Transforming Supply Chains with AI: Insights from Schneider Electric's Madhu Hosadurga

Can AI revolutionize the way we manage complex supply chains? Discover how Schneider Electric is leading the charge in transforming inventory management, production scheduling, and customer experience through artificial intelligence in our latest episode of the Chain Reaction Podcast. We sit down with Madhu Hosadurga, VP for Artificial Intelligence Internal Office at Schneider Electric, as he shares his journey from General Electric and Wipro Technologies to founding Oxalytics and eventually joining Schneider Electric. With a diverse and extensive background, Madhu offers invaluable insights into the interplay between AI and supply chain optimization in a global leader with 220 factories worldwide.

Madhu breaks down the intricacies of optimizing inventory management using AI, explaining techniques like single echelon and multi-echelon inventory optimization that help maintain optimal safety stock levels while minimizing waste. Learn how just-in-time inventory practices, bolstered by AI-driven recommendations and ERP systems, streamline operations for greater efficiency. We also explore the role of digital twins in providing real-time visibility and control, predicting maintenance issues, and addressing quality concerns—ultimately contributing to Schneider Electric's mission of sustainability.

Finally, we tackle the critical and complex subject of AI governance. Madhu discusses the importance of ethical AI deployment and the governance frameworks necessary to mitigate risks and biases. Hear how traditional forecasting models are adapted for non-CPG industries at Schneider Electric and how AI is enhancing customer experience by predicting potential issues through "weak signals." This episode not only elevates your understanding of AI in supply chain management but also underscores its pivotal role in enhancing both operational efficiency and customer satisfaction. Don't miss this fascinating conversation that promises to shed light on the future of industry automation and energy management.

You can follow Chain Reaction on LinkedIn, Twitter and Facebook

Chapters

• 0:00 Supply Chain Advantage With AI Optimization
• 11:18 Optimizing Inventory Management With AI
• 24:01 AI Governance
• 35:33 Customer Experience Enhancement Through AI

Transcript

Tony Hines: 0:00

Hello, tony Hines h you're listening to the Chain Reaction Podcast, all about supply chain advantage. Great episode coming along. In just a moment we have the VP for Artificial Intelligence Internal Offers from Schneider Electric, and that's Madhu u. Schneider Electric is a global industrial technology leader, specializing in electrification, automation and digitization. The company provides products and solutions for various sectors, including smart industries, resilient infrastructure, data centers, intelligent buildings and homes, and I'll let Madhu say more about his role, and he can tell you a little bit more about the company in the introduction. So, without further ado, it's my great pleasure to welcome Madhu Hosadrga to the Chain Reaction Podcast. Hello, madhu, great to have you on the show today. Can you tell me a little bit about your role at Schneider Electric and perhaps, for the listeners out there, you can say a little bit more about Schneider Electric and what they do?

Madhu Hosadurga VP AI Schneider Elec: 1:04

First a quick introduction of myself. My name is Madhu Hosadurga. I've been in the AI sector for close to 22 years now, started my career with General Electric, spent about six years with them helping them in the Six Sigma manufacturing quality, and then I moved to a company called Wipro Technologies, spent eight years with them helping various Fortune 500 companies. And then I started. In fact I went to do my MBA at Oxford, took a break, came back, started my own startup called Oxalytics, which I ran for about four years and did not have a great success. So I ended up joining Schneider right after that in the capacity of head of.

Madhu Hosadurga VP AI Schneider Elec: 1:53

AI for what we call as global supply chain function within Schneider Electric, and then, after serving in that role for about three years, I moved on to my current role, which is called as AI Hub. Now, within that, I play the role of head of AI for internal offers and, as the name suggests, I head a function which leverage AI for our efficiency within the company, employee experience and customer satisfaction. So this is my overall experience and current role at Schneider.

Tony Hines: 2:30

Thanks, Madhu. Can you tell us a little bit about Schneider too?

