A (2:47)

So this is what year that you start.

B (2:50)

I started dabbling, let's call it dabbling in AI was 2002. So it's almost 24 years ago.

A (2:55)

24 years. But what you were playing with in 2002 was an extremely primitive version of what we have now.

B (3:04)

I think it was very relevant because the way I see it, should I have skipped all the foundations that I learned over the years and just gone to the current state? Maybe. But when I look back, I think that foundation really helped me because back then, and surprisingly, by the way, neural networks, when I talk about that, it's still very valid and relevant within AI. The entire foundation is neural networks. So I think that foundation really helped and I still find it very relevant and I apply it day to day.

A (3:32)

Yeah. Imagine I'm having a conversation with you 20 years ago and I say, what are you up to? And you say, I'm playing with this thing, neural networks. This early version, Would you have used the term artificial intelligence?

B (3:47)

Probably not. I probably would have used something called statistics, which is everyone is aware of. Back then it was more statistical, so you don't have these big algorithms at that point. But then something happened. I don't know if you heard of this company called Kaggle. They used to host these sort of data science competitions and anyone can participate and if you do really well, you get a prize. That was a good motivation just to see, okay, I learned something. Let me apply it and see how good I am. I'm getting at it. So I think that was my first entry point where I really got hands on into AI. And that probably stayed with me for a while. I think that was back in 2006. That's where I started getting hands on. And the funny thing is, when you look at these Kaggle competitions, the use cases, they used to give actual industry applications. So you are really dealing with business problems, applying AI to solve it. And then, you know, wait a minute, a medical company is using it, a manufacturing company is using it, banking is using it, and this is 2006. So it already started and then it just. Yeah, today it's a different ball game.

A (4:47)

Yeah. Now, so you, you came to Heineken when?

B (4:52)

2020.

A (4:53)

2020.

B (4:53)

Right. Middle of COVID or right after.

A (4:56)

Yeah. And Were you brought in to be the Chief AI officer? Was that the. Your correct title change?

B (5:00)

The title changed, but yes, I was the global leader of AI at the time.

A (5:03)

Yeah.

B (5:03)

Yeah.

A (5:03)

And what made you want to take

B (5:05)

the job was actually working for Amazon at that point. But when I looked at Heineken and I thought, okay, this is a legacy, traditional company. Right. And AI was not a capability embedded at that point of time. So it's a great opportunity if I can start something from scratch, really build it across the entire value chain of Heineken. I mean, that's probably the best job anyone can even ask for. Yes. It's, of course, a lot of responsibility that you have to make sure that you really build the right products and right capability. But that also happened. So I look back, it's quite fulfilling.

A (5:39)

But it's also, if I might play devil's advocate for a moment, you're also taking a risk going into an established. How long has Heineken been around?

B (5:48)

162 years, to be specific. 1864.

A (5:51)

It's a very different proposition walking into 160-year-old company and saying, I want to bring the future to the way you operate than it is with the startup.

B (6:00)

That is true, but it's also a challenge. It's a good challenge. And also the Heineken is also looking externally. There are companies that are picking up speed and embedding and adopting AI. So should we fall behind? Not really. So we also need to pick it up. So I thought it was a good challenge because the use cases were there, the opportunity. I could sense the business really was having the appetite. Let's do something different. When we apply AI in a corporate setting like this, it's super important to understand how the business actually works. What's the value chain looking like? What are the nuances? Where can I. And once you get an understanding. It took me some time, by the way, to understand the full business and the complexities. But once you cross that threshold, you figure out what's feasible, what's not, then. Then it opens up. Wait a minute. Within the value chain, I see 10 areas I can optimize.

A (6:46)

You say, once you understand the business, describe the business. What is the Heineken puzzle?

B (6:51)

So, well, puzzle. Let's see if it's puzzle after I explain. Yeah, it's actually. We start with the procurement, where you get the glasses, cans and all the raw materials. Then it comes to the brewery where the magic happens. That's where the Heineken beer is produced. Then it goes to the distributors, basically supply chain takes over, then it goes to the customers. And what we refer to as customers are the bars, restaurants, retail stores, mom and pop stores, convenience stores. And that's where actually consumers then come and actually consume the product. So that's actually the value chain. It's actually pretty linear when you think of it. But there are nuances depending on the country and the market. There are some specific rules and guardrails that you have to be aware of.

