we could use.
7:53 Speaker 00 And so for this, I need to identify right partners and also test this solution concretely on our problems. Okay.
8:05 Craig S. Smith? Yeah, uh, and um so which solutions are you looking at right now? Or maybe I'm sorry, I'm I'm I'm not doing this very logically, but I'll edit it. Maybe take us first of all within Schneider Electric.
8:27 Craig S. Smith? The range of AI applications that you guys are using. And then how what what are the AI applications that you're selling? And then we'll talk about what you see on the horizon.
8:44 Speaker 00 Okay, so um, okay.
8:49 Speaker 00 So if we take a setback, so for if we take the application, we have, you see, we have two kind of application, internal efficiency and then external for external offers that we said. Internal efficiency, maybe eighty percent of our use case today. So it's it's to help day to day activities and with generative AI today. What we empower is everything related to knowledge management, information retrieval. Okay, so that's the type of topics that we are working on. On external offer for our clients, there is a wide diversity of tasks that we are trying to solve.
9:31 Speaker 00 One of the example I can take to make it more concrete. We have an offer called a microgrid, ecosystem, microgrid advisor. So it's basically for residential building to help homeowner who have a modern home with maybe an EV charger, a battery, a solar panel, some heating system as well, to control this automatically to answer the demand. As a demand which might be in twenty four hours, the car need to be charged and the temperature in the home need to be set at nineteen degrees when the homeowner come back. Okay, and so do respect the demand, but also the constraints.
10:16 Speaker 00 So minimal costs of energy and trying to use as much green energy as possible from the renewables. Okay, so this is a problem that mix in fact forecasting of the demand and supply of energy as well as optimization of this complex problem under constraints. So that's the type of AI application that we are selling today to our customer. We have a first version out, but we are still working, for example, on making the model more robust, improving these forecasts, modeling the uncertainty.
10:58 Speaker 00 And so for this admin, for example, we are looking at how generative AI for time series in particular can help us to model all of these different scenarios.
11:12 Craig S. Smith? And then internally, when you were saying.
11:18 Craig S. Smith? Yeah, can you walk us through some of the processes that you apply AI to internally? So from, from maybe filing expense reports or managing corporate finances to
11:39 Craig S. Smith? controlling robotic arms, picking products and placing them in bins in the warehouse.
11:46 Speaker 00 Okay, yes, so here maybe I can give an example. Which, so the example I can give here is for customer care center, in fact, okay? So we need to satisfy our customer and to satisfy the customer, we need to understand exactly what are the problem that, that, that he met. So one, one thing that we are doing is leveraging our historical base of FAQ, question, answer of problems that occurred in the past, as well as all our documentation, but product manuals,
12:26 Speaker 00 documentation that can be maybe a bit less official as well. So leveraging all these diverse set of documents to help customer care experts to answer more efficiently and so with a better, better accuracy as the clients.
12:54 Craig S. Smith? And, and then on, I'm interested in the robotic side. Can you talk about how you're using AI in robotics or am I mistaken on that?
13:07 Speaker 00 And I think we don't have, I'm not sure, Craig, on this, I'm not sure what I can talk about, actually.
13:16 Craig S. Smith? Okay, okay, that's fine. So, I mean, just what interests me is, you know, I talk to people a lot about enterprise adoption of generative AI. And while there are a lot of pilot programs, it's been extremely slow, it's been extremely slow in, in.
13:37 Craig S. Smith? In other flavors of AI supervised learning even.
13:43 Craig S. Smith? And Schneider just seems way out ahead or one of the companies that's way out ahead on adoption. Why is that?
13:54 Speaker 00 It's a good question, Craig. Thanks for asking this. I think maybe there's two components. I would say there is organizational components. So we have made the choice three years ago to have a central AI hub to support business and function. And so we have put in place this AI hub and spoke model where we help business units with our expertise, we help business units to structure their problem, formulate it, make sure that all the constraints, the environment is well described. Okay, and we go up to through different stage from exploration to industrialization,
14:38 Speaker 00 leveraging the 350 experts we have internally, but also leveraging the different partnerships that we develop with big companies like Microsoft, AWS, Databricks as well, and also smaller companies, startup, innovative startup as well. So that's the organizational part. I think that works quite well today. And the second part is also that we invest a lot of effort in AI awareness education for everyone at the company.