Madhu Hosadurga VP AI Schneider Elec: 2:34

Okay, I'll quickly tell about what Schneider Electric is into. Schneider Electric, as you may already know or have heard, is a leader in two main sectors. One is what we call as energy management and the second one is what we call as industry automation. So it is an organization with close to 160,000 employees across the globe, with close to 220 plus factories across the globe, with one of the very complex supply chain. So that is also a 150-year-old, one of the reputed companies.

Tony Hines: 3:14

Yeah, I mean I'd done some background research and I was surprised to see how old that Schneider actually is, because you don't think about it. And then I checked on a few things, and when you talk about energy products and so on, I even have some of those products in my house here. So yeah I am aware of you know yeah but it's the unseen part, isn't it? We don't see them.

Madhu Hosadurga VP AI Schneider Elec: 3:35

It's all kinds of stuff around really yeah, in fact, um, you know schneider, while people people may not necessarily recognize the name, but it's actually a mix of many brands. For example, many people may use the products like APC and they may not necessarily know it belongs to Schneider Electric. So, for example, even in India we have products under the brand Luminous, which is one of a very local popular brand, again part of Schneider Electric. Many people may not know. So it's a company which has many brands within itself. So, yeah, that's the nature of the company.

Tony Hines: 4:19

Yeah, and if we go back to the idea of some of the projects that you're working on currently in your particular role with Schneider, I mean, I read about some of the sustainability work that's happening and some of the stuff to do with how AI is helping with ESG and that kind of thing, but I think your particular interests are going to be in some other aspects of what's going on at Schneider, so perhaps you can explore some of those with me.

Madhu Hosadurga VP AI Schneider Elec: 4:49

The one top problem or challenge which I'm trying to solve in my current role is, of course, efficiency.

Madhu Hosadurga VP AI Schneider Elec: 4:57

See, the reason is, schneider is a very large organization, a complex organization, so it's very important for us to keep a tab on our efficiency, and that is the problem me, in my role as AI vice president for internal offers, trying to solve. If I had to give you an example, we sell about 400,000 products across the globe, produced across 220 factories distributed across 100 different international DCs. It's a very complex operation, and then the efficiency of our supply chain of course, depends on how much inventory we carry at any point in time in different locations. So this is a very complex problem. It's not something humans can solve it with simple processors and IT systems. This is one of the problems we solve using AI. One of the problems we solve using AI, right. So we use, you know, optimization techniques, using machine learning to recommend how much to stock in which location at any point in time, right, not only finished goods, also the raw materials which go into it, so that is one problem.

Madhu Hosadurga VP AI Schneider Elec: 6:21

Similarly, we also leverage AI for things like to forecast demand at any point in time, which product, how much to produce, how much to stock, and then we also optimize our production scheduling.

Madhu Hosadurga VP AI Schneider Elec: 6:38

For example, when you have 220 factories, each one of them hundreds of production lines within it, and now you really have to, on a daily basis, uh, plan those production schedules, yeah, and then each of those uh uh production lines need to have the right amount of raw material, workforce and all of that, yeah, imagine, uh, uh, you know, humanly trying to plan this, it's not possible. And even if it's possible, it will never be efficient and optimal. Yeah, this is another example of the problems we try to solve using ai. Yeah, in my current role and again you know that is mostly what I told you or is on the chain. Similarly, we have similar challenges in finance, which is another major function for us. We have similar challenges in sales, marketing, e-commerce within your enterprise, it, data and governance, so on and so forth. So, across the company, we have a lot of challenges, which includes forecasting, predictions, optimizations, as well as dealing with a lot of natural language, what we call as unstructured data and then using all of this to automate, optimize, predict.

Madhu Hosadurga VP AI Schneider Elec: 8:02

So, if I had I mean happy to give you more examples, but I'll let you ask me if any other questions you may have.

