Fueling AI-Driven Intelligence and Enterprise Systems with Logistics and Commercial Data

Henry Ames: For this session, this afternoon's session, we have some wonderful participants. We have Guy Courtin back from yesterday's session, and Guy is the vice president of Industry and Alliances for Tecsys.

We have a new participant today, Henrik Killander, senior vice president, product manager, integrated supply chain from A.P. Moller Maersk. And we have Evren Ozkaya, CEO and founder of Supply Chain Wizard AI.

Let me invite all three of them here to the panel session for this afternoon. Just to get us started, I thought it might be fun to give a little fun fact about each individual.

Evren is an enthusiastic traveler and has spent his time going across over 66 countries around the globe, has plans for quite a few more.

Henrik loves music and has built fan sites to attract his favorite band so he can interview them. That's a way of actually getting to meet the people that you admire. He's toured Sweden with his own band from top to bottom. And Guy Courtin his great great grandfather was a Belgian artist, and his father's face is on the Belgian 500 franc note.

Guy Courtin: Right. Great great great grandfather.

Henry: Three greats? OK.

Guy: Three greats.

Henry: We've got to go back three generations. We've got a great group here. Today's discussion is going to be on AI-driven intelligence and enterprise systems with logistics and commerce data.

Before we begin, if we could, now that we know some fun facts, I could ask each of you to give us a little bit more about your background, a little bit about the organization which you represent, and then we'll dig right into the questions. Then, of course, we'll turn it over to the audience for some Q&A. So, Evren, off to you.

Evren Ozkaya: Well, if you were downstairs in George's story of, like, Chapter One, Two, Three of his life, I can probably say the same thing.

Chapter One, I was a consultant at McKinsey & Company. I've served seven different industries, not pharma. Then I got headhunted into pharmaceuticals. I started working at Sandoz, Novartis as supply chain strategy director, and I was the head of Track & Trace program globally, so that's like Chapter Two.

Chapter Three, I started Supply Chain Wizard and SCWAI as a startup organization, and it's been 10 years at it. Right now, I'm looking forward to the next 10 years with the AI and intelligence and connectivity. So, good to be here.

Henrik Killander: I'm Henrik and I'm with Maersk for the Integrated Supply Chain Engine organization. I have a background, 10 plus years in supply chain where I've led corporate ventures around digital freight forwarding and initiatives around supply chain visibility and recommendations and which is what I'm working today with my team in Maersk.

Guy: I'm Guy. I feel old now because I've had 25 plus years in supply chain. Started out at Forrester Research late 90s. I have a couple of Jeff Bezos stories if you ever want me to tell you. He used to come in and brief us and it was this cute little company out of Seattle selling books and CDs, but that's what really got me started in supply chain.

Tecsys, where I work today, we are a supply chain execution software company based out of Montreal focused on health care, distribution, and retail.

Henry: Wonderful. Thank you all. During our preparation, we talked about the complexities of the pharmaceutical supply chain, the lack of digital integration that exists today, and the opportunities for driving process improvement and cost savings. I could open that up to any one of you to take that. Maybe you could play off of each other a little bit and elaborate.

Evren: Maybe I can tell you a little bit of story of how I came to pharma and what I found. After serving seven different industries and McKinsey as a consultant, I thought that I know about supply chain and all the intricacies. I came to pharma, and my first job was five weeks from now.

Day one, I'm on the job. My head of supply chain, my boss tells me, "Evren, your first task is organize the very first S&OP meeting, and it will be for the president and the direct reports." I said, "OK. Let me see the previous S&OP meeting decks. Let me see your database. Let me see the data, you know, sources."

He said, we stopped doing S&OP. There's no deck. There's no database. Here are the bunch of people you should talk to. Then once I realized talking to those people and literally our factories are all totally in different systems, most of them are on paper, the lack of information about our CMOs, which we had, like, 30 plus of them for North America at the time.