15:20 Speaker 00 At different level, of course, awareness for leadership, awareness for users, people we will use the solution internally because it's key for the adoption and also keeping our AI experts at the forefront of the technology.
15:41 Craig S. Smith? And so the lab works with all the various business units. Do you have sort of representatives from the lab embedded in each business unit?
15:55 Speaker 00 So what you call the lab is AI, right? So the AI, yes, so it's still, so you know, we we're still a central organization. And we have a reference in each of the business units. That talk to one person in the AI hub, okay? So currently, we don't have a mini lab in each business unit. It's still very centralized.
16:25 Craig S. Smith? Yeah, I'm sorry, I said lab, I meant hub. Right. And then is this coming from?
16:36 Craig S. Smith? Obviously, it's coming from senior leadership, but do you have a CEO who's particularly
16:44 Craig S. Smith? focused on AI adoption? I mean, it is, you know, an electronics company. So I presume that people are paying attention what's happening in technology. But I just find it unusual, the speed at which Schneider has adopted this stuff.
17:04 Speaker 00 So you you know, the AI lab before even becoming a hub, it existed before. We always had an R and D department looking at AI. And so senior leadership indeed decided to scale it to 350 people today.
17:31 Speaker 00 Sorry, Craig.
17:34 Speaker 00 So can I redo it?
17:36 Craig S. Smith? Yeah, sure.
17:39 Speaker 00 Okay, so but basically my my answer is that um this comes from senior leadership, this comes also from signals that now AI is a key um solution if we want to bring more efficiency and more sustainability in our products and for a company of one hundred and fifty one thousand fifty, so sorry, I I I I'm confused, for a company of one fifty hundred thousand people, we had at least, we needed at least a few hundreds people to support
18:20 Speaker 00 correctly our business unit and function to scale. So this really comes from the fact that we cannot play and be a leader in our domain without AI.
18:33 Craig S. Smith? Yeah, yeah, uh the um uh
18:40 Craig S. Smith? I mean, I you know, I'm not sure that you would be aware, but are there other companies who are as deep into AI application internally as you guys are that you're you're familiar with?
18:57 Speaker 00 I think in our domain, we are quite well positioned.
19:06 Craig S. Smith? Yeah.
19:07 Speaker 00 Yeah.
19:09 Craig S. Smith? And so let's talk about where you what you you you see on this twelve to eighteen month horizon. I mean, obviously, multi-agent systems are going to be a major part of it. What what kinds of things are you working on, both, well, primarily for Schneider internally and then for products?
19:37 Speaker 00 Hmm, so it's it's a good question. I like it. So one thing that is very important for for us at Schneider Tric, so we manufacture physical products, okay? And so these physical products, they live in in the world and so they are
19:57 Speaker 00 They are submitted to the law of physics, if you will, okay?
20:01 Craig S. Smith? Yeah.
20:02 Speaker 00 So it's really how, it's not just AI data driven, but it's also how we embed this physical law in our system. Okay, so that's what we call a hybrid AI. And we are not the only one to to look at this. So lots of people, like a lot of company like Nvidia, for example, they work on building this simulator that can create a more realistic representation of the world in which we can train AI model. Okay, so that's one first thing of hybrid AI that we think is important.
20:46 Speaker 00 Everything also for us related to expert knowledge, how to embed beyond embedded knowledge from documentation, already formalized in documentation in our system, how to include the domain knowledge in our system. And so, of course, we talk about a large language model fine tuning.
21:13 Speaker 00 So that's one thing, of course, that we are exploring. But we're also looking at how we can have systems that are more interactive with the user to to kind of embed their knowledge that is maybe not so well formulated in our system.
21:36 Craig S. Smith? On the first, the simulation, are you talking about building a digital twin of Schneider Electric and then using that to explore different scenarios for decision making?
21:51 Speaker 00 Yeah, I think that's the idea. And again, we are not the only one to do this. That's extremely important to to build this kind of simulated environment or digital twin, because.