Tony Hines: 8:09

Yeah, well, that's interesting. I mean, you're working across the piece, so you're not only focused on supply chains as such with your role, but you'll be working with people from finance, so you might be looking at accounts receivable, that sort of thing, or you might be looking at accounts payivable, that sort of thing, or you might be looking at accounts payable, uh, in terms of stock and so on, but I suppose we're most interested in this particular um talk today about the impact on the supply chain and and so, from that point of view, perhaps you could explain a little bit about how you control the inventory through ai and perhaps give us some some examples of of the successes you've had with keeping stock levels down, or indeed, perhaps with disruptions and the move towards resilience, where lots of people are talking about coming back to near shore rather than offshore, so trying to reduce the complexity in that sense, how are you getting the stock to the right place at the right time?

Madhu Hosadurga VP AI Schneider Elec: 9:08

Yeah. So if you look at the inventory topic within supply chain, you have multiple parts to it. You have components which come from various suppliers and that has to be stocked at the right levels for us to be able to produce what you want to produce. At the same time, you should have your production lines ready to be able to produce them, and then you need to have the finished goods stocked at the right level. So this know what we mean by inventory optimization yeah, and if you really look at it, how much to stock?

Madhu Hosadurga VP AI Schneider Elec: 9:48

while it sounds like an easy question, it is a very complex and a pure function of demand and supply. Yeah, so how do we keep it at the minimum level? Right, in fact, we have, uh, two kinds of stock level which we maintain. One is, of course, the stock which we sell, so we maintain an inventory of that. Second, we also have something called as safety stock, which we also maintain at a certain level, mainly for our resilience and to cater in case of situations which are not in our control. Cater in case of situations which are not in our control. So now let me talk about first the inventory optimization of the finished goods and then I'll come to the safety stock. So, in terms of how do we maintain our inventory at the optimum level On the safety stock level and this is kind of coming from the resilience angle you mentioned earlier, right, we in supply chain, of course, we deal with a lot of complexity, not only internally, even external complexity, like sometimes geopolitical, sometimes shortage of some raw materials, sometimes as simple as dependency on the external containers and ships and things like that.

Madhu Hosadurga VP AI Schneider Elec: 11:18

So one way we try to be resilient despite all the complexities, by maintaining a certain amount of safety stock in our inventories, yeah. So now again, the big question is how much of safety stock to be maintained in different location, different components? Yeah, and then there are different techniques, like single echelon inventory optimization, multi echelon inventory optimization. What does that mean? Single echelon is you consider a single distribution center and then you optimize your safety stocks around it, whereas multi-echelon is you consider a network of distribution centers and then you see how much to maintain, considering the network. For example, if you have a nearby DC, you can almost borrow it from you know that DC. So again, again it's. It's a complex AI problem which we leverage AI to maintain an optimum stock.

Madhu Hosadurga VP AI Schneider Elec: 12:24

Again, in my view, maintain an optimum stock. Again, in my view, this is also something which contributes very well to our sustainability goals, because maintaining inventory at the suboptimal level is not a good thing for the business. If you keep more inventory, of course you have left a lot of value on the table which you may never use. At the same time, if you maintain less inventory, you're not able to cater to your customers, and for us, if we are not able to sell to our customers, that means an opportunity lost for our customers to be more sustainable, because the products we sell will help our customers optimize their energy consumption right. So that's where I think it's important for us to maintain both the finished goods inventory as well as safety stock inventory optimum at any given point in time. So this is one of a great AI use case, I would say.

Tony Hines: 13:26

Yeah, that's a very important use case and you've explained why it's important and you've explained its link to sustainability. I think that's great, and I think also perhaps I'd ask the question about how do you manage things like just in time then? Have you got any particular policy towards just in time in the company or in particular lines in the company? And if so, I suppose how does that fit in with the overall AI intervention into inventory management?

Madhu Hosadurga VP AI Schneider Elec: 13:57

Sure, we use methods like Kanban and stuff like that to manage our daily tasks, tasks and just in time goes in hand with the inventory. What I talked about, the optimization of inventory, and here you are looking at mostly on the component side, right? So while we talked about normal stock of finished goods, here, the same is applicable on the component side, and that's what helps us to produce things and procure things just in time.

