From Supply Chain Analytics to Agentic AI | How AI is Transforming Supply Chain Decision Making

[music] Thank you for joining our open webinar today hosted by the MITx MicroMasters program in supply chain management. As academic year begins here in Cambridge, Massachusetts, we are excited to have several of our courses opening at this time that align with our discussion today and more details will be shared through the chat. I'm Eva Pon. I'm the director of online education at MIT, the center for transportation and logistics and the founder and director of the MIT omni channel supply chain lab. Today we are going to explore how AI is transforming supply chain decision making with VJ Sankar Raman Chief AI officer at Novan Health which is a leading healthcare organization. Welcome BJ and thank you so much for joining us today. >> You are pleasure to pleasure to be here. Thank you uh for having me here. It's a bright day. I mean, especially the way we all have, you know, registered ourselves into the the new transformative AI era and how we are powering ourselves to move through the supply chain and more things. I'm very happy to share that. >> Thank you, VJ. Really an honor to have you joining us today and we have a big audience joining live too. So, welcome, welcome everyone. In this session, we will start by looking at how AI and advanced analytics are being used to make better predictions and improve inventory decisions. Then we are going to explore the evolution from predictive AI to generative and agentic AI. And we are also going to focus on how these technologies are changing the way supply chain decisions are made. We will also talk about what these changes mean for supply chain professionals and the skills that will be important moving forward in the future. But before we begin, we really want to learn more about our audience. So let's try to launch a fun poll in order to know how often are you using AI for work. Let us know in the last month how often you use AI for work. Let's see. And while you are responding to this fun poll, let me briefly share the plan for this webinar. We will start learning about BJ's leadership journey and his current role. Then we will explore two use cases that show how AI tools are transforming supply chain decision making process. Then we will explore two use cases that show how AI tools are transforming supply chain h decision making process. And finally we will discuss what capabilities supply chain professionals need to develop to lead these transformations. Let's try to have a look to the first poll and see how often are you using AI for work. Okay VA more than 60% of our participants use it daily or almost daily. Um 25% several times per week. So definitely most of our participants are definitely using AI for their work. But now I we really want to know a little bit more about that use and please let us know in which areas of your work you are currently using these AI tools. Let's see if it is for demand forecasting, inventory management, transportation, supply chain risk management. So poll two is already uh open. Please erh go to that poll and let us know. Um while you are responding to this poll, I'll be sharing some guidelines for this webinar. So if you have questions, please [clears throat] use the Q&A feature. We want this webinar to be interactive. So be sure that you log with a name and we will also launch additional polls during this session. We so try to to to keep engaged with with the speaker and and with me during this webinar with several of our courses having just started or opening soon. We will also share program and course resources in the chat throughout today's event. So keep an eye also to the chat and let's try to take a look to the results of the second poll. Let's see. Okay VJ this is interesting because I think we have 22% of our participants use AI for demand forecasting and planning we are going to talk about that 20% for inventory management and optimization 16% for transportation and logistics 14% used for procurement and sourcing and I know we are going also to discuss how you are using Aentic AI for procurement and sourcing seen at Nobang Health, 13% used in the environment of manufacturing and operations and 15% for supply chain risk management. Super interesting. Thank you. Thank you for sharing with us uh how you are using AI in your day-to-day job. But now let me formally introduce our industry speaker. So BJ is the chief of AI officer at Novang Health and he has been in that role for about two years. His previous experience was in the retail industry. He was the vice president of product and technology at Lois for five years and he also had the role of being the head of product management and technology at Walmart for three years. BJ has an extensive experience working with technology advancement product development in different uh sectors. So BJ healthcare is a sector very different from retail. H working to improve people's life definitely open a completely new paradigm. Could you tell us more about about that and also about your leadership journey to get there? Eva again thank you for having me here and a kind introduction as well right I mean first things first the way I have brought my upbringing and been humble and fortunate to be part of a number of institutions you mentioned but all through my journey I've held on to one belief where who we serve and what problems are we solving are far more important Then the means to getting there because it evolves. Given that the perspective that we have held is working through a variety of consumer focused companies in Walmarts and the Home Depots and the loes has given