A (7:31)

So you have that process going on all around the world and across multiple brands.

B (7:37)

Multiple brands, multiple countries, multiple operating companies.

A (7:40)

Yeah. From your perspective as someone who's, who's, who is the chief AI officer, what are the tasks in front of you? What's your opportunity there?

B (7:50)

Any process that you think that is maybe not digitized or maybe not data driven, you can optimize. I look at it like a pendulum. So one side of the pendulum you have complete gut based decision making. The other extreme is completely data driven. So the idea was, can we swing this pendulum a little bit towards data driven from where we are?

A (8:10)

Give me a specific example of a problem you set out to solve or address.

B (8:16)

Yeah, sure, there are quite a few, but if I want to pick one,

A (8:19)

give me the most fun one.

B (8:20)

Fun one, maybe most complex one. Let's take that one, I think. So we spend quite a bit on advertising and Heineken is largely a lot of it marketing company and we really care about our brands and products. We are almost obsessed with it. Let's take an example. Let's say you have X million dollars as your budget and you have two brands, let's say Heineken and Dosakis. I think the audience here will be familiar with that. And then you have three touch points. TV, YouTube, Instagram. And I want to optimize my advertising budget between different brand and touchpoint. So hi Nikken on Instagram, how much should I spend? It's a very easy question to ask, but to actually solve this you have to study historically how these performed and then create a model and then predict. If I allocate my budget in this way, that's probably more optimal. Before it was more like somebody took a gut based decision saying, okay, here goes X million, here goes Y million. And we say, no, no, no, that's not the right proportion.

A (9:19)

What did the AI tell you about the accuracy of those spending decisions? In the past we looked at the

B (9:26)

return on investment from this advertising. So how much incremental volume or volume of beer are we selling or revenue are we trading? And we find out can we improve that Right. It's the moment you apply AI. And when we look at that, there's significant improvement. In some cases we have 30% uplift. 30%, 30%, 3, 0. Not everywhere. In some places we've gone. But it ranges between 10 to 30% uplift depending on the type of AI product you're building. And that impacts the top line. So it's very easy to also realize that value people get to see it.

A (9:57)

So you say, oh, we can do a way better job if we spend X more or X less in this particular area. Give me another example of a.

B (10:05)

Another one would be we have a very large sales force within Heineken. So these sales reps, what they do, they go to the outlets, the bars and restaurants and they maintain that human to human relationships with these, our customers. That's super important to maintain that. And they go solve the customer problems. Let's say someone is out of stock, somebody's about to churn or there's a price mismatch, something like this. And before they used to go like this. Let's say a sales rep on a day to day job has to visit five places. A, B, C, D, E, five different outlets and he used to go abcde. Turns out the model tells you on any given day, if you optimize taking into account the traffic conditions, instead of going from that linear route, you go to D first and then to B, then to C, then to E and then to A. And the reason for doing that is the model tells you if you visit customer D first, he has the biggest problem that needs the most amount of time to be solved. Solved. And that's how optimized. And also the sales reps now, they are becoming so educated with some of these AI models, they are now becoming business advisors. So they are no longer just solving little problems, they are having the time to say, what else can I do for you as the customer? So that I think it was a big change within Heineken because it impacted a lot of people that were using that.

A (11:23)

But in that instance it requires not just building a model that can be smarter about how people should spend their time and what they should say, but you have to obviously educate your salesforce to believe in what the AI is selling. Tell me about that piece. Is that a piece that you, that you're a part of or does someone else.

B (11:41)

Yeah, that's also part of. Because that is super important. I think we can build the best models, best algorithms, highest accuracy doesn't mean anything if it's not used the right way. So what we do, we have within our company a pretty big upskilling programs. So bring everyone along in common, understanding basic understanding of what AI does. Not everyone needs to understand neural networks or algorithms. Right. But what we do is give them a handheld device and an app which is driven by AI. Play with it, have fun, see how it changes your life. And once you start liking the product, liking the UI ux, then you start getting more. And the insights also tell you the story because once you start getting the value, I am not having to pitch my models anymore. The sales reps and the markets, they are pitching on behalf of us. And that's such a good place to be.