22:07 Speaker 00 That data acquisition, for for example, is extremely costly, okay? So if you can build with the knowledge you have on a problem, physical equation, you know, on how how how the system work. If you can build this digital twin and using to simulate different scenario.
22:30 Speaker 00 To to to then, you know, maybe train control system or this robotic arms that we were talking about. And that can also adapt to to new environment, new scenarios, that's extremely valuable to us, of course.
22:44 Craig S. Smith? Yeah, this digital twin idea is something that's kind of fascinated me and I've never really understood. How you do that? I mean, is it in a single system? So you have one model that
23:03 Craig S. Smith? that that simulates the entire organization. So you can
23:11 Craig S. Smith? ask it about, you know, your supply chain and it would. If you have a, if you're considering a new vendor, then it could look at the supply chain and and, you know, explore all of the implications of switching vendors. So is it an overall digital twin or do you build simulations of specific business units or or transactions?
23:43 Speaker 00 It's a good question. I I think it's like for generative AI, it starts maybe by a single application for for a business unit or even an entity in a business unit. And then trying to make maybe this digital twin communicate with each other to unlock even more value because, for example, you might, you in certain application, you might want to have information about the design of a product to to better maintain it, maybe to to know also how to operate it. So from design, operation, maintenance,
24:28 Speaker 00 you need to have all of this database communicating, so hopefully through a digital twin.
24:36 Speaker 00 I think it's a complex problem, so just like what we see in generative AI, probably it will start with more side load application, but hopefully breaking the side loads in the future.
24:52 Craig S. Smith? Yeah, and and to build that simulation, is that purely with generative AI or using other technologies?
25:02 Speaker 00 No, it's with different technologies. So I cited Nvidia, for for example, you might you might be familiar with Nvidia Omniverse Toolkit that allow to be this kind of simulator. And generative AI is one component of it. For example, if you want to generate from this environment different videos, for example, or images to train your AI model.
25:34 Craig S. Smith? Hmm, the I mentioned multi-agent systems. Are you working on that?
25:43 Speaker 00 We we are we are starting to to look at this, of course. And it's kind of the same thing. So we we we start by making application that that work for a well-defined and more siloed application and starting to to to see how we can make all of this communicate with multi-agent system.
26:09 Craig S. Smith? Hmm, and for multi-agent systems, are you guys, do you build that? Would you build that yourself using the tools available? Or would you adopt software as a service? I'm thinking, I mean, it's not widely available yet, but everyone talks about Manus, you know, the Chinese multi-agent system. Uh, yeah, how do you approach that? Is that something that you guys do internally? Or do you look to an external? Vendor that has a product that you can adopt.
26:49 Speaker 00 Uh huh. It's it's a good question. It's it's a general question. What's what is our make versus buy versus partner approach? In fact, so Schneider, you you understood we we we are um we have uh this team of AI experts. But we we are not uh building uh generic AI products. So uh products uh that AI products of the chef AI products that we can adopt and use directly without customization. We don't hesitate. We buy these products. Okay.
27:26 Speaker 00 Uh we can also partner uh with um uh provider to build uh to to to to. Sorry, we we can we can partner with um provider uh to to use their solution and customize on our needs on top of this. So for um so for Agentic AI, we are at the moment looking at the ecosystem uh with a different partner. And we are uh at the same time uh looking at all our requirements, all the different type of agents that we need to build. How we will need to also integrate this uh with our current software, sometimes also hardware.
28:11 Speaker 00 Uh and and we will make the decision with all these different criteria um of of of of business, scalability, technology that we answer our need.
28:23 Craig S. Smith? Yeah, uh and again, looking out twelve to eighteen months, and I mean, I'm sort of obsessed with agentic workflows. But uh you mentioned simulations and so so uh what is the?
