DSCInsights in Action: Integrating AI into Your Supply Chain

Welcome to Digital Supply Chain Institute Insights in Action. We keep the actions going and as always promised, we keep the focus on the things which are applicable, meaningful, and also which can change the way of how supply chains function. And I'm very happy to have with me today Marisa Brown, who is the senior principal research lead for supply chain management at APQC as our long-term partners. Welcome, Marisa. >> Thank you, Marko. >> And Vivek Gelani, as we call him a usual suspect, Digital Supply Chain Institute director of research. Welcome, Vivek. >> Thank you, Marko. >> Let's deep dive into the things which, you know, we we have been cooking for last uh few months together. And especially towards what we can share with our audience about practicalities, about the deep dive research we did in the in the AI segment. I know today, you know, there was also a great webinar about it and I think this DSCI insights can actually help a lot in a sense of aggregating the key messages we would like to share. And I would like to start first from based on your based on the research we have done together, you know, what common partners or patterns, sorry, or gaps do you see among the organizations, right? That that struggle behind moving towards the pilots. Because we see a lot of them thinking about it, testing, trying. Mostly they are about pilot phase. Not that many go into production, which is meaningful output for, you know, internal companies and for the customers. And then any data points that you would like or you can share about it. So, we can start with Marisa first and then Vivek you can you can jump in. >> Sure. So, I think when it comes to moving beyond pilots for AI and somebody once referred to this as pilot purgatory, companies that are stuck endlessly piloting and never get through to full-scale execution. I think what we see across the research is three patterns that come up over and over again. And the first is there's no clear link to business value. And so pilots can be interesting from a technical perspective, but if they're not tied to a metric that leadership is actively managing, whether it's cost or service level, cycle time, something like that, if it's not tied to a business outcome, then it's really difficult to get scaling and justification for scaling. And then the second big thing we see is fragmentation in the underlying data and process management. Organizations often underestimate how much standardization is really needed for AI to be successful. Because otherwise you get AI working in a pilot, in a bubble if you will, but it won't scale across regions or functions because the processes aren't consistent or the data's not consistent. And then the third thing I would say is a lack of ownership and governance. There if there's no clear decision-making structure and authority for who decides what moves forward, things often stall. So I think that's what we're seeing as far as the biggest barriers aren't necessarily technical. They're organizational barriers. It's lack of collaboration, it's lack of governance and what ends up happening is very little enterprise impact. >> Yeah, I I completely agree with Marisa and I I would add to that. I think trust in the technology as well has a major impact in terms of adoption. Uh we have seen through one of our members as well um that they had successful pilot, uh, it's going to scale, but the end user is not trusting the results that's coming out of it, and it's the adoption rate just dropped to significantly low, then it's just a waste of time and resources for them to even actually use that. And then they had to scratch that pilot and the uh, and the end implementation after that. So, I think trusting the trust the technology, right? The decision-making that's coming out of it. Um, again, we're not saying blindly trust it. Right? You have to you uh, the team will be there to go through the process of understanding what's happening behind the curtains, how the decision was made, right? What are the what were the stakeholders and everything. But once everything is up over the level and and and there needs to be a trust there needs to be a trust, otherwise the um, the technology can't fix that issues for you, and then it won't um, result anything at the end of it. >> You know, that you remind me, Vivek, we can't forget that implementing AI is a change, and all the historical lessons learned about change management is still important, right? So, engaging people in the change, having them be part of it versus having AI something that happens to them is so important. >> Yeah. >> Thank you. Thank you, Marissa. Thank you, Vivek. I think, you know, like the way you how you structured, let's call it three plus one approach in a sense of the segments that companies can focus on improving the business case, you know, finding the way to scale, govern and understand, and then, you know, build internal consensus, and then plus one, have the trust not only in technology, but in data. I love, Vivek, how you you cross-share that. You know, presents a complex system, but everything what is complex is actually, at the end of the day, resolved in a simple manners. And I think simple manners we can move towards are what performance metrics, which are usually the simple things to evaluate and and tackle the problems we spoke about. Do you recommend companies track in early stage of AI implementation? Because actually those metrics will prove the case in you know, one, two, three plus one. >> Um yeah, I can I can I can answer that Marissa if you're okay with that. I think in the early stages, I think tracking in any of the pilot needs to be there because other than you can't evaluate is it working or not. But in terms of tracking the KPIs, I I think companies should track adoption and impact both of