me the perspective to look at what uh an everyday person including myself putting myself in that shoes. What friction am I going through in order to complete what I want to complete and how do I how do I get past that faster? Health care makes it more personal. Personal in the sense that each and every one of us at some point during our lifetime andor for our loved ones may have to make a trip to even to get a flu shot or or something of that sort. Right? So when you go through a variety of phases in our uh you know on our on our lives if you will those journey through the health care systems especially especially in the in the era that we are all in is laden with friction. Friction in the terms of how easy or not is it to get an appointment how easy or not is it to you know complete through a surgical journey. How easy or not is it through pay a bill understand the the the economic bandage. So when you go through all of that there is an element of supply chain there is an element of marketing there is an element of uh you know operations all runs behind the scene how can AI and technology and the advancements help move and eliminate the friction to make it easier for us that's the core purpose of uh me doing what I'm doing >> that's a very interesting angle VJ how to eliminate this friction for for patients and for people that really need uh to have a smooth process through through that. Uh let's uh let's start h by t taking a step back and looking at how supply chain analytics has evolved and and how companies have been using these tools to support decision making. So I personally I have been working at H in supply chain management for 25 years and truly statistics, forecasting, optimization, operations, research have always been kind of at the heart of supply chain decision making. But in in many ways AI is building also I believe on on that foundation of all of these operation research techniques uh for decades. Correct. So BJ in the last let's go back the last 10 15 years. Uh what did supply chain analytics look like and and how has this evolved since then? Uh mainly my my question focus more on what were companies primarily primarily using analytics for that at that time. >> Yeah. You know, I think if you all of us date ourselves back to your point, you know, say see 15 years ago, even even in that in the line of it, we were very busy creating dashboards. We were very busy creating u scorecards. We were busy in understanding you know what happened last month, what happened last year, what happened yesterday, what was my inventory, how much traffic did I run through, how much do I need to keep up with what potentially could come. All this was descriptive in nature. So essentially if I have to encapsulate in one word, it was descriptive analytics. Now that was in vogue, that was popular. We all heard the word big data. We have all heard the word uh ERP. We've heard a lot of things. And by the way, all of them are still around. Nothing wrong about them. But the word largely was fixated about leveraging that. The other piece I'll bring to the connotation is back in the time 15 years ago or so, supply chain from the genesis of it from my eyes at least largely was a pure cost center. Your purpose in life as a supply chain leader was to keep the cost as low as possible. which is perfectly fine. But as technology evolved, as the advancements evolved, more and more we walked into what we called as the deeper analytics aka the the early innings of AI, machine learning and all the fun stuff we can talk later came about and at the crux of it the move went from being what I would call it as purely descriptive to bit more predictive. Now, now I'm I'm not just merely looking backward with the available technologies in front of me as a planner, as a supply chain leader. Now I'm also being bold enough to go lot more predictions. Now in the process, I'm not just merely obsessed over cutting cost. I'm also looking at where I can make profit. We can talk more about it. So that those are the big changes. >> Yeah. Yeah. and and 100% know about this evolution from descriptive to prescriptive analytics and also what has changed is not simply simp simply that we have more data that we have or more computing power we also have that but also the types of patterns we can identify and the kind of decisions we can support is is also evolving. So what what how do you see machine learning and what change machine learning is bringing to the table in that evolution? >> Yeah. Um see first things first the the all the definitions of machine learning out there my simplest way of explaining uh to to to uh at the time my fifth grade now she's she's grown up to be a junior here um if a job is best done by a machine let the machine do it. >> Mhm. With that philosophy in mind, the advancements allowed the machine to be more and more I use the word machine in a in a very figurative way to become more and more immersed in what we do and learn the patterns, learn how we do what we do. And to your point, there are a few other friends of the family of machine learning came about. Cloud computing became more prevalent. Cloud computing became cheaper. All that allowed for gargantuan amount of data to be ingested. You're no longer married to your computer, your physical uh you know limitations of your computer. The cloud was wide open for you. So what that did was and the other pieces were in the traditional world at least 15 years ago or so you think about demand planning, think about inventory, think about transportation, everything that we saw in the in the poll few minutes earlier. These were humans with at the time available uh