28:38 Craig S. Smith? What is the thing that you think is going to have the biggest impact on Schneider Electric but enterprises generally in the next 12 to 18 months? Build the future of multi-agent software with AGNTCY. The AGNTCY is an open source collective building the Internet of Agents. It's a collaborative layer where agents can discover, connect, and work across frameworks. For developers, this means standardized agent discovery tools, seamless protocols for inter-agent communication,
29:21 Craig S. Smith? and modular components to compose and scale multi-agent workflows. Join Crew AI, LangChain, Llama Index, BrowserBase, Cisco, and dozens more. The AGNTCY is dropping code, specs, and services, no strings attached. Build with other engineers who care about high-quality, multi-agent software. Visit AGNTCY.org and add your support. That's A-G-N-T-C-Y dot O R G.
30:01 Speaker 00 So for me, the biggest impact now, I think it's for internal applications, so internal process efficiency, I think it will come if we manage to break these data silos.
30:19 Speaker 00 And this is not just a question of AI model. This is also, do we have the right data connector for this? And can we also standardize this data so to understand what relates to what? So this is where I think is the biggest value coming, also in everything related to knowledge management. I think we have the solution now available, but we still have to adopt them to take benefit of this solution of knowledge management. And then for external products for Schneider Electric.
31:06 Speaker 00 A big part will still, I believe, not necessarily relate on generative AI, because in the end, we need a solution that we can also embed in our product, in specific hardware. And so for me, the biggest value will come more if we can integrate, as we say, this more physical law, this expert knowledge in this system, but also respecting the hardware constraints, the latency constraints, and being able to, for sure, to maintain this system and integrate them well in the current software stack.
31:49 Craig S. Smith? And on that end of it, on the external side of it, can you give us a use case that you're looking at?
32:00 Speaker 00 We are looking...
32:05 Speaker 00 So I mentioned hybrid AI, so how to integrate a physical equation in the system. So all of this, you can think of it in the context of predictive analytics. So we are developing a solution for predictive maintenance for our clients to understand when there is a failure, a risk of failure in our equipment, but also when, how to optimize the maintenance of this equipment. And so we need to understand what is, for example, the remaining useful life of our electric appliance.
32:48 Speaker 00 So this is an area where there's a large body of literature explaining how to measure remaining useful life for these different types of equipment. And I think the challenge is to see how we can integrate all of this work or of this knowledge into more modern AI system.
33:13 Craig S. Smith? Yeah, when you said hybrid, you're talking about traditional control theory. You know, hard coded equations marrying that with with generative AI.
33:28 Speaker 00 Yes.
33:28 Craig S. Smith? What you mean?
33:29 Speaker 00 Yeah, I'm really talking about, so you have data driven approach that can benefit from all the lab data, operational data that you are gathering. And on the other hand, you have all expert rule physical equation. And so how can you create a hybrid system to to take benefit of both?
33:51 Craig S. Smith? Yeah, it is.
34:00 Craig S. Smith? Don't worry, I added.
34:03 Craig S. Smith? This idea of breaking data silos and simulating specific applications and then eventually hopefully simulating the entire enterprise. Do you see a day when Schneider Electric or enterprises in general are going to be largely managed by AI and the senior management or the the technical
34:42 Craig S. Smith? workforce
34:45 Craig S. Smith? will be
34:49 Craig S. Smith? tweaking and and improving and watching the AI operate. Do you understand what I'm saying?
34:58 Speaker 00 Yeah, I understand. I currently I'm more thinking AI is augmenting our workforce. So AI will probably automate things that are very much standardized. But what I see everywhere is lots of exceptions, also environments that are changing. You see also Schneider Electric, there's a wide diversity of clients, of applications. So I don't see this as miraculously, we'll have a multi agent system that will handle everything, but more really.
35:39 Speaker 00 Helping empowering our employee in their day to day job to automate completely things that you know that are no added value
35:54 Speaker 00 that are very standard and the rest of it to have the expert reconcentrate and the reconcentrate on spending more time with the clients or preparing also the future.
36:10 Craig S. Smith? Yeah, and how far out? I mean, obviously, this is an evolving process, both in the adoption and in the in the technology that you can apply. How far out do you do you see
36:30 Craig S. Smith? an enterprise that's that really has taken care of all of the standardized work or
36:40 Craig S. Smith? non value added work with AI?