it in the early stages of pilot because that's also helps you when you're you're going to scale. You can definitely track if the model is working technology is working. But also tracking the human aspect of it, right? You can connect it to business outcome at the end of the day what value was created depends on different pilots that you're creating. But tracking the impact of it right in in in terms of user impact, efficiency gain or um um or removing you know, the decision points or the nodes in the process that you're you're implementing AI. That is a very important. But also adoption, how user is using it, is it actually helping them in their day-to-day life, is it solving the problem that they are facing? That is more important nowadays because that's the best way to have user to use it. Now, cost saving, cycle time, those KPIs are always true. Businesses need them, you need to connect to them because unless you connect to them, you can't even scale because then you don't get a buying from a sales sales with leader or leadership on and to actually scaling them. So those are important, but in terms of getting confidence of your team, that is more important as well. So I can tracking that is important. The third thing I would say don't buy in the KPIs that vendors are selling you ever. Like vendors will sell you their own KPIs on the vendor data that they are trying to demo you on it. So never buy that. Let them work on your own data, your data like real world data. Let them track the KPIs that is um important to you and then see if that makes sense to you. So yeah, I would I would suggest just work on those KPIs first. >> Thank you, Vivek. Marisa, please. >> Yeah, I agree. I agree. And I think one of the biggest mistakes organizations can make is tracking technical performance only, like the model accuracy, without tying it back to business outcomes, like Vivek said. And if you can't draw that line back to the business outcomes, then again, it's difficult to scale. >> So in a nutshell, at the end of the day, everything comes towards defining your own metrics in a right way, moving away from the vendor's script, and then as as you said, Marisa, you know, building the consensus around it and understanding, you know, how it impacts your business directly. And that brings me to the next question, which I think is very important. How do you see the organizational structure changing or evolving, and what should supply chain leaders do? Because there is a great mystery about, you know, AI or people, and we don't see that in the reality, right? So in which direction you see it's evolving, and you know, what supply chain leaders should do through your research? >> So I think we're seeing, I would say, less about completely new structures and more about how responsibilities are evolving. And I I can think of two major shifts we're seeing. The first is cross-functional ownership. Because AI doesn't sit neatly in any one part of the organization. When you think about supply chain, it cuts across planning, procurement, logistics, IT. So, we're seeing organizations creating these shared ownership models, and a lot of them are using things like a center of excellence or similar type structure to provide the governance and standards that are needed. Um and then a second shift is more about shifting our roles from execution to decision-making because we're seeing that AI can handle a lot of the we'll call it kind of ordinary work. And then what people are starting to focus on um is exception management, scenario planning, strategic decisions, things that are more um higher-level thinking, if you will. And companies are starting to focus on skills development for folks in those areas with AI fluency and data-driven decision-making. And I think it's both a bit of a work shift a workforce and an operational model shift to making sure that you're treating AI as if it's a team member and having it be collaborative um and still ensuring for this kind of cross-functional governance and center of excellence kind of a model. >> And then >> I completely agree with what you're saying. One thing about governance is that that is also important when you're doing pilot. Um in the guide that we shared, I think one of the interview also shared an example around the any pilot they do, they also go through the governance exercise first before the pilot takes actually starting. Because once you don't do the governance, even if the pilot is successful, they may not see the end end goal of scaling because the governance was not set up properly. So, I think that is more important. In terms of roles and responsibility, I agree we we just did a research internally for one of our members and and the few of the learnings were very very highlighting in terms of the middle layer for so many of the team members were just vanishing away just because the manager now can handle so many people. So, on an average they used to manage like 25 per people, now they can manage 40 people because of the AI enablement enablement that they are getting. Right? So, the middle layer is going away. The roles and responsibility we try to map like the demand planner and and the new demand planner we try to do demand forecasting and demand new demand forecaster. The roles and responsibility are are changing. They are not just sitting into as Marisa said into one function. They're going to work cross-functionally across other function. And there there are also new roles that are coming up like agent AI agent AI for agent AI especially we have seen agents ops coming up, the new governance structure ladies coming up for agent AI. Uh so, those are type of new roles people will need to change to as well. Um so, there's definitely an shift also happening um in terms of okay, your role will change from this to that. But it how is going to impact in terms of your deliverables? That