It was a $3 billion business running on pen and paper and Excel, and that was the most difficult five weeks of my life. I thought I know supply chain. I thought, like, after that experience, I said, OK, well, this is an industry to stay. There's so much opportunity and I end up staying for almost 15 years now.

Henry: Thank you. Henrik?

Henrik: I think we've had an amazing journey in Maersk with both failures and successes. I've had to enjoy a couple of them. Some of you might know TradeLens which I was also involved in evaluating that business where we try to, through blockchain and AI technology bring together the industry, to create this global interconnected logistics platform.

I'll share today some of the learnings that we've made from that experience and what we're building today and why we're more confident on the approach of integrating global logistics and our vision of creating visibility and AI power recommendations for our customers to bring more control to our customers.

Guy: I think one of the things we need to think about when it comes to pharmacy is really sort of a bigger issue around health care and the exchange of information and the mentality of health care when it comes to supply chain.

Until recently, most health care associations, pharmacies that we've talked to, supply chain is sort of a back office function. It's, oh, we need it but it's those guys and overalls will do it. And we're focused on the patient which is great. Like, yes, focus on the patient.

I think we're starting to see this change where all of a sudden the supply chain is being recognized as this is instrumental to serving the patient and then with pharmacy, the same.

But then what's happening is that we're relying on old systems you mentioned pen and paper, etc., that we're only starting to now figure out, wait a minute, If we want to be a modern business, taking care of the patient with pharmacy and health care, we have to have modern systems behind it.

That's the big transition that we're certainly seeing whether it's in health care, pharmacy in particular. I think it's that mentality that's starting to change. I think it's good, but there's a lot of work ahead of us when it comes to that. Exchanging information, having the right platform, is one of the fundamental parts of that as well.

Henry: Thank you, Guy. I would like to drill down a little bit further on that. In one of our prep calls over the last couple of weeks, we talked about changing market dynamics in pharma. We talked about this focus on the core competencies of these organizations.

The need to expand drug pipelines, the outsourcing activities that are going on, allowing real specialists like the Kuehne Nagels of the world to do the work that they do best and allow the drug manufacturers to really do the work that they do best.

We've talked about this drive towards releasing shareholder value and we've seen it in the market whether it's Baxter and Baxalta, Maersk, Organon. You've seen Johnson and Johnson and Kenvue, you've seen, these divestitures and these acquisitions.
How does that play into the complexity of AI and the use of data and the collaboration across these massive entities that are constantly changing?

Guy: I'll start with that. Part of it is exactly what we've been hearing last two days here. Which is it's all about the data, but it's also about the cleanliness of the data, the data hygiene, the data governance.

Now you're starting to look outside your four walls as well. You're relying upon whether it's Kuehne Nagel or other 3PLs, whether it's manufacturer. How are you communicating with those folks, if you will, in your network to make sure that you're all striving for the end goal?

Oh, and by the way, there's a consumer or the patient in this case at the end of that chain. How are you communicating or how you gathering information from them as well? I think that's the challenge. Is that we're opening this up to a network where you're going to have different ways of sharing information. You're going to have different philosophies.

You're going to have different needs and wants. And the question becomes the challenge becomes for the industry is, all right, how do we optimize on that? How do we build an ability to share information and decision making that we can all agree upon and do so in a way that again, we're protecting the patient.

I think the difference is -- I work a lot in retail -- the difference is, I'm going to pick on Amazon. If I don't get you your T shirt on time, OK, it's not the worst thing in the world. You might get a little pissed, but you know what? So what? If I don't get you the right medication on time, that can affect your life.

Or if I get you the wrong medication, that can affect your life. That's why I think this is much more of a challenge is because if that information is wrong or if I don't have the right information at the right time, it can affect your life. That T shirt? OK.

Evren: Maybe I can add the manufacturing side of the story a little bit. During the prep talk, I think you mentioned it. J&J, you heard in one of the discussions you had. They have, like, 56 different types of ERP systems in J&J.