computing power computing knowledge ERPs had dealt with 1 2 3 four call it eight variables at a time. For example, I'm looking at uh you know last year sales history. I'm looking at uh the lift I would get with a promotional uh uh you know that I might be able to do. I'm also looking at uh you know um a causal factor like a weather. three or four factors were largely dominating how a science married with the art of the human would would help predict a transportation planning or a demand planning allowing inventory. Fast forward machine learning comes about learns and and ingest and all of this. Now you have an element of physicality or the physical intelligence getting into it. RFID, IoT, data sensors, the list goes on and on where the data that was only available otherwise on a weekly. How many of us can put your hand up and say, "Hey, you know, do you remember the weekly batch runs?" You know, every Saturday night and a Friday night, something would run to predict what needs to happen that week. Gone know those days where the amount of learning now you the compute power and the intelligence where you could do anything you want within the hour. I can I can predict the one route. I can predict the one store. You get the point. So what happens is that now collapse the window of how often anything can be done and situate more and more into the dis shifting the decision powers little bit more equally. Now in the process the variables are no longer limited to handful. you your variables could now be in my in my in my quest as a supply chain planner to predict how long would it take for me to go from point A to point B. >> Mhm. >> I have I can look at social events to say what you know uh what else is going to be happening in the route that day. Which event is happening? Maybe there is a a presidential campaign that's going to come by. I'm just going to throw one out there that that may have a traffic impact to something. All those factors are now lot more sourcable at the at the tip of an intelligence hands factors that and gives far more prediction. So I think machine learning created that uh element of jump from being reactive to lot more uh predictive if you will with the level of accuracy that you know in the in the last 15 years we've seen significantly go up. >> Yeah. and and one thing I found particularly interesting is how you connect the dots between prediction and better decision making. Um a forecast by itself doesn't create value but when we connect this prediction to make better decisions is when we really are bringing this value that you were mentioning. So you you saw this firsthand h at lowest. So um can you walk h us through the approach you use then and more specifically how did you move from this predicting demand to use those predictions to make better inventory decisions just to connect these two dots >> absolutely I'll give you a specific use case when we think about lows there are many products that makes up one key category that is uh highly highly valuable for the company, highly valuable for the consumer out there is what they call as the appliance sector. Your refrigerators, your washerd dryers, your dishwashers, what have you. No points for guessing. Lowe's remains the market leader in the US for the appliance category for a long time and obviously would love to remain in that space. But buried under the supply supply chain aspects of it, appliance is also one of the most highly unpredictable sourcing categories. It is unlike certain other categories where the supply is a plenty. This is [clears throat] the opposite of it. Especially the fact that some of these have to be sourced from far east. some of they have to be coming from an Asian you know manufacturing facilities if you will you know and you you're now looking at different ports they have to carry through the traversing through varieties of uh you know pieces have to come together the number of no longer a refrigerator is a refrigerator with a sort of kind of a um bespoke models coming through so all of that put together it's a constrained supply chain so when it becomes a constrained supply chain for us back in the time the problem was no longer I could rely on traditional inventory management and transportation to keep the product. For context, we have roughly about um we had roughly about 1,700 locations where the point of sale would happen plus the giant.com which means I could ship anything anywhere. Think about every zip code in the country and you only have certain amount of shipping points to get there. So physically the idea was to go multiply the last mile shipping points but the last mile shipping points cannot become an inventory holding point cuz then you're now you're you're only multiplying your constraints. You're not you're not deducting it. So then the science was applied to say look how do I predict what do I need when what categories within the appliance are needed which ones for example you think about a certain categories of refrigerators or what we call as uh distress look you know I I run an apartment I just broke my refrigerator broke I don't care which one I need I need something ASAP I'm not looking for the the $2,500 bespoke model that's a category which is separate from Arana household. Uh look, I don't buy a refrigerator every other year. I buy once every five, six years or 10 years, what have you. Uh and I want to make sure I pick the right one and you know keeps my wife happy or my wife thinks keep the guest happy. So that is a different category. So understanding and nuancing out who these are. How do I then apply reusing the word bespoke bespoke supply chain through each of these and