A (12:32)

Yeah, it's interesting. So in that instance, when you're designing a more efficient form of interaction and fruitful form of interaction between sales reps and customers, I could see a version of that where it is really clear, looking up from a high level, that things are working better. But it might not be clear to the salesperson. Is the salesperson who's now following the direction of the AI aware that they are more efficient?

B (12:59)

They are, they are.

A (13:00)

Why are they, how are they aware?

B (13:01)

They're more, they, they realize few things that they were visiting customers just because they had to visit, because it was in the schedule. Now they go there and they find out, well, wait a minute, I never tackled this big problem that was not being addressed and they solved it. And the customer feedback also comes back saying, we are really happy. So for all these products, we get the feedback not just from the sales reps, but also for the customers. Do you really like the recommendations we are giving you? And that's the best validation you can think of? Because that's firsthand feedback.

A (13:31)

When the AI is doing this ranking, it wants you to focus on the customer with the biggest problem first. Or is it much more complex than that?

B (13:39)

It's a little bit more complex than that. But usually it's rank ordered in terms of which one is the biggest problem or that needs the most amount of time. That's how it's rank order. But sometimes you can also override the model. We also give options to people you don't have to all the time, 100% follow the recommendation. If you have some urgent priority, you can override that.

A (13:59)

That's also possible with something like that. Is there a next level you can go to? So you design this system and you say, oh, I can make our sales staff a lot more effective in the way they operate with their customers. And then you see that it works and then it comes back and then you say, okay, what's 2.0? Is there a 2.0?

B (14:19)

It could be. So it's always about innovation. Then you think, okay, today we go and solve the problems that have already happened. What if we solve the problems that are likely to happen? That would be the next step. So this customer hasn't been very active for a while. There's a high chance that that customer might churn out of Heineken. So what actions can I actually recommend to make sure. And we do this, by the way, we also gather a lot of customer feedback and complaints and feedback and we use LLMs to extract and glean information. Okay, what are the real pain points? What's the theme and the topic that needs to be addressed? And once you do that, then also you can prepare ahead of time. We are already there, by the way, when I say 2.0, we're already testing it. You solve problems or you try to solve problems before they even occur. So I think that's a little bit of a 2.0. Then we have to see what else we can do with it.

A (15:09)

Tell me you've had this partnership at Heineken with IBM for, Since I think, 2013.

B (15:16)

2013.

A (15:17)

So you came in and there was already a strong working relationship. Tell me about how that relationship started and what does it mean on a practical basis. You're building all these tools. How does the interaction with IBM work?

B (15:30)

Yeah, so I think good to give a little bit of context. Heineken started this digital transformation journey in 2020 formally, but the tech was already there. We had our systems, platforms, data, everything was there. So all the IT for IT systems is where IBM was partnering with us from the get go from 2013. And it's a very long standing partnership because as we found the tech is evolving, our partnership also kept evolving because we need to keep up to the speed. So it was more about IT for IT systems, cybersecurity platform, data, incident management, service level, you name it, all of that. IBM was supporting us both in terms of hands on and also in terms of strategy to create it together. But that also evolved like I said, when we went into this digital transformation journey in 2020, then we started building this digital core, which is the central nervous system software system of Heineken. That's where IBM is really partnering with us and helping us not just shape the whole thing in terms of building it hands on, but how do we strategize that so that it lands well? So that's. Yeah, it's a long trusted partnership I think we are going to go a long way together.

A (16:40)

Yeah. Heineken space just outside of Amsterdam.

B (16:44)

The head office is Amsterdam. That's correct.

A (16:45)

So the IBM people who work with you, are they on site?

B (16:48)

There are some on site, yeah. There are some teams in India, some teams spread across the globe.

A (16:52)

Yeah.

B (16:53)

But for, for tech, I think the location doesn't matter, but you still need people on site to actually talk with the business and really understand what the problem is. For those interactions are also very important.

A (17:03)

When you said you wanted, when you got there, you wanted to build a digital nervous system, what does that mean?