Madhu Hosadurga VP AI Schneider Elec: 14:32

For example, if I know the demand and if I know how much finished goods to stock, if I know what's my production schedule, I would know how much components to stock as well and how much to procure from our suppliers. And this way we can also collaborate with suppliers in giving them enough lead times to produce and service to us, again end-to-end. If you look at it, it's a very well-orchestrated engine assisted by AI, and that's what helps us.

Tony Hines: 15:09

you know, do things, uh, you know just in time and keep enough to serve our customer needs yeah, and does this help people in your warehousing, in your dcs, in your various locations throughout the globe have visibility about what's going on in supply chains? Can they actually see into the system?

Madhu Hosadurga VP AI Schneider Elec: 15:33

Absolutely so. Every, in fact, one of the principles of AI within Schneider Electric is every algorithm we build, every model we build, the output of that goes into an ERP right. The reason we have adopted this principle is that way we ensure each of our solution gets adopted very well. Yeah. For example, when I talk about inventory optimization and I talk about recommendation of the right stock levels, it's the same thing which goes into ERP. Sometimes the users may not even be aware that it's done by AI, but what they see is okay this is what I need to maintain and they follow the instructions At every level, from procurement to manufacturing, to logistics, as well as in the sales yeah, so so that that's where the visibility comes.

Tony Hines: 16:33

So that's really good. I can see, and being unaware of these things, I think, is quite a good thing in one sense, because they're seeing what they need to know without worrying about the technical aspects, and you're handling that behind the scenes. I think that's that's the way things are working, and I think also in terms of um, I'm just wondering where you are with your notions of things like digital twins in supply chains. Do you have anything to say on that?

Madhu Hosadurga VP AI Schneider Elec: 17:01

yeah, again, digital twin is another vehicle, how we use technology to be able to manage such a large supply chain. And you know multiple factories, what we have, in fact, you know that gives us real-time visibility. You know, thanks to the, the iots and the edge computing and being able to see what's happening at any point in time, being able to monitor and also sustain operations. So all of our factories are smart factories. All of them are connected. We have data coming in every second into a cloud, which we are able to see what is happening, take the right actions, have a full control of the different operations going on at any point in time.

Madhu Hosadurga VP AI Schneider Elec: 17:58

So, for us, digital twin is the way we monitor and run our factories and the DCs monitor and run our factories and the DCs. And then, of course, there are a lot of AI that is being put into the data that's generated in digital twin, so we are able to predict any maintenance issues, we are able to predict failures, we are able to predict quality issues while we monitor using digital twins. Just saying digital twin mainly helps from a visibility point of view. And then that data, which is because digital twin can use you a digital footprint of everything which is happening in real and when you have a digital footprint's like a gold mine of data for you to use the eye. Yeah, yeah, so that's how I see digital twin.

Tony Hines: 18:54

Yeah well, I think that's a good way to view it. I think that's you know. That's the whole idea, isn't it that you've got the, I suppose, almost a simulation of what's going on in the reality?

Madhu Hosadurga VP AI Schneider Elec: 19:04

uh, as, well, it is a simulation and a reflection actually.

Tony Hines: 19:06

Yeah yeah, exactly so. So that's really good. So, when we think about some of the skills that are involved in ai and the and the sort of people I mean, obviously one of your roles will be to build a team of people around you who can solve the problems in the organization. And when you're looking for people to join uh and the sort of career opportunities that might present themselves, can you say something about that, because that might help people who are also listening to understand, if they're looking for roles in this area or they're looking for opportunities, how to get themselves in a position to do so.

Madhu Hosadurga VP AI Schneider Elec: 19:43

So is your question more around what kind of skills we look for within the AI team? Yeah, Sure, sure, sure, See, within our hub we call it AI hub we have mainly four kinds of skills, if you like.