I actually said, oh, wait a second. 13, 14 years ago, one of my mentors from McKinsey took a really senior job at J&J. When I was talking to him, he told me they had 77 different instances. They are making progress going from 77 to 56. Probably divesting can be along the way.

One of the largest companies in manufacturing, they have all these different ERP systems, which is the main system how you run supply chain. Imagine when you go to the shop floor, how many different legacy systems that you have, how many different sites you have.

The problem is with the, you know, we're trying to tie everything back to AI. We need to have that foundation first of data and data quality. When you have all these fragmented systems, it's a very difficult thing to achieve, and it's very expensive.

If it will be, like, super easy to unify to SAP S/4HANA in one instance of ERP, they would have done it by now. They're still going through that massive process. It's a real challenge to address. It's not an easy thing. We can talk all we want about AI, but we cannot talk about generalized AI at the company level until we start addressing those layers of data and systems.

Henrik: Very well spoken there, and I want to touch on what you also mentioned about the end consumer, the end patient, or and also maybe help Amazon here a bit in jobs. I think something that really inspired me when he says is that, he was asked, what will change in the next 10 years for you? And he basically answered, well, it's easier to answer what will not change.

We'll still need to deliver goods faster. Customers will have lower price, etc. I think, although we have to build these fundamental capabilities in the stack being focused on those things that won't change can help us navigate an evolving market situation.

Henry: Thank you, Henrik.

Evren: Maybe one other comment. We are talking about the pessimist scenarios, all these fragmentation issues, but what is the optimist scenario of, like, how can we actually make AI work? One big mistake that I see, especially with large technology firms and the large manufacturing companies and pharma coming together, and they have this great idea to build a data lake.

They create this hundred million worth of project. Let's create a data lake. Let's put all of our data from all these systems in, and then five years later, somebody in the Power BI department will build some dashboard for us, and it will, you know, that doesn't work

If you're thinking about a big database lake project, that won't work. But what you can do to make AI work is pick a very specific use case. Come from the business aspect first. Identify your objectives. Make the decisions on what you need to change in the business, what visualizations you need coming backwards.

Then, which data would feed those dashboards and decision making processes, only collect those data points and those systems, then you have a layer that can add up all the way to an AI and automation and intelligence, which is very small slice of the overall supply chain rather than trying to, what we used to call in McKinsey, boil the ocean.

We try to do everything all at once. It's very different approach to do it very use case driven and value driven, and then AI can be starting to put some value into the business.

Henry: Could we dig into that a bit more? When we talk about the supply chain, of course, everything's interconnected. So where would you start? Where or is it dependent upon the organization? What have you seen in your own experiences?

Evren: I can tell you from our experience, like, maybe I can tie it to Track & Trace. When we were in the early days of Track & Trace, we were implementing these packaging line upgrades, and one client, basically, after putting three, four machines on the packaging line, their entire efficiency start going down.

We actually measured it before and after. It was about 30 percent loss in throughput in one packaging line. The client was obviously concerned. They didn't want to roll this out to 12 different lines, and the whole factory will go down by 30 percent. So we start measuring. How do you measure the efficiency on the line?

Well, we have the sheets at the end of the shift we fill out with pen and paper, and that was like, OK, a clear some specific use case that we then dig deeper and said, OK, how do we collect the right data, put some sensors, put some user interface, operators are interacting.

Then we start building a small database on why the line stops, why it doesn't stop, how the productivity is, and then start getting insights, which led to an insightful conversation and decision making on how to improve the performance of the line and bring that 30 percent loss back to zero and possibly into positive.

That was one slice that I'm talking about. The framework that we used for this with the client, the change story we used for this was 3D transformation, data, dashboard, decision.

You need to collect the data on the specific use case, turn it into right dashboard, understand the insights, and then make the right decisions, improve the business, get the results, then feed it back again, and then you keep fueling that and then build the use case from there. But you need to pick one area to improve, and data can be the key in that area.