now I need the science in this case a machine learning model to be performed through each of these and that's how we kind of look through >> that's an excellent example BJ thanks for sharing that. Um uh we have almost 900 live attendees. So [laughter] is really very interesting. So and we are receiving some excellent questions from the audience. Let me bring one from so habit. He's from Pakistan and his question is very interesting. He's saying classical models consistently beat more sophisticated models like recurrent neural networks for example in retail demand planning. at least in his experience. What are your thoughts on that? Because definitely I I I think he's highlighting a an important point. Sometimes more sophisticated models don't automatically mean better results. What What are your thoughts on that? >> Absolutely right. I think I think spot on, right? Um horses for courses. You don't you know just because we have a very sophisticated model that can ingest a whole bunch of variables uh you know and spit something out doesn't necessarily mean that it's going to be that much better than everything. It all depends on what is the situation we are trying to predict models for. I'll give you an example. If we are looking at an everyday uh you know supply that we are trying to you know stock up in a in a in a in a store or stock up in a warehouse a traditional model in fact you could even get away with the way back when minmax and nothing wrong about it now you're looking at constraint now you're looking at something that has even perishability that gets into it now you look at other variables that may have an apply and I'll give you one context to that back in the time in my Walmart, as many of you would know, Walmart still runs the largest grocery chain in the country. You walk in there, you would see strawberries and bananas. Every strawberry that you see there is not made alike and every banana you see is not made like in every store. What I mean by that is when they come from the sources before we had any level of intelligence we applied almost a decade ago. Now all bananas would generally be routed treated alike and go to every store. The the store that was closest to the the the supplying warehouse got the freshest pack and the store that was farthest out got a day or two older. Hence the perishability of that particular batch of bananas were faster. In other words, their shelf life was smaller. So now a simple logic that was put in place to say we will also factor that into which one gets routed all of a sudden change the narrative and there's more to it. So where I'm going with that is you're absolutely right. It comes down to what use case are we solving for then pick the right one. In the in in our world we use what we call as ensemble model. So not every model runs everything. The ensemble decides which one I need to use. >> Excellent. Excellent. Yes. Not every technology fits for all the problems. H so h let's move now to the sec to the second part VJ. So you have also frame very well with real examples that we really value that h the predictive the use of AI as predictive descriptive and predictive tool. Let's try to move now to generative AI and agentic AI. So from a supply chain practitioner's perspective, how would you explain the difference between these types of AI AI generative and agentic AI? >> Yeah. Yeah. It's it's very interesting. We are gradually increasing our vocabulary of what these AI are, right? So I [snorts] mean in the simplest way I've learned myself is when we think of predictive AI it tells it answers the question what will happen you know I look at all the content I want to know how much is going to happen when how fast can I get somewhere what will happen is the question predictive AIS do all the models and all the data that looks into it generative crosses another boundary and says Help me to create something new. Get me use all the context you have. The word generative here literally allows us to go create something net new that I don't have it before. Synthetic included. Agentic is simply saying that I get that I get what you're going to know when I get to know. I also know you need to create something. Let me do it for you. >> Right? I put it all pieces together. It is not one is more intelligent than the other. It's just the fundamentally predictive, generative and agentic in the order that you have increasing scope of autonomy that we decide to give. >> Yeah. The use case we one of the use cases we look in supply chain for us is when I look at how much I need to purchase to fulfill a category within within a within a store or a hospital or a DC wherever the need is predictive tools come into play. Now if I have to go use that to help negotiate with a supplier, negotiate with uh you know um how much do I you know uh you know uh v vary based on the contracts look at that and look at the contract and adjust for prices. Generative AI now helps me to create that draft that conversation that amount of uh uh articulation to get there. But if I'm saying all that is fine, I simply don't have the time to go look at thousand suppliers every morning and and and do all of this. Agentic, you take it from there. So it's a matter of autonomy that opens up. That's how we look in supply chain >> and and you have seen this transition uh VJ very clearly in your current role at Noman Health where you are using AI agents in in healthcare system and in the healthcare industry. Can you give us an example of how Agentic AI is being applied to procurement in particular erh for in that environment? >> Absolutely. Now first things first being in the healthcare I want to