that is also changing. So, we mapped that out for one of our members and that's has been a great exercise in terms of understanding. I know people always fear losing a job to AI. But as long as um I think you can map it out what you could do with the help of AI, that also help the team members as well. So, I think more and more um people need to also think about that. Like how you you think your job can change if you use the AI properly in a right way. Uh that can, you know, um elevate your responsibility but elevate your decision-making in the company, your role in the company. So, that's what we are seeing with one of our members that we did it did research with. It needs to be um invest more time in what AI can do for you instead of just worrying about AI will take your job. >> Mhm. Thank you very much and I like the notion where you explained actually it's not the change of the structure, it's the change of the roles which are going from more vertically if I can put it that way into more horizontal and more focus on cross collaboration which was always one of the challenges and hearing both of you what I see evolving here is you know also a capability and capacity of the orchestration in between the human and energetic and I think that might be a topic for us to explore in one of our next conversations because that might be a a unique skill set which will be needed across the board where actually people and technology will not compete but cross collaborate in a right way. And this gives me a great way to try to round up this conversation where I will try to squeeze two questions in one, right? But they are they are closely correlated and they go towards like everybody's talking about it. We spoke what are the criteria what are the elements, what's happening with the organizational structure? But for those who would like to go forward, who would like to select partners or vendors along the way, what criteria should companies use from your view and learnings, you know, to evaluate tools and and vendors, you know, especially through AI comparing to, you know, just a simple automation because people are trying to say well sell all as AI today. And then based on your learnings, what are the mistakes you see organizations content constantly making when they are selecting because these two will help people understand especially those who didn't start and even those who started and made a mistake not to be, let's say, frightened not to go again. >> Yeah. >> So what I I seen in terms of which which I like from the interview that that we did who is actually a coach of DSA as well. His His one point I think really made made an impression about when he evaluates as a CIO any technology vendor not just AI like any technology vendor across the board. His one thing is any vendor I work with needs to be easily replaceable, right? If I'm if I'm putting money to implement a technology and in 6 months because technology is evolving so rapidly, there is a new technology that I could use next time. He wants to just get rid of a technology like really like a ripping a band-aid off. That's what his word like really ripping a band-aid off right away so I can just put a new band-aid on and just go and and do my do my work. So I think that that's one of the things I his criteria. And I think it's great because if it takes you years to implement a technology and years dismantle it, then you won't be able to catch up with the with the way technology is accelerating, right? So that's I I think companies should think about when they are um implementing technology. In other comments I said before I think make sure they're not selling and you're not buying into that promise of wonderland and heaven, right? You just need to think about what they can do and work on your actual data. Um actual KPIs that you are important to you and and then then evaluate them and not just go blindly trusting anyone about that. Yeah. >> Yeah, that that's exactly what we're seeing in the research. Is that the biggest mistake is evaluating a vendor based on the technology versus the outcome. So companies feature focus on features and the demo and the technical sophistication, but often don't validate whether that solution's actually going to work in your environment. So the tool looks impressive, but it doesn't deliver measurable impact. So I think that requiring your vendor to use your own data, your processes, your specific use cases so that to use Rebecca's words, words, they're not selling you heaven and wonderland so that you can actually see it in your environment. That then will be a strong indicator of whether you can achieve the results you're looking for. So shifting from what can the tool do to what results can this tool deliver in our environment? I think is the important shift. >> We can cover both in saying like don't buy heaven and wonderland, but actually define beforehand how your heaven and wonderland looks like from resolving your internal issues and then give that task to vendors to play with with your criteria and your data and if they can deliver, you might be having a scalable solution. If not, you you just go further and explore. So thank you very much for a great conversation. We'll definitely build up on this because there are multiple elements which will be coming as an addition, I'm sure. I want to to give a big thank you as well to APQC as digital supply chain institute partner for long time in the area of really hands on research in supply chain growth, development, and digital transformation. And we look forward Marissa working more with you and your team and aligning new things and and valuable insights that we can share with everyone. So thank you both very much. >> Thank you. >> Thank you Marco. >> This was another episode of digital supply chain institute insights in action and we'll come with more. >> [music] >> We