36:45 Speaker 00 When will it occur? That's your question?
36:48 Craig S. Smith? Yeah, I mean, is that something that you imagine? And I mean, on a widespread basis, both at Schneider and globally, do you see that something happening in a year, in five years, 10 years?
37:05 Speaker 00 For me, it's it's a never ended process. We are already automating some of the things. The technology keep on changing very fast. So this will change, you know, hand in hands. And in 10 years, we will still have have have more to to augment, you know, and our job is just that our job, of course, will be different in 10 years from now. But we will, we will. Continue to to change.
37:35 Craig S. Smith? Yeah, um, in keeping up with that change um. And and exploring uh new technologies and. Uh, you know, surveying the literature, how much AI do you use personally in doing that? Uh, you know, you now have these fantastic tools like. OpenAI's deep research or DeepSeek's R1 model, or, you know, every, every major. Uh, model provider has a bunch of tools. Uh, to do that, I mean, do you personally use those tools?
38:13 Speaker 00 Yeah, I for personally, I I use a lot uh generative AI. For for information retriever uh for sure, though, um, you know, for keeping up, keeping up to date with new tools. I I still rely a lot personally on on LinkedIn, for example, where where um I have a subscription uh to. Uh, to thought leaders, okay, that's a tie for all, um, then, um.
38:45 Speaker 00 Then, then, then one thing, uh, I think we.
38:50 Speaker 00 Um.
38:54 Speaker 00 No, no, so so sorry, uh, I'm done, Craig, on this, yeah.
38:58 Craig S. Smith? Okay, I mean, the reason I ask.
39:01 Speaker 00 You can cut it before I had another idea, but.
39:03 Craig S. Smith? Sure, sure, and and you'll see an editor before it goes out, uh, but.
39:08 Speaker 00 Yeah, I hope we can edit your the answer because I'm not convinced with a lot of answer I provided.
39:14 Craig S. Smith? Yeah, that's fine, that's fine, uh. The reason I ask is I I try to keep keep up with with uh some of the literature at least. And what I do is I'll. Take a paper, copy and paste it into uh. ChatGPT or Perplexity or or one of the. Models I subscribe to and and ask them to give me a summary first. And then uh, when I read through, if there's something I don't understand, I ask it to explain it to a layman. Uh, because I'm not a practitioner and and it's, you know, a fantastic educational tool, but.
40:00 Craig S. Smith? But just reading papers directly to me is onerous. You know, I can read the abstract, but to read through all of it because I'm not a mathematician.
40:15 Speaker 00 Yes.
40:16 Craig S. Smith? Yeah, so it's a fantastic tool that way. And I just wonder how many people in the field do that? Because there's just so much stuff to read.
40:30 Speaker 00 Yeah, you're right. So I agree it's extremely useful when you want to learn and follow a new AI technology on a domain that you're not familiar with because you don't really know where to start. You don't know the key labs that are working on the topic. So here, here, yes, of course, we rely on this.
40:57 Speaker 00 Yeah.
40:58 Craig S. Smith? Yeah. And are you using personally? Are you using agents yet at all? Because agents, I'm not yet, but they would be fantastic for just paying attention to what's getting attention in the in the in the field and then packaging it and delivering it to me so that I don't have to be constantly scanning headlines and that sort of thing. Do you use anything like that?
41:35 Speaker 00 Not yet personally.
41:37 Craig S. Smith? Yeah, I'm looking forward to that happening. Another thing I wanted to ask, and I realize it's a little off topic.
41:47 Craig S. Smith? If you were a non-practitioner, if you had not gone to graduate school,
41:58 Craig S. Smith? a lot of young people are very confused right now about where they should spend their energy. Should they learn to code? Should they just focus on learning, you know, the field of AI tools and becoming experts in tools? Or should they? Uh have a more general uh understanding of AI and and where it's going. I mean, things are changing so fast. Uh it's it's difficult for young people to know where to step into it.
42:37 Speaker 00 Yes, so here, what I truly think, um so you need to learn how to learn.
42:47 Craig S. Smith? Yeah.