B (17:08)

Maybe good. To give an example, let's say iPhone, Right. It's a central platform, but you can download thousands of apps there and all of them, once you download, seamlessly integrates with the system and you don't see any difference. This is the same thing. So what we want to build within Heineken is a central software system, which the old school way of saying it is the erp, Enterprise Resource Planning. It removes the fragmentation of different platforms. It brings it all together. It makes sure all the business applications within supply chain, commerce, finance, hr, all in one place and coordinates them, everything orchestrates them. The benefit of doing that is 2:1. Across the value chain, you have one way of doing business because everything is standardized. But we also have multiple markets globally, across the multiple markets also it becomes one way of doing business. So it's both ways. And once you standardize it, we can embed new apps which will seamlessly integrate and then you just keep scaling further. And we scale very quick. And having that digital core will really help us scale because the value from AI and Insights is not just building one product in one place, it's how quickly can you scale it and are

A (18:18)

you still building it or is it an ongoing thing?

B (18:20)

It's ongoing thing because there are nuances in markets, there are nuances in tax systems and currency systems. So it takes a little bit of as much as we want to standardize. You also have to bake in some of the nuances, otherwise people cannot use it. Yeah, so those sort of outliers we have to also bake in.

A (18:38)

You must learn something when you suddenly. Not suddenly, but when you standardize a bunch of things that have not been standardized before, presumably you have a basis for comparisons you couldn't make before.

B (18:49)

That's correct. So we also get a lot of external inspiration. So sometimes these large projects, we don't start by our own. So we get inspiration from partners like IBM or someone else. How have they done it in somewhere else and where it's really working. So then you get those ideas, the learnings, and you start building that way. And while doing that, you figure out that, wait a minute, we might have to do something different and maybe it's even better than what others have done.

A (19:15)

Yeah.

B (19:15)

So it also creates creativity.

A (19:17)

Yeah. I'm just curious whether there was an insight that you learned from that process that comes to mind.

B (19:26)

A big one, I think we don't look at tech for the sake of tech and embedding it. You know, just. Just as one would say, it's one single core, one single platform, everything coordinated. What's the big deal? The big deal is bringing people along to actually believe that there is a benefit of doing one way of business. And that actually means the entire company, not just the leadership team. So to bring everyone on board and say this, tell us how this platform should look like, what are the components we should build, it's a pretty big task. That's where the change management comes in.

A (19:58)

What was hard about that? Did you have bumps in a row?

B (20:01)

We did. I think it's about convincing people the benefit of doing this. Why do we say if we standardize something, we can go at high speed in scaling? It's not very easy to visualize it at first, but what you do is you show some proof of concepts, and that's. I won't call it a trick. This almost bread and butter of what we do. Show a small proof of concept, show that it works, show that we can scale, and then automatically people start having the faith and they will say, okay, I see it. Yeah, it makes sense.

A (20:29)

Sergi, at least half of what you've talked about is not about the tech itself, but about being a kind of evangelist for the tech. It is half the right percentage. How much of your time is spent convincing an organization and people in the organization to see the value in what you're doing, as opposed to building the thing that has value.

B (20:46)

Yeah, I think that proportion changed over time. When I first joined, I was very much into the products itself. I was to review quotes myself, let me check what's going on. And over time, of course, then you focus on somewhere else. You realize, like I said, best codes, best models are useless if not used the right way. Then I said, okay, now my time is to actually inspire and show people the value of it. And what I realized is explaining AI in very simple language really goes a long way because you take away that anxiety that a new product is coming in and we humans a little bit, have This, I don't know if it's the right thing to say, but it's inertia of rest. We like status quo. We don't sometimes like change that disrupts our. So every time you build a new product that will change our way of working. Yeah, there's inherent little bit of anxiety. Take that away. Yeah. Job becomes a lot easier.

A (21:36)

Are you a good evangelist?

B (21:39)

So far it's working. I think I can be better for sure, because it's about understanding what's the reason people sometimes might be reluctant to actually onboard or adopt a new technology. And once you sort of understand that, then that anxiety goes away. It becomes easier.

A (21:56)

How many people work for Heineken?

B (21:58)

About 85,000 to 90,000 globally.

A (22:01)

So you have essentially a city pretty much. And if you look at that universe of 85,000, is there anyone in that universe who is not touched by what you're doing?

B (22:12)

I think the way we do it is we prioritize based on the size of the market and the potential opportunity. Yes, if I had infinite resources, I would go everywhere within the Heineken company and do everything. But we cannot, we don't have infinite resources. So we say, let's be a little bit picky and choosy where the biggest opportunities are, but it's a matter of time. Right. Today we touch upon the big markets biggest scope. Over time, it's going to be pervasive through the company. But the appetite is already there. So people are really, even if they have not really embedded some product, they're asking for it, which is a fantastic place to be.