That's one example from manufacturing. I'm sure there's more in logistics and retail side of the things as well.

Henrik: Oh, sure. I think in our view of the world, our cold chain business is, of course, something where we really value the business and trust that our pharma and the cold chain customers give us.

We have these Hypercare control towers for large pharmaceutical companies where it's, let's start with the basics, ensure product quality and safety and we do that through the condition monitoring and generating alerts at a timely way, but also adding control.

It's this future where we are assisting our teams, our customer experience staff to take action on behalf of our customers to mitigate, for example, a deviation in the temperature or in the humidity or condition that could compromise the product.

We're doing more advanced of that in in some of the cold chain examples like bananas, actually triggering ripening throughout the process based on AI. These are some of the use cases, customer backwards that we are exploring among others, which we'll talk more about.

Guy: Yeah. We always talk to our customers when you think about AI and I think you mentioned it. It's narrow AI. It's really being able to focus on what dataset can you leverage AI for.

For example, if you think about the warehouse, one of the things that we can look at are things like route planning in the warehouse. So how do you pick in the warehouse? That's a narrow focus. Why? Because I have a finite set of data. I know historically how my pickers go pick. Can I leverage that and start applying some intelligence to that data to figure out other better route optimizations within the warehouse?

I would always tell anybody when they're looking at AI for anything, whether it's pharmacy, healthcare, retail, whatever, is think narrow. Think narrow AI first. Now there's a lot of hype around other AI beyond that, but to get the value today, I think it's really that narrow focus.

Henry: Thank you, Guy. One of the topics we thought we wanted to cover today was AI and its role in improving workflow, operational efficiencies. I'd like to dig into that just a little bit. We talked about leveraging AI and smart system design for creating efficient workflows within an organization, redefining those workflows, and improving the overall change management process.

Could I ask each of you to sort of weigh in on the core issues associated with this?

Henrik: Yeah, I can give it a go here. At Maersk, we have our goal of, of course, integrating and simplifying global logistics for our customers, but we also had our fair amount of failures in one-to-one migrations or feature parity when migrating old systems to new systems and benefiting from AI and data and platforms.

You really need to understand deeply the customer's needs, the objectives, you both spoke to this, and the value chain steps. To use this, perhaps silly but I found it very good analogy to dumb it down for me, it's like pre-washed salad. What's the job to be done to get that done? Well I want to serve a healthy salad with my dinner.

Then you understand the value chain steps that go into that, whether it's selecting different salads, it's cutting it, washing it, drying it, plating it and so on. I know it's a silly example, but that's like taking multiple value chain steps, which is the promise of AI. That's something we're thinking, we're applying.

Deeply understand the workflows, the slice that we mentioned, and understanding the job to be done and the success criteria for that and applying that to doing an import customs clearance.

How we automatically label commodity codes and speed up the import process or how we proactively or predictably identify the merge and detention issues to help resolve and close loop action on moving out goods that are incurring costs that might land in your margins, so to speak, a landed cost. Slices, but also with a view to the larger job to be done and the components of that value chain is key.

Guy: Yeah. I mentioned it in the last question. It's narrow, the narrow focus. Building it to your point too, Henrik. Like, when you think about the warehouse logistics, like DHL does things where they're leveraging AI for last mile routing. Why? Because they have a finite amount of data.

They know that, like, here in Barcelona, the streets are pretty much the same. Yeah, there's going to be things that change, but overall, you can start optimizing based on the data you have and with the end goal in mind.

The question becomes back to your point, Henry, is now the change management side. One of the things that we've discovered too is if you think about, let's take your warehouse again, and I start telling you how to re optimize your pick route or even in transportation, I tell you how to re optimize your last mile.

Well, your drivers might know things that the AI doesn't know. Your driver knows that, hey, on Tuesday I can park in this spot because school's not in session so no one's going to be there. But on Wednesday, the kids get out that time so I can't park at that spot. But the AI engine might say, well, no, no, no. You got to optimize the route this way.