42:48 Speaker 00 That's for for sure. Uh and and the second thing, you it's important to learn also a scientific methodology. So I would say, no matter whether you are doing mathematics or physics, or well, pick pick whatever you you like, right? Uh but learn rigorous scientific methodology. I see this because when I I evaluate tools, uh so AI solution for partner. I really rely on rigorous methodology to do this evaluation. Uh uh proof of concepts that are conclusive.
43:28 Speaker 00 And what's everything that I learned in my study, in my PhD, this this is helping me a lot, and I see the difference. Compared to people who didn't have the education. I think then is it AI or not AI, uh there are so many resources today, so many tools. If you have learned how to learn, you you will you will ramp up with this.
43:49 Craig S. Smith? Yeah, that's interesting. Uh Terry Sonowski, I don't know if you know him, but. Uh he's an early uh AI pioneer with Jeff Hinton. Uh has a course on Coursera actually called Learning How to Learn.
44:07 Speaker 00 Okay.
44:08 Craig S. Smith? That's specifically about that. Uh, okay, you know, I've been asking these scatter shot questions, uh.
44:17 Craig S. Smith? But what is it that you really wanna talk about? Is there?
44:22 Speaker 00 Oh.
44:23 Craig S. Smith? Yeah.
44:23 Speaker 00 So, so you know Craig, it's, I'm very happy that you you ask all of these question. I I did not expect it as much as much question indeed on the innovative side. I wasn't so much prepared on this.
44:37 Craig S. Smith? That's okay, yeah.
44:38 Speaker 00 We'll have to see if my team is okay with the information disclosed.
44:43 Craig S. Smith? Yeah.
44:44 Speaker 00 But but so as if we still have a few minutes. I I can talk about, you know, partnership evaluation. This methodology that that has been developed. If it's interesting to your audience, we can, we can take the rest of the call to to go through this.
45:03 Craig S. Smith? Yeah, absolutely.
45:05 Speaker 00 So, so this is, this this topic of AI solution evaluation is something I I've worked on a lot in in the last year. And so we've, we've developed a methodology to choose, to select and and choose AI solution. And so we we've developed a methodology based on three criteria. The company, the technology and business and scalability. And this is, this are criteria that I would encourage your, your audience to to follow. If if they want to evaluate an AI solution. So if you're interested, I can go through these different criteria.
45:49 Craig S. Smith? Yeah, absolutely.
45:53 Speaker 00 Okay, so.
45:56 Speaker 00 On the techno, on the, on the, sorry, so on the company criteria, what we are doing is a kind of a background check of the company. That's to to understand if they have the expertise in in the domain. Where they are coming from, where they are in terms of the founding round, for example. And so we do this a lot with our Schneider Electric Venture team. Okay, who are used to run this background check. And the second element of this is more on understanding whether we have mutual interests.
46:36 Speaker 00 We have a common interest, a mission that is aligned and values that are aligned. And we realize this during our first interaction, okay? So company criteria, background check plus is mission and value aligned. Second, it's the technology part, evaluating whether the technology solves the problem. And this is what I mentioned to you just before on the scientific methodology. What I would like your audience to keep as a message is that's not try to make a P O C successful, but making a P O C conclusive.
47:20 Speaker 00 So with the different metrics that you want to evaluate on real use case, domain data,
47:29 Speaker 00 checking rationally if this align. And of course, we all want that things with a partner work and we go to the next stage. But I would encourage really to first try to reach conclusion, rigorous conclusion on whether things are working consistently or not on several use case of our business. And the last, yes.
48:00 Craig S. Smith? Can I, can I just stop you on those two steps? Do you use a rubric? Do you use some sort of tool to, or is this just in your head and in conversations?
48:14 Speaker 00 We have our checklist, we have our checklist. So when I talk to a company, I, this is kind of interview question, but it doesn't happen like this. And then this helps us to look rationally at whether a solution is promising or not. Then it doesn't mean that a company, a solution needs to respect all the criteria. But at least we understand what are the risks that we are taking.
48:46 Craig S. Smith? Yeah. Okay.