A (22:48)

Yeah, yeah. What's been your biggest disappointment so far?

B (22:52)

So far it's been very fulfilling, I must say. But I think what I would look back. Can we do things a little bit quicker?

A (22:58)

Quicker.

B (22:58)

Can we go a little bit at high speed? And that's why this whole concept of digital backbone, can we standardize everything? If we can speed that up, if we can really scale quickly, I think that will be the best. Because today I have a very good problem. People are asking for products. Sometimes I say, yeah, I need to put it on a timeline and a roadmap, because I cannot just cater to it immediately.

A (23:19)

Presumably, that's one of the things that the people you're working with at IBM can tell you. They can give you a sense of how quickly others have adopted some of these technologies.

B (23:29)

That's correct. And that's actually one of the benchmarks that you were referring to. We see. Are we losing pace and which of the things can we go forward and in some case, when you look at the digital core and the backbone, maybe specific areas we can speed up, because those are the areas that are maximum potential value. And some of them we can depriorize a little bit that we do all the time. Just a pragmatic approach to solve.

A (23:51)

I'm curious about Heineken's specific question, which is. So here you have a legacy brewer based in the Netherlands, 160 years old. If I were to say, I want you to take an entirely new job, I want you to do what you're doing, but I want you to do it for an American company in a completely different industry that's 30 years old, maybe a company that makes vacuum cleaners, 30 years old in America. How much of what you're doing? I guess what I'm trying to say is, are there things that are particular to Heineken that have made your job sort of challenging or interesting or that just wouldn't be an issue in another environment?

B (24:35)

So it's a good question, and thanks for the enticing offer, but I will politely.

A (24:41)

I tried to make it as unadverted

B (24:42)

as possible, but I will politely reject the offer. But I'll tell you why I'll reject.

A (24:46)

You're going to be in Nebraska making just one kind of vacuum cleaner.

B (24:50)

Well, I went to school in Iowa, so. Exactly. So I'm quite aware, I think there's a culture difference. We are all very passionate about our brands and products and there's a lot of it is connection based in the sense we create these connections with our customers, sometimes consumers, and it's all about maintaining that. And once you get a feel of it, you feel part of the family. That's a very good feeling to have. And the fact that today where I am, if I look back, I probably would happy to say very fortunate to have probably one of the best jobs in the world in the current times. And there is no end to innovation, by the way. And even within Heineken. Yes, it's a traditional company. Who is stopping innovation. There's a lot more to do. So I'll be very busy for next few years.

A (25:30)

Yeah. What are the Dutch like? This is one of the oldest and most successful commercial cultures in the world. A tiny country that's been astonishingly successful.

B (25:40)

That's correct.

A (25:40)

I'm just. I'm curious about innovating in that. That kind of environment. How is that different from innovating in a huge country like the United States or in a different kind of national culture?

B (25:51)

I think it's a Question of opportunity. Because within Netherlands, by the way, Netherlands has one of the most highest number of startups within Europe, if not the highest. So there is this culture of innovation that's already embedded in there. It's happening all the time. Companies like Philips, asml, some of the very big, big players already there. So it is could be a little bit different. I think in Netherlands we want to make sure what we are doing really is going to work. So there's a little bit of discussion, alignment, it's more structured but also agile in a way. We do things and us was more like let's do, let's go, quick, experiment, learn, fail. So I think there's pros and cons on both sides, but so far it's quite good.

A (26:32)

Give me a sense of what your what's a day in a life like for you. What does it look like to have the job that you have in a place like Heineken?

B (26:42)

First of all, it starts with the calendar and the number of meetings I have, which is usually filled for 40 hours or longer in the week. So that's the starting point. And then I have to pick and choose which meetings I need to prepare for what. And usually these meetings are mostly about where are we with the product, what are the challenges, how can I help and solve it? And then sometimes also pitching new products or convincing something and also sometimes change management. I also do sessions where I present internally, quite often go to different places because it always helps to be in front of the audience when you're presenting something. We also started something recently which we call AI Bootcamp, which is you use Genai as an interface for all these big AI models and people can interact in a very fun way. That's our new way of really convincing the rest of the company that hey, this is fun to play with and let's go. So I. Yeah, it's.