Part of it too, from the change management is at the end of the day, this is just a tool. The tool is only as good as the people that use it and the people need to buy into using it before you really get the value out of it.

I think that's part of the change management when you start using these AI tools on a narrow focus. Because also to make sure your labor, the people that are going to apply it, are going to apply it properly. That's the big change management aspect that often is overlooked.

Evren: You mentioned the workflow. I'll give you maybe, I'm sure some of you are from pharmaceutical manufacturers, some of you are not, but maybe if you go into a pharmaceutical manufacturing plant, the workflow is actually very simple. It's sequential.

It starts from, let's say, the pills that we swallow. They are called oral solid dose, medicines. Start with dispensing. It goes into granulation, goes from granulated, product goes into compression or encapsulation, then maybe some coating, and then packaging. Five, six, seven departments in a sequence.

If you ask a contract manufacturer what's your lead time, beginning till the end, they quote to their customers a standard 90 days. If you order now, Pfizer orders to a contract manufacturing company, say, whichever, x, and then they say 90 days, and then they deliver in 110 days.

So what's going on in that value chain? If you actually really look into every single step, every single step takes hours or days. The total value added time in that process is maybe 10, 15 days. Why are we sending stuff out in 110 days? Because there is no system that ties the whole thing together. Everyone looks into their department.

Let's say I'm a compression manager. I don't even know what's coming from the granulation. I just focus on the work orders in my shop. Then I do it fine. I do setup. I do run. I do cleanup. I did enough job and I'm happy and everybody's like, yeah, good job in compression.

Look at the whole system, you're delivering in a 110 days for a process that could actually be done in 15. That narrow slice, if you put that slice through end-to-end, not vertical, because everybody looks vertical, heads down, nobody talks horizontal, if you connect those dots horizontally, you would unlock so much opportunities.

It's not even AI. Just, like, visibility will unlock potential. Then you can bring the AI to make it look how can we improve that automatically.

For example, if granulation is not done yet for another two days, why am I standing up my labor force in compression waiting for the product? The other waste that is very common is they do the job, granulation, they park the product in front of the doors of the compression, but then somebody forgets that it's over there.

They don't process it for 30 days, and then you have a deviation. You need to retest or dump the entire batch, and it's like million dollar loss on maybe product just because there's no visibility. There's no data. There's no connectivity.

I mean, so basic things like this in terms of shop floor visibility. Manufacturing is more like a mini supply chain. You extend it to a broader supply chain, like, this was in a single company, single manufacturing sites control. Extended it to hundreds or thousands of parties trying to work together, the difficulty is exponentially more.

But then the challenges could be resolved relatively easily if you focus horizontally and tie those things together, and then add the AI later to make this even smarter.

Henry: It's interesting the way you describe that. I think about the visibility being a big part of that, but I also think of people that were here yesterday know that Guy and I both went to Babson for MBAs.

We had a professor there named professor Kassarjian and he would often say be careful what you measure because you will drive certain behaviors. In those examples, I imagine those individuals that were driving those activities to drive efficiencies in that one area were not being measured across the organization itself.

Evren: Yeah. If you just measure in one function, then you get a total totally different result. Like, Amazon, I actually keep telling people about Amazon. If you ask the Amazon head of warehousing what's the best way to run Amazon warehousing, he would probably put one gigantic warehouse in Seattle, and that will be the most optimal, cost efficient way of running Amazon.

But today, they have multiple warehouses in every state. Why? Because end-to-end supply chain, it's better to deliver products in a few hours to customer. It's better for the system. Then, warehousing guy is like, OK. Well, I guess, I need to manage hundreds of warehouses now and tons of robotics and people.

Your cost of warehousing is significantly more, but your entire value chain delivers significantly more value to the end customer. It's not about the individual functions and what they do optimally in that function. It's a systems thinking, and the problem in supply chain is that we have a system problem.