Madhu Hosadurga VP AI Schneider Elec: 20:06

One is what we call as AI product management. Right, and the product management is all about having a good knowledge around how to build AI products, which includes stakeholder management, being able to create a business case, being able to create a product roadmap, so on and so forth mainly around product management. And the second area is data science People who are skilled in data science, being able to create data science models. And the third one, what we call as engineering team, who are basically take this data models or data science models and embed it into a good engineering solution. Yeah, so that's the third type of skill.

Madhu Hosadurga VP AI Schneider Elec: 21:00

And the fourth one is mainly around what we call as MLOps, which is a skill which is required to run the machine learning operations. These are the teams who build the required platforms, run the operations in terms of CICD pipelines and then create different components of the platforms. We need to deploy these solutions at scale and then, on top of it, fifth one is also what we call as AI scrum masters, who are able to bring all of these people together in building use cases and the products. These are like four or five key skills, uh, what we usually look for, and then they constitute what we call as a help today right, okay, that's.

Tony Hines: 21:54

That's really interesting, madhu. I think, uh, I think that's given us some understanding of some of the opportunities and some of the roles that people might be thinking about as they develop their skills in the in this particular area, and of course I mean operational research, mathematics. All that kind of background in terms of developing those skills is going to be really useful to these people, as as, indeed, our computer skills, aren't they?

Madhu Hosadurga VP AI Schneider Elec: 22:20

absolutely when, when we say data science, it's all about you know the mathematics, it's all about the machine learning, the deep learning. Again you know nowadays, uh, you know when, when we talk about uh, you know the mathematics, it's all about the machine learning, the deep learning. Again you know nowadays, uh, you know when. When we talk about uh, you know machine learning. People have probably moved on from the, the mathematics. Now they kind of use the, uh, the models which are pre-defined, pre-trained and stuff like that. But again, the underlying principles are still same. Now it's it's a very complex mathematics, mathematical algorithms, exactly.

Tony Hines: 22:51

I mean it's algorithms that drive the world now and they drive the supply chains, and obviously data science is a newer term that encapsulates all the skills that would happen under that particular umbrella, I suppose. Yeah, absolutely happen under that particular umbrella, I suppose?

Tony Hines: 23:08

Yeah, absolutely. So I'm thinking about some of the big developments in AI and one of the things that strikes me and one of the concerns that people have and have raised publicly is to do with the ethics. And I'm just wondering about the role of ethics in AI and how important is ethics in AI? I mean, I was reading Suleiman's book on you know his experience as a deep mind and so on, and I just wondered whether that was something that's right up the agenda now for everybody working in AI.

Madhu Hosadurga VP AI Schneider Elec: 23:42

Yeah, no, it is a very valid concern, tony. It is a very valid concern, tony, especially when you have machines doing day-to-day tasks, some without human intervention, making sure they do the right thing is very critical. And AI, because of its nature, has some issues like mainly around the bias. So there are different types of bias. There is usually, you know, a recency bias, which happens because it learns, you know, from the recent events. Sometimes that could be misguiding, sometimes it could also lead to some unwanted biases, like gender bias and stuff like that.

Madhu Hosadurga VP AI Schneider Elec: 24:27

Again, it learns from the data, from the historical transactions and it tries to mimic the same way for future, which is a problem because we are trying to change a lot of things in the world and human beings today are going through massive change, management, transformation from you know, inclusion and many other aspects. And then if you want machines to kind of learn from the past, which could be a problem and which could lead to some unwanted you know biases, which is one problem and the second, the more and more you have kind of machines do day-to-day operations, you are always at risk of cyber threats. Imagine, if I have to give you a very layman example. Let's say you have self-driving cars today. It's a reality, people are using it. Imagine if your car gets hacked. You are on a highway, car is driving on its own and your car gets hacked.