A (27:35)

How many people would you cycle through that kind of boot camp at any one time?

B (27:40)

Usually we keep it a small group just to make sure everyone is doing something hands on and nobody's just listening. So usually 20 to 30 people max and then you go from one place to other and it's all hands on. You cannot sit and watch. You have to participate.

A (27:54)

Are you directly involved at all in the design or creation of any of these tools?

B (27:59)

So I review it at the end. I used to review also the codes before and now I'm mostly like trying to get the feedback from the people that are using it because that's my best validation point. If I get high net Promoter score on these products. I know the job is well done, but I do check accuracy of models. Some of the basic things you check in AI, is the model drifting over time, what's the accuracy, how is it hosted on a platform? These things I check. But we also have mechanisms on those. So it's not like every time you have to dive deep and look into everything. So once you have these mechanisms in place, then these sort of tasks become easier.

A (28:33)

You've used the phrase that you want to make Heineken the best connected brewer. What does that phrase mean?

B (28:38)

Yeah, so I think it started with the ambition in 2020 when we said we are going to digitally transform. Remember the pendulum I was talking about from. From gut based to all the way to data driven. And in today's world when you think of digital transformation, there are a few components. Cybersecurity being one of them. The digital core, like I was saying, is one of them. Simplification and automation of systems is one of them. Our breweries, how can we simplify? And then comes data and AI which is the really one of the biggest components. And when you think of best connected brewer, the idea is we have been serving our consumers and customers for 162 years. What's different? If you leverage tech in today's world, I think you can really enhance the experience the customers have. The example I was giving you earlier about the salesforce going in different places and optimizing the route, that's a good example where the relation is maintained just simply by data driven insights. So if you can connect all the different applications, all the platforms, remove fragmentation, scale very quick, make sure your company is cybersecurity, things are simple and automated. That's what we call the best connected drawer. That's the ambition actually.

A (29:46)

How do you measure the success of what you're doing? In other words, do you expect that your efforts will have a measurable intangible effect on the bottom line of the company? And can you actually figure out what the impact of your efforts is?

B (30:03)

Yeah, we do. I think that is super important to measure because the first one I was referring to proof of value, that when I'm embedding some model, does it really work? So we do a B testing which is basically you keep aside some sample and you actually launch the product on a different sample and you see the difference between the two. The assumption is those that had the product and those that didn't have the product, both of them went through the same experiences because of market, seasonal, et cetera. That's one good way of doing it. And if you cannot have the luxury sometimes of doing AB testing because everyone is having high appetite, give me the product. I don't want to sit aside. Then you do some sort of causal models, like we say. So you kind of look at what would have happened if the model was not there. And then you predict that. And since the model was there, something else happened. The difference between the two is the incremental value the model is creating. AB testing is more accurate. The causal model is the other one, like I said, which called time series models a little bit less accurate. But directionally, both give you the sense that yes, it's working.

A (31:04)

What happens if you do a B testing on a new idea and you don't see a difference?

B (31:09)

In that case we will move on to something else because it means it's already optimal. Then we say, good, check that. Now let's move on to something else. But we need to just make sure that the process is still running optimally. So time to time, you keep doing AB testing anyway every six months or whatever the time frame is.

A (31:26)

Yeah.

B (31:26)

Just to make sure that it's still still relevant.

A (31:29)

What if there's a.

B (31:30)

This is.

A (31:30)

We're getting on a little bit of a digression here, but it's, it's something I've often thought about. What if the value that is being created is not measurable? So I'll give you a dumb example. When you were talking earlier about the salesman and giving them a new, you know, better instructions about how to basically spend their day. What if you tested that, discovered it didn't have any effect on the bottom line? But in fact, what was happening was that the salesmen were a lot happier with their jobs and more satisfied and were excited to come to work. Do you measure something like that?

B (32:06)

Something intangible measure. One way is NPS score, which I said net promoter score. Are you really happy with the product? Has it changed your life? That gives you a good indication. And then, and by the way, it's a numeric output, so it gives you a score between -100 to +100. And sometimes it's not even tangible. Let's say we do something for corporate affairs because they want to get external signals of consumer insights and then just glean some information. Maybe we act on it, maybe we don't. But this is for a good cause. Sometimes you just want to study the market. There's no immediate value if you don't create a product out of it or something to do with legal. If there's a Reputational risk for Heineken. Can I extract some insight that will prevent us or create the best briefing or summary or external briefing using AI that will help us protect ourselves. That's also reputational damage.