Everybody is...nobody's a rogue player, I assume. Everybody wants to do the right thing, but together, they're doing the wrong thing. If we come together, how do we come together? Horizontal connectivity, like the MINT and TraceLink trying to do that across the supply chain.

Within the individual logistics or retail or hospital and manufacturing, they also need to do the same thing internally and digitize it because TraceLink will never be able to connect to a piece of paper in a shop floor. You know? That needs to be digital.

[laughter]

Henry: These 30 minutes go by so rapidly, but I also want to open it up to questions from the audience. I see one right here. Awesome.

Audience Member: I'll steal one. I'm out in front so I can ask the first one. I'm going to use the example of J&J because it is put up there.

When looking at systems like MINT and these connectivity systems, does it make sense when looking at AI type applications to maybe look a level down away from the pharmaceutical manufacturer and look at datasets from integrators like the Kuehne Nagels of the world that might have better cross functional data that's not siloed when looking at that narrow blend of AI?

If you're looking at the complexity of the supply chain, are there people who are better apt to start with these AI type tools?

Guy: I would say this and I think we've mentioned earlier. Let's start with the end in mind. What are you trying to solve?

I'll give you an example. I was at a conference beginning of the year, and I was talking to a colleague of mine who works at Integrator. Someone came up to him and said I want an AI solution. For what? For my warehouse. Well, what in your warehouse? Nope. I just want an AI solution to solve my warehouse.

It's like, that's just the warehouse, not the...but so I think it's what part of the supply chain? Are you looking at better demand sensing? Are you looking at logistics?

Audience Member: Specifically logistics sense.

Guy: If you look at logistics, I would say then focus on account what? Are you looking at your trucks, your routing, or your warehousing? Within that, are you trying to focus on, let's say, your labor or your put away strategy, or your slotting strategy.

Because then as you can see, you sort of boil it down and then to your point, your question is, well, yeah, maybe if it's my slotting, then maybe I do want to go to a Kuehne Nagel who might have all my data about what my SKUs look like in my warehouse and I can apply some AI intelligence to that.

Maybe it's my 3PL. Maybe it's DHL who's doing all my last-mile routing. I want to work with their data set because they're going to look at that last mile. It's really how do you boil it down to that narrow focus, and then that tells you which data set probably makes the most sense.

It's probably going to be a combination of your data set and someone else's or maybe even others. But start with that understanding of what am I trying to solve because that will tell you, I think, then what data set to go after.

Evren: Exactly. Don't do data lakes, please. Just go after specific data that you need for that use case. Don't try to couple up everything just because you may need it in the future.

Henry: One of the things we talked about was the promise of AI. Obviously, you know, Gartner has their hype cycle. I think each of you felt that there were opportunities for improvement near term with regards to logistics performance. Maybe you could comment on that.

Evren: Logistics?

[laughter]

Henrik: Something we're exploring and having some traction with is to be able to understand disruptions that are happening, whether it's a strike, which is very current or it's a port congestion or a weather or a natural disaster, and be able to using our network, we have this framework of visibility, recommendations, and actions.

Visibility, we try to identify the issues or potential that is exist within a customer supply chain. Remember, Maersk is now no longer just a shipping line, it's an end-to-end operator. That visibility layer which is a data lake, sorry to say, [laughs] but we with enough slices in that data lake to then identify what areas could be improved.

Then the recommendations or AI layer then helps us understand, given our network of assets that we own or our network of operators that work with us what are the routes that could avoid this port? Are there other sailings? Could we drop it off earlier? Could we slow it down? Could we keep it on the ship and take it on the route back?

These are some of the things that we're exploring within logistics that can help realize the power of AI and scenario planning within logistics.

Henry: Wonderful. Thank you, Henrik.

Evren: I can give three bite-sized examples. Two from my company that we're actually using today, which has nothing to do with supply chain, but AI is really adding value, and one may be with the customer situation in this digital factory.