Madhu Hosadurga VP AI Schneider Elec: 25:28

So same problem can happen in organizational operations because it is digital and then digital comes with this risk. So that is why, in my view, ai file comes with a lot of advantages. It also comes with a little bit of risk, just like any other digital operations. Here the risk is also more because you are kind of using it as an alternate to human intelligence. So here you are trying to keep human involvement to the minimum.

Madhu Hosadurga VP AI Schneider Elec: 26:01

So the only way this can be avoided or controlled is through a proper governance. So this is why, even at Schneider, every product we build go through a very strict governance process, very strict gate review process, to ensure every use case or product is certified from all of these threats. So a lot of questions get asked in each of these use case. For example, does it use any personal information? Does it use any gender information? Is there a recency bias which we can avoid? For example, when COVID happened, most of the machines start to learn, thinking COVID will repeat yeah, right, so which was wrong? Because we had to even remove the COVID period data from, you know, machine learning, because we had to remove it to avoid that bias.

Tony Hines: 27:05

How interesting. So you had to take out the COVID data to remove that kind of bias in the system.

Madhu Hosadurga VP AI Schneider Elec: 27:11

Yeah, I know it is very interesting and also a bit scary at times. And this is where we believe especially if we have to do it at scale that governance is not an option, it is a must and every organization have to have a very strong governance under the umbrella of AI. In fact, this is where I strongly suggest organizations to see AI as a function on its own, not necessarily a small part in some corner of the it organization, because the moment you see this as a separate function, I know most likely you will have a c-suite leadership, you know heading this, and then, the moment you have a c-suite so you can have a good process, you can have the required, you know the resources to run it at scale and at you know, ethical fashion.

Tony Hines: 28:13

Yeah, so it's very important. I mean, what the statement is that you're making is a plea for companies to have people with those necessary skills on the board, so they need to be in those c suites. And once you're in the c suites, as you quite rightly say, you get the resources which you don't necessarily get the resources you can mandate as well, correct, yeah?

Tony Hines: 28:38

understand absolutely. So I think that's that's a very important point, madhu, I think. And the other thing about recency. I mean, I'm going to flip a little bit from what we're talking about right now and just go back to the inventory situation. I think that's one of the problems with inventory. If you've got a model and you're looking at recent situations and say you're looking at pinch points around the Malacca Straits or anywhere else in the globe, or you're looking at disruptions that are taking place and you're looking to try and predict, say, container boxes coming from X to Y and backhauling and getting them to the right place at the right time, yeah, if you went on a modeling exercise that tried to encapsulate that and you looked at the recency, that would perhaps give you not the answers that you might wish to act on yeah right in fact.

Madhu Hosadurga VP AI Schneider Elec: 29:28

Um, if you it's. It's a very interesting question, tony. If you really look back, the origin of forecasting comes from a CPG world. Cpg meaning consumer packaged goods, and if you look at the nature of CPG business, they have a repeat kind of consumer goods. For example, a person buying a shampoo would buy every third month. A person buying a toothpaste would buy every month. It's a repeat nature of business. And then, as long as your population is kind of not changing too drastically, forecasts work very well in those businesses. Absolutely, absolutely, yeah. But what happened?

Madhu Hosadurga VP AI Schneider Elec: 30:17

All other companies started to build on this principle of history-based forecasting and that is why many companies struggle to get good accuracy in a non-CPG kind of business, especially companies like Schneider, if you look at we are not in aPG kind of business, especially companies like Schneider. If you look at we are not in a CPG kind of a business. For example here, if I sell a UPS, that would last five years and if somebody buy, let's say, a breaker, that would last 10 years. So here the question of repeat does not necessarily come into our yearly forecasting and near-time forecasting and stuff like that. So now, how do we solve that? How do we still kind of predict our future sales.