underscore the fact that we believe in having a human in the loop to get the right remarkable patient experience. There is no compromise on that. Which means when we look at where AI and the forms of AI like an agentic AI or agents come into play is exactly where it has the most relevance where it is closer and closer to a clinician or a patient. Closer and closer the human in the loop is involved. Given all that when we look at procurement one of the use cases for us was similar to many other large healthare systems here we are made up of thousands of contracts thousands of supplies and each of these supplies have a genesis of when they started why they started I'll give you a a very poor example but to hit the point think of um surgical sort of instruments or scalpels that we use for grins. We have about seven different uh orth orthopedic surgeons operating at our seven different the seven is just a a notional number different I mean by each of these surgeons through variety of choices they and practices they bring to the table they have a particular choice of uh an instrument they want to use and some of these instruments are made by three different suppliers. Now pretty quickly the number of same s the same procedure done by different surgeons choice of scalpels made by different providers uh suppliers I mean by multiplies and each of them may carry a contract or two depending on when they were written what was negotiated. So suddenly you see for one product you got multiple contracts before you know it and there is variance in each of these. Dr. Eva may have signed up for you know $50. I'm making a numbers appear scalpel. Dr. VJ may have done for 45 and maybe somebody else would have done it for 65. And given the breadth and the scale of things out there, it is not humanly feasible to be able to look at each of these and and and syn align and synchronize. Maybe maybe should we get everything a 45 or the quality is better for 55. All those things has to be looked at. We have deployed agents to help analyze those variances, help make a recommendation what makes the most sense for a particular category. that obviously ensures the human in the loop, the clinician in the loop ensures, okay, that's the right thing for us to use and then help draft a conversation with the the vendor and the supplier and say, hey, how could we help align closer to so we reduce the variance in terms of procurement I have to do? So that reduces the contract variances. So that's one of the one of the spaces where we look at agents. >> Yeah. And this raised also an important question about how much autonomy we autonomy we should actually give to this system because you also mentioned at the very beginning the human needs to be in the loop and and here is a very interesting question from singing from France. He said in procurement which decisions are most suitable for agentic AI you you just bring that example but which decisions are the ones that should remain under human control. >> Yeah. No, that's a great question right not just in procurement but I can also extend other parts of supply chain but in in general when a decision which is routine reversible and it's a bounded cost you are you are more likely than not at some point to help an auto automation or an agent to take over. But when something is whatever rhyme a reason uh irreversible or or something that is uh too too one of something that truly is more let's call it strategic to the aspects of the company aspects of what you do. You're branching into a brand new um you know venture of business within your organization or you want to try something out at large at scale. those things significantly require a human in the loop because the reversibility aspect to me is the most important. >> Yeah. And and I think this distinction [clears throat] between what AI can do and what AI should do is really important and this is very important point you are bringing here BJ there is another this is probably a more technical question but also very interesting from Reene from India. He said, "How would our agents when the data is good but unstructured?" >> Yeah. So, so here here's the beauty of data, right? We talk about AI a lot. We talk about agents, everything, all the glamour and the glitz that that aspect of the science brings to the table, which is very popular out there. But the reality is none of this is any good if the underlying data is of no use. That's the weakest pole in the link, right? Given that the question is all about if you had more unstructured data, how do you use it? Now, by nature, lot of the data that we get to see in our lives is unstructured. I'll give you one example. In the clinical world, I'll give you an example. I'll give you in a in a retail world. In the clinical world, many of us go get X-rays. Your X-ray by design is an unstructured data. It's a picture. Well, it's an image, right? It's radiology image. Having to read an unstructured data like an image, like a picture and deduce information from it through all the knowledge on it is is a highly complex piece to it. You are you as much as an AI would have in it, there is an element of human required to ensure that's validation. So, you got to you got to balance who you are. In the retail world, when you think of an unstructured data, you can think of lots of product cataloges. How many of us here can think of a retailer? You don't have to name it. Think of a retailer who you saw a picture on the on the on the catalog on the web and what you got at your home two different things, right? When you think about any which of these use cases, the idea here is what ultimately what use case are we