A (32:55)

One last question before we go to questions. I'm curious. When you look at the very beginning, you talked about this linear value chain. Where along that chain are you having the most impact and where are you having the least impact? I'm more interested in the second half of that.

B (33:14)

Yeah. I think we covered few things, but one area I think we can do more is really understanding consumer sentiments. And the reason for that is Heineken is people go to the bars and outlets and you're not really leaving your first hand data there. Right. You're enjoying a beer, then you walk away. I don't know exactly what you did. I can get some aggregated data to make some sense out of it. But if we can really get consumer insights as to what the consumers like and dislike, what sort of ad you like, how should I design my Heineken campaign so it resonates with a cluster of individuals that would be a little bit of holy grail as the next step. Like you were talking about 2.0 and to get consumer insights first party data, it's not super easy. So what we are trying to do is create digital twins of consumers. So at an aggregate level they give you a sense of also with agentic AI, which is also you hear a lot about to get a sense of how consumers might react to certain campaign or certain product. And that should give us quite a bit of insights that right now we don't have access to. I think that's one of the areas we could really do a lot more.

A (34:20)

Yeah. Yeah. So if I sat. Last question. If I sat down with you, it's 2026 now. We did this over five years from now, 2031, you're sitting in his chair. Tell me what the kind of. What's going to be the next big score?

B (34:36)

I think one area will be how we make our lives as employees of Heinecke a lot easier. So the repetitive boring tasks, manual tasks, can we automate those things and just use the time to do something more creative and think big about the business itself. That would be one area most on the productivity side. But the other area would be indeed, when we look at Gen Z and this is fact, I'm not saying something my own opinion. There's a trend of distinct trend of alcohol as a beverage. The consumption is on a decline. So then what's the next best thing for the new generation consumers. What will resonate, those are the pockets we need to find and I think that's where we will transition very quickly over the next five years. And if you get there, I think there will be a big success.

A (35:20)

Do you think that your specific department responsibility can help the company in discovering what the answer to that question is about?

B (35:29)

Definitely. That's the ambition. That's what we're trying to do. That's what we are really trying to get this consumer insights. I think that's the last mile, that's the one part that is left.

A (35:38)

Sergey, this has been fascinating. Thank you so much. I should say.

B (35:41)

Thank you for your questions.

A (35:42)

My uncle was a Heineken salesman in Jamaica. He was the local distributor and I have so many childhood memories of going to Jamaica and he would show up in his Heineken truck. So I, we're, we're resonating deep in my memory with this conversation. He would come and he would have a Heineken right there on the, on the table and would drink it at the end of the day. But we have, we have a few moments for questions. They're all on the screen and I don't have my glasses. Can you read them, sir?

B (36:12)

Yeah, I can, I can read them. Should we go in order? The first one?

A (36:16)

Yeah. No, no, no, no, no, no. Rookie error. Never do that. Okay, read the first four and pick the one you want to answer.

B (36:24)

Okay, good tip, but I gave it away already so I'm going to now do what I said. No, I think the first one is quite relevant. So it's a question for both of us. If you were advising a 20 year old, what three skills would you tell them to start developing right now to stay relevant in an AI driven world?

A (36:44)

Oh, wow. Well, you don't. You have a 20 year old or a near 20 year old. You have a 15 year old?

B (36:50)

You told me I have a 15 year old. Yeah.

A (36:51)

All right. What do you tell your son?

B (36:52)

I thought you were going to answer

A (36:53)

this first, but my kids are 2 and 4. I tell them to put away their toys. You, this is more, you start. This is more real. Is your 15 year old a son or a daughter?

B (37:05)

His son. And he's already tinkering with AI, he's doing his own Python coding, et cetera, which I couldn't imagine when I was 15. So I think I'll give a high level answer to be, to be actually successful. Depending on whether you're hands on within AI building models yourself or not. There are Three things I think are super important. One is having that tech background, having a common understanding of what AI really is. It always helps. Not everyone needs to have the details and algorithms and how models work. Not needed. But having that basic understanding always is good. Then you know exactly how to gauge what AI is really doing. I think the other thing is if you are in a corporate setting and you are doing something for the business, work backwards from the business and understand whatever you're building should actually touch the business and make it beneficial for them. It's not AI and modeling for the sake of it. That's for a separate research and development. If you're in a corporate world, try to build something beneficial for business. And I third one I think which myself, I learned quite a bit in my. In last six years. It's communication, talking about AI. If you use a lot of tech jargon and mathematics, sometimes people lose you. So it's about how you really narrate the story in a very simple way so people can relate to it. I think if the combination of these three has worked very well for me, so I can say that. Anything you want to add?