Madhu Hosadurga VP AI Schneider Elec: 31:09

So here is where we augment historical sales with the events happening around the external world. What I mean by that? For example, who is our customers? If I look at Schneider Electric, one of our big customer sector is real estate. Yeah, so now there are a lot of external micro and macro indicators which tell me the growth in the real estate sector, both residential, commercial infrastructure and so on and so forth. So that is a very good input for us to augment with the historical data what we have. Yeah, right, so we kind of use the historical trends within the company and then we also augmented, enrich it with the external data to make a better forecast. Of course we also have some consumer products within the company and then we also have, you know, these massive projects kind of products. Yeah, I think you know the answer is in using internal, in augmentation with the external data to be able to forecast better.

Tony Hines: 32:21

Yeah, yeah, I can see that, and I suppose the proportions of the data that you choose to use in the forecasting models will vary according to the particulars of the product going through the process yeah, again in the, the way we build models is in such a way that the model itself will understand what is the nature of product.

Madhu Hosadurga VP AI Schneider Elec: 32:43

Is it a fast moving? Is it a rare moving? Is it, like you know, somewhere in the middle, and then it's able to forecast.

Tony Hines: 32:53

Looking at all of that into consideration, plus the external information, yeah, I think that's very interesting and I think one of the key things that you said about the forecasting and the development of forecasting, which is very important when you're looking at supply chains and you're looking at demand, because one of the things you want to focus on is the customer at the end of the line, you want to get the goods in the right place at the right time to that particular customer. And if you're looking at the forecast historically, as you rightly pointed out, the past is not necessarily a continuation into the future. So that recency or that bias that's coming through the chain needs to be somehow moderated in the process, and that's where the modeling comes in, and it's also where the AI needs to learn through the machine learning I suppose that's taking place so they can actually adjust and become more nuanced in how it deals with these things.

Madhu Hosadurga VP AI Schneider Elec: 33:48

Absolutely, absolutely. You're right, and this is where machine learning at scale really helps.

Tony Hines: 33:55

Yeah, now another thing I was interested in when I was looking across the pieces. Schneider was about the microgrid project and that seems to be a very important piece of work that's taking place in the development of the microgrids. Are you involved in that also place and the development?

Madhu Hosadurga VP AI Schneider Elec: 34:14

of the microgrids. Are you involved in that, as also I'm assuming? Yeah, I have a colleague of mine who looks into all the ai which goes into our external offers, while my focus is mostly on the internal efficiency. But again, it's a you know, similar principle. We use ai and micro gates to kind of manage, you know, energy consumption and make it efficient at any point in time. So the principles are same more of optimization techniques, being able to predict, forecast and optimize.

Tony Hines: 34:48

When you talk about the internal offers and so on in which you're involved. Just explain a little bit more about what that would actually cover in terms of how you develop AI for those internal offers. What's the thing that tasks you or?

Madhu Hosadurga VP AI Schneider Elec: 35:09

challenges you. So, in fact, uh, yeah, uh, you know the, the key objective for internal offers is being able to drive efficiency at the organization level, being able to create a good employee experience and being able to, of course, create a good customer satisfaction levels. So now, how do we do that? So we, as an AI hub, work in what we call as a hub and spoke model, and the spokes are actually the different business functions we have in the organization. Supply chain is one such function. Whom we work with in enabling AI. Similarly, we work with finance, sales, marketing, field services, e-commerce, hr, enterprise IT, data and governance. So, put together, we have close to 11 or 12 functions and five regions. The Schneider has kind of grouped different parts of the world into five regions. So the five regions and 11 functions put together is my scope. I work with all of them in a hub and spoke model in solving various problems.

Madhu Hosadurga VP AI Schneider Elec: 36:27

That would lead to one of the objectives, as I said efficiency, employee experience, customer satisfaction.

Tony Hines: 36:34

That's great, madhu. That's a very concise and very good explanation of how you're working. And I think, when it comes down to the customer experience, what sort of data are you working with from customers to know how well you're doing, how well Schneider is performing, and so on? And you've obviously got systems in place that do this, and the AI is very important in that process.