having to have this AI to solve for and what is the underlying data? If the underlying data is unstructured, how do I put the decision loop in a way that it is meaningful versus doing for the sake of tech? Never [clears throat] do anything in my opinion, never do anything for the sake of tech. >> Yeah, I'm glad that Reing brought that question because also brought the how important data is, data quality and the type of late data. I think that's a great place to to leave this part of the conversation because we have gone from AI helping us to predict to recommend and now potentially to act and and and take action. But as these technologies take a one take more and more work I think the question and the the concern becomes is about what happened to the people doing that work and what skills become more important. So let's now talk about this evolution from supply chain analytics to agentic AI and how this is impacting the workforce and let's start hearing from our global audience because we still have almost 900 people live. So please share with us h launch poll three and let us know how do you expect AI to affect your own career during the next three years and while uh the audience is [clears throat] responding to this poll BJ as AI takes over more routine analytical and execution task how do you see supply chain jobs and the workforce changing? Yeah, I think see reality is as more and more AI and automation come about in everything we do and everything we see the quantitative grind shrinks meaning the need for somebody to run a dashboard every day. The need for somebody to make thousand phone calls to to do something what have you. The quantitative grind will shrink. When that quantitative grind shrinks the judgment gets elevated that will be a shift not an elimination in my opinion of functions shift I mean by the need for now the humans in the loop to say is what's coming out the right thing do I how do I ask better questions so the way I think about is that most important important skill that we all should prepare ourselves as we get ahead of us is realizing the the automation and the AI can do what they can do with the power of compute and all that fun in place there is going to be a part where the fundamentals aren't changing let's not let's not assume that is going away if we don't know to ask the right question you're not going to get the right answer no matter no matter what level of intelligence that's going to be out there. If you don't know to feed the right data, structured or unstructured, you're not going to get the right answer. >> To do that, you need a level of human intelligence to play in it. >> It's a shift that is required to help to get to that that part of the the stage. So uh my ask of rising folks who wants to participate in the AI in the supply chain what to do from here would be not everybody needs to be the the most strongest AI developers or coders. In fact many of us don't need to be it's a it's a there there are places for it. What is required though is the ground understanding of the business fundamentals, the ground understanding of what supply chain regardless of the function you're part of that is not changing. How do you use that and create in in the world of AI we use the word prompt, right? That is how do you prompt for the right question? How do you prompt yourself to get to validate the answer? Just because you get an automation answer doesn't mean it's always right. Right? How do you know to ask the right question? How do you know to see what things are to be able to get to that? That is the skill that that's the muscle and that's the skill we have to develop. >> Yeah, >> I see a shift not an elimination >> definitely and I think this is 100% aligned with the responses. Let's try to share the results of poll three. But most of our attendees, they see this 42% more as an opportunity than a threat and 33% as a significant opportunity which is great. You also brought BJ the important of business accumment and this understanding of of the business that need to be there for sure as well as the quality of the data and the type of data. Um yeah one of our our attendees from Vietnam he said okay if how do you see companies now hiring junior analysts to learn on the job? um is the expectation of new graduates to already work alongside AI agents from day one. I think you already answered the question because definitely the prompting and bringing the right questions to the to the tools h is part of that but I don't know if you want to elaborate more on that interesting question. >> Yeah, I mean definitely an interesting question and and this is the turn in the transformation era we're all witnessing in front of us, right? I mean quite quite candidly a certain population gets to become a bit more vulnerable than the others particularly the newer entrance to the business because they're new they're graduating from wherever they are they they're just walked in walked into the workforce and the most entry-level jobs sometimes become the most obvious candidates for automation AI to sort of take over. So you know it's it's almost like a like a like a a mishmash. So my ask of that situation or the entrance in this case is do not skip the fundamentals. Do not do don't assume everything is only going to be AI and AI is only going to be as good as how you're going to know how to use it. So there is always going to be a role. It's just a shift in the narrative. be prepared, understand enough to know how to how to ask the right questions, how to tame it. You don't have to be an expert in building dashboards. You obviously need to know what how to ask the right questions, right? So, I