A (38:27)

It's funny because I met this guy who's the headmaster at a Jesuit school in Manhattan. We've been chatting and I want to do a little program at his school. And it's all about asking questions. Because we're now into the era of asking questions, right?

B (38:46)

That's correct.

A (38:46)

AI is this incredibly good tool, but you have to ask it the right questions. But this is not just true of AI, but it's also true of the. The world we're living in is a world that's so interconnected and everything involves so many different people that your distinguishing feature in many contexts is not whether not the answers you have, but the quality of the questions that you ask.

B (39:08)

That's fantastic what you said. I fully agree. I think it's about asking the right questions that really tells you, looking for that unique thing that you're missing. I fully agree.

A (39:20)

Maybe I'll invite you to this class and like you have that kind of time on your hands. If you brought up, you know, Heineken for all the kids in the school, that would. That would really.

B (39:33)

Yeah, sure. We have to build a special product for that. But let's see.

A (39:37)

All right, next question.

B (39:38)

Let's go to this one. Surajit or Malcolm. Is there a particular AI capability you are each excited to explore?

A (39:47)

That's for you, my friend.

B (39:49)

I think in the short term, I'm really looking forward to Agent tki, you hear a lot of noise and the hype and there are a lot of feedback that I'm getting from a lot of companies. Have you really embedded Agent Ki within your systems? There's a very mixed feedback. Some say yes, some say no. I think the potential of Agent tki, when we look at these tasks we do day to day. Let me give some example. Invoice management or transactional finance or very repetitive task. If you can really automate augment those things with Agent tki, I think it's going to be a game changer. If you free up 30% of our time just by embedding these things, then I can really think big. Everyone can think big. What's next? Then the creativity comes in. Otherwise all day you are stuck with the repetitive task. So I think that's what I'm really looking forward to. And this is very short term. Within the next few years.

A (40:40)

Yeah, we have, I think time for one more question. Sergi. Go for it.

B (40:47)

Let's see.

A (40:47)

This is. It's got to be the last one, always has to be the best one. So let's read for people hearing the

B (40:54)

phrase for the first time, what is the real example that shows Heineken being the best connected brewer? Basically you're asking for a proof point. Are we really becoming the best connected brewer? When we look at our markets, Heineken Mexico is a very good example. Across our value chain, if you work backwards from consumers, customers and so on, we have advertising optimization for consumers. For customers, we have next best action. Actually for customers, we have pricing and promotion optimized for the sales force, we have next best action. For the breweries, we have connected brewery. We are getting signals from these machines and optimizing them. I think it covers a significant portion of a value chain that's fully automated, end to end. So that would be a good example where we really saw the benefit of taking it to the next level when it comes to automation. So Mexico, Heineken Mexico is a good example.

A (41:45)

Thank you so much for joining us. Thank you to all of you who came to listen.

B (41:50)

Thank you.

A (41:51)

Thanks.

B (41:52)

Thank you very much.

C (41:55)

That's it for the first episode of season seven of Smart Talks with IBM. But stay tuned. There's so much more to come this season as we dive further into how AI and quantum computing are creating smarter business. Smart Talks with IBM is produced by Matt Romano, Amy Gaines McQuaid and Jake Harper. Engineering by Nina Byrd Lawrence. Mastering by Sarah Brugera. Music by Gramoscope, Strategy by Cassidy Meyer and Sophia Durlon. Special thanks to Sergey Ghosh and Michelle Gange. Post from the Heineken company Smart Talks with IBM is a production of Pushkin Industries and ruby studio at iHeartMedia. To find more Pushkin podcasts, listen on the iHeartRadio app, Apple Podcasts or wherever you listen to podcasts. I'm Malcolm Gladwell. This is a paid advertisement from IBM. The conversations on this podcast don't necessarily represent IBM's positions, strategies or opinions.