48:49 Speaker 00 And so the last component really is about business and scalability. So it's about understanding how the tool will integrate in the final application, whether standards will be met, whether all the question of responsible AI, cybersecurity. And also, I think, a very important point to insist on, whether the solution will be used. And this is where we come back to our discussion on AI adoption, raising awareness, AI education. This is also key when you evaluate AI solution to make sure this solution will be understandable
49:35 Speaker 00 to the final user. Because if it's not understandable, if the user doesn't like it, the UI is not good. All the rest is for nothing.
49:46 Craig S. Smith? Yeah, and you do that last step in collaboration with an expert in the business unit.
49:57 Craig S. Smith? And you, for example, evaluating the UI.
50:01 Craig S. Smith? Do you have a team that is very familiar with all the UIs that are coming at you and can look at it and just intuitively judge whether it's a good UI or not?
50:19 Speaker 00 The team is the final user. So all the POCs that we are doing, it's with the final user. I don't do POC just with AI experts.
50:33 Speaker 00 So that's the core. In fact, I should have started by this. When we do this evaluation, we form a team of AI experts from the AI head, but also of the business unit. That they know the business problems they need to solve and also the final user of the solution and we work end in end.
50:59 Craig S. Smith? I see and in you have vendors that are pitching to you constantly, I would guess. Do you, does it start with a business need and then you go out and scan the horizon for solutions that purport to address that business need and then evaluate, you know, three or five simultaneously or are you looking, do you narrow it down before you go through this more rigorous scientific evaluation?
51:39 Speaker 00 So it's a good question again. So there is in fact two approaches. So either top down approach, we start from business case and we narrowed down the different solution in the market. It's better if we test different solution, but sometimes we have narrowed down and there's one solution that look promising and we just go ahead with this solution. I mean, we don't have to find necessarily the best solution in the world. There are so many of them. We need to find a solution that work well for us. So that's a top down approach,
52:19 Speaker 00 business case and then restricting the problem and testing. And then there's the other approach, more bottom up approach. We are contacted by different vendor. There's new technology in the market constantly. And so we also try to see what this technology, what type of problem this can help us solving. Okay. Things that we are doing in a certain way today, should we revisit or not? So it's a bit more techno push. So we have to be cautious in this in always coming back to a problem we want to solve, but we have to do this as
53:00 Speaker 00 well, because maybe our process today, it's not efficient and we can have huge gain.
53:07 Craig S. Smith? Yeah. And how often, I mean, there's a certain bias that develops when you're
53:15 Craig S. Smith? Talking to a vendor and looking at the product. Uh, at a certain point, you've invest invested so much time in looking in the product that there's a bias to to accept it. I mean, how often do you do you go through that process and decide uh that this tool really isn't the right fit?
53:37 Speaker 00 Yes, so this process, uh this methodology, we we try to do it um to apply it a lot on because we have a wide diversity of application and problem we need to solve. So we apply it a lot. And on your points, um uh we we try to go up to the end of the process, but for for example, recently, a few weeks ago, we um we we we did a technical evaluation uh with a startup on two different uh use cases. And consistently, uh we saw that uh the solution was not working very well. Okay?
54:21 Speaker 00 So uh what's uh we we decided is not to go uh forward today. Okay, so it's it's on pause. We didn't go uh to the end of the process. Uh and we concluded that the solution is not mature enough today. And so probably in one year from now, or we we'll have to revisit uh and see whether with new tools we can uh solve this problem.
54:46 Craig S. Smith? Is there something else that I we haven't covered?
54:51 Speaker 00 Um, so you know, Craig, I think I would really leave uh you with with this um message. Let's not try to make P O C test successful, but really conclusive and all our discussion with a scientific methodology. Actually, you suggested a course to follow on on Coursera to develop this methodology. I think this is the the core of what we should do because this, what we are discussing, it won't be automated by AI agent tomorrow. This we share with with us and this is where we will have our value.
Previous #277 Alex Salazar: Arcade's Vision to Make AI Agents Secure and Scalable Next #279 Matthew Carroll: Immuta's Approach to Secure, Scalable Data Access in the Age of AI Page updated November 1, 2026 · transcript and summary are machine-generated