think it's it's preparing yourself than than than uh relying on what what was working 20 years ago. >> I couldn't agree more on on the need of understanding these fundamental techniques and the foundations in order to build on that and be augmented by by these tools. Now let's move to the last poll because we want to learn from our global audience about specifically that which skills do you believe are most critical for future supply chain professionals. Um BJ you already mentioned the importance of foundational supply chain skills analytics statistic optimization supply chain fundamentals. um what should also companies be doing in order to prepare their workforce for this transition? >> Yeah, I mean here is the core argument for me. You can't judge what you don't understand. Meaning the population that drives the companies in any which level has to ensure that they are equipped with the knowledge of what is going to be success for that function to be which includes understanding the core aspects of the function, understanding the the principles where the companies want to go, all of the above. only when when you have either a physical or or or a virtual automation. You're not talking about just AI coming in the form of an LLM or a chat GPD with you. It is also the robotics and the and the physical sensors that they're going to be hitting you all day every day. Regardless of that, there is an element of how do you know if that automation is doing the right thing? Only if you know the function, you'll be able to ask the question. So the companies have to prepare their audience grounding themselves again and again in the solidarity of what they are going after with the basics and and give them the enrichment for lack of a better word with AI and automation along the way. AI and automation cannot be a substitute or or a or a or an alternative for not having a grounded fundamentals. >> Yeah. Yeah. and and also to to just challenge this output to understand the output and being able to challenge that output. Let's try to see now the results of this poll because our audience 75% of them vote for critical thinking that is totally aligned with what you are bringing here. H 55% analytical skills. Yes, these fundamental and and and basic tools that we need to understand in order to to understand the solution and the technological and AI literacy 57% uh vote for that one with that uh VJ what advice would you give to professionals who are building and advance and also advancing their capabilities in supply chain management? I mean the first and the foremost I would tell you is that the business of how we or the art of business the conducting the business is evolving as well. The business itself is not static. What was otherwise let's call it 20 25 years ago a linear way of doing things. you plan, you then make and then your ship all that is now largely with the automation AI in place is is becoming more secular in nature. I'm going to be planning lot more factors based on how that's going to get shipped, how that's going to uh you know get potentially returned and every function is looking much more circular in nature. That transformation and how the business is also being conducted is an extreme element that we as professionals should adapt to. AI, automation and whatever other fun things that are going to be coming ahead of us that we do not do not know, do not predict will always be there. Technology is only there to advance. But the most important thing is if we get a grasp of how the business is shifting, what questions do we ask? it is less important what answers I'm going to give because there are tools and automation out there that can give a good answer but if I do not know how to ask better questions the the the answers are not going to be no matter what AI is out there the answer is going to be raj irrelevant right so first my only advice from myself and anybody in this audience is stay grounded with what you can learn in terms of the business stay grounded with how effective of a questions we can ask that improves the response that comes from wherever it comes from. >> Yes, become a lifelong learner. [laughter] Thank you BJ for such an insightful and engaging discussion. We really appreciate you sharing your experience and giving us a practical perspective on how AI is changing supply chain decision making and also how this is impacting the workforce. And thank you to our audience for all the great questions and for being part of this conversation as you reflect on the topics we covered today from supply chain analytics to generative and agentic AI and also the changing skills required in the field. If you are thinking about continuing your learning journey, our supply chain analytics and supply chain fundamentals courses part of the MITx micro masters program in supply chain management are now open for enrollment. You have a we have also shared more details in the chat and also invite you to explore to continue investing in your learning journey that we have been discussing today is really critical and important. I would also like to thank my team Elis Jonica and Caragrini for supporting this event behind the scenes. And finally, we would love to hear from you. We are launching a final poll now to get your feedback on today's webinar. Your inputs help us to continue improving these events and provide a valuable learning opportunities to our community. So, thank you again to our global audience for joining us today and keep an eye uh for future future webinars. Let's continue to grow and learn together in this exciting field of supply chain management. Thank you BJ and thank you everyone.