Madhu Hosadurga VP AI Schneider Elec: 37:03

I would have thought yeah, so there are many ways we use AI to measure the customer experience. Yeah, I can give you some examples. So we use AI, for example, to read every customer query we get and then AI is able to, for example, classify into the right category and it transfers it to the right agent for a quick response. So this is one of many use cases which would address customer queries much faster. Similarly, we have another AI which looks at all the customer you know, queries and issues, and then it is able to summarize it is able to help our customer care agents as well to be able to resolve much faster.

Madhu Hosadurga VP AI Schneider Elec: 37:57

Yeah, and then we also use the customer interactions data, for example on our websites and e-commerce shops, to improve the experience for our customers. Recently, we even have introduced a capability like ChatGPT on our website, where users can converse with our website asking questions like hey, can you recommend me a product for electric charger for my home? Then it would help you navigate, I know, in a very conversational fashion for searching a product. So there are many, many use cases in the customer satisfaction space which involve a lot of language and structured data, and we use a good amount of Gen AI capabilities in solving those problems.

Tony Hines: 38:59

Yeah, that's great, isn't it? I mean, that's where I can see it really coming into its own, in a sense, where people are coming in and just using ordinary language to find the way around the website and get some guidance on the site. So I think the future in that way is is is with us. It's happening now, isn't it?

Madhu Hosadurga VP AI Schneider Elec: 39:17

not many people understand uh, you know what they need. When they mean I need electrical setup for my home, they don't necessarily understand everything which goes into it. Yeah, so this is where we can guide them a bit in a very conversational fashion within our website to help them make the right decisions.

Tony Hines: 39:39

Right, okay.

Madhu Hosadurga VP AI Schneider Elec: 39:43

And then there are a lot of.

Madhu Hosadurga VP AI Schneider Elec: 39:44

In fact, one of the first use cases was what we used to call it as weak signals. In weak signals, what we used to do. You weak signals, you know, in weak signals what we used to do you get the customer, you know issues or, uh, you know requests, and then we have ai, look into, you know the category of issue and then at the same time, we try and link it to the quality issues we may have had in the manufacturing facility that time which the engineers may have put some comments in the manufacturing facility, that time which the engineers may have put some comments in the complaints and things like that, all the way to the supplier quality data. Even our suppliers are obliged to give us quality information what failed in their production line, nature of failure and things like that. So we used AI to connect the dots from supplier to manufacturing quality, all the way to the customer issues. And we used to call this as weak signals, any signal which is able to help us proactively identify any major issues that may come in the future.

Tony Hines: 40:52

Yeah, that's great.

Madhu Hosadurga VP AI Schneider Elec: 40:54

Huge potential.

Tony Hines: 40:56

Yeah, fantastic potential. Yeah. Well, I think you've given us a very good indication of your role, some of the work that you're doing and, of course, what Schneider's doing and how it's taking things forward, and it's been a very illuminating session really, and I've been very interested to hear those developments. And thank you for coming along and talking to us today. It's been a real pleasure. So thanks very much, Madhu.

Madhu Hosadurga VP AI Schneider Elec: 41:23

Thank you, tony. I mean it's a pleasure talking to you as well. Very interesting discussion.

Tony Hines: 41:29

That's great, and perhaps at some future date you might come back and talk to us again if you'd like, to my pleasure always. Thanks very much, Madhu. I'm going to sign off now.

Madhu Hosadurga VP AI Schneider Elec: 41:41

Thank you, take care. Bye-bye.

Tony Hines: 41:42

Bye-bye. Well, there we are. It was a very interesting conversation with Madhu Hosadurga, the VP for AI Internal Office at Schneider Electric, so a big thank you and a shout out to Madhu and to Schneider Electric. Well, that's it for this week. I hope you've enjoyed the episode. I hope you've learned something, and if you don't already subscribe to the podcast, subscribe so you'll be first to know when a new episode is coming your way. I'm Tony Hines, I'm signing off and I'll see you next time in the Chain Reaction Podcast. Bye for now, thank you.