We started using in our company as a software company, SCWAI, ChatGPT for translations. We now can offer in, like, the next client we are rolling it out in Korea. We are using basically ChatGPT to translate 95, 96 percent accuracy, and then having the clients review the exceptions, and we are done. We are launching our first product in Korea.

The same thing with Saudi Arabia in Arabic. We used to pay a lot of money to official translation service providers. Now it's literally almost free. That's one.

The second thing, we started six months ago using AI notetakers. I actually saved so many hours of extra meeting just to tell people what we discussed in the previous meeting. I just forward them the meeting link. AI summarizes the action items pretty accurately. Fathom.video, if you want to start using free. I don't get any commissions. It's an amazing tool. That's a note taker. That's already delivering value from my company.

Third one is maybe in the manufacturing floor, we integrate it with OpenAI, and we are transforming the operator comments that are unstructured text, free text fields, into classifications so that we can analyze the whole comments data feed which is organized and standardized through the OpenAI filtering and AI is basically adding value right now because it's very hard to analyze unstructured text in AI models or in any analysis, but that's another simple use case.

These examples are small, but they add value. Starting small is always a good idea in technology, then you can build the bigger AI model.

Henry: It's wonderful. We're almost out of time, so I'm going to throw one last question. This is one that the three of you seem quite, I'd say energetic about when we talked about it in preparation, and that is overcoming historical baggage.

When you talk about sort of implementing change, how do you make recommendations to overcome maybe failed activities that have occurred historically? How do you gain buy-in? How do you gain that support? How do you overcome that historical baggage?

Guy: Wow. I mean, we could write a PhD dissertation on that. I think first and foremost is open communication with your stakeholders. It's also open expectations. It's all about expectation setting. It's also about, again, change management. Understand the people you're dealing with and who you're working with.

I'll give you a very quick example. Those of you who heard my talk yesterday, I apologize, I think I said it yesterday. I remember I went into a warehouse once trying to sell them a WMS and they certainly needed a new one and, you know, so selling technology.

Guy basically told me, he said, hey, I'm retiring in three years. If I put a new system in today and it doesn't work, I get fired, I lose my pension. Guess what? I'm not doing anything for three years. You want to come and sell my next guy? Go right ahead. But for me, I'm not touching it.

I think it's an example of, it does come down to setting expectations, working with the people, to change management because you're going to have that baggage. I'm a software vendor. No amount of my marketing spin is going to change that.

I talked about being at Forrester. One of the things I remember seeing at Forrester during the dotcom boom is we got burned by the dotcoms and it took a while to get over that. Now we've come back and now I'm seeing the same thing in AI. This crazy investment in AI, we're going to get burned by it too.

So just be careful, set the right expectation, work with your people, Set the goals and be reasonable with them. But at the end of the day, there's going to be baggage, and there's no magic bullet to get rid of it.

Henry: That's very helpful. Thank you. Comments?

Henrik: Yes. Short comments. I think it's all the hygiene stuff, buy-in from the executive committee sponsorship, all of that pre mortems, identify all the risk in advance with that end in mind and backtracking that to address those in a change management project.

Then also have the incentive structure in place. There are no quick fixes. No marketing spiels. You can kind of silver bullet that you can fire but it's, and in a company like ours, 100,000 people, it's a challenge still. I'll be honest, but we we're working with some of these tools to accomplish that.

Evren: I'll just say one thing. It's like the very first thing that you said. The senior leadership sponsoring and supporting that and committing to that change.

We have examples. Same company, same client, two different factories. One is wildly successful, fully adopted, fully functioning, scoped everything in. The other one failed in the pilot because they gave it to the lowest ranking engineer.

No site head involvement, no operationalizing. The same technology, exactly the same. One great success, one couldn't take off after the pilot. So people first. Technology is just an enabler. No technology will give you all the improvements. You need to make the people believe in it, change it, and then start adopting and use it for their own good.

It starts from the top and the people first. Technology can always be found and improved.

Henry: All right. Well, thank you. People first, start small, and logistics has near-term opportunities for driving significant value with AI.