How a 160-Year-Old Startup Uses AI: The Heineken Company’s Ronald den Elzen
Me, Myself, and AI · 7 Jan 2025
Source: https://pdst.fm/e/traffic.megaphone.fm/AMMTO5643444732.mp3?updated=1743618618
Transcript: https://podcasts.opennash.com/shows/me-myself-and-ai/how-a-160-year-old-startup-uses-ai-the-heineken-companys-ronald-den-elzen-8cedb3/ (via OpenNash Podcasts, machine-generated)

Stay tuned after the episode as Shervin and I discuss the key points from today's guest.

A 160-year-old global brewing company still operates like a startup today. How does this help them put customers first in over 80 markets? Find out on today's episode.

I am Ronald den Elzen from the Heineken Company, and you're listening to Me, Myself and AI.

Welcome to Me, Myself and AI, a podcast on artificial intelligence and business. Each episode, we introduce you to someone innovating with AI. I'm Sam Ransbotham, professor of analytics at Boston College. I'm also the AI and business strategy guest editor at MIT Sloan Management Review.

And I'm Shervin Khodabandeh, senior partner with BCG and one of the leaders of our AI business. Together, MITSMR and BCG have been researching and publishing on AI since 2017, interviewing hundreds of practitioners and surveying thousands of companies on what it takes to build and to deploy and scale AI capabilities and really transform the way organizations operate. Hi everyone, today, Sam and I are excited to be talking with Ronald den Elzen, Chief Technology and Digital Officer at Heineken Group.

Ronald, welcome to the show and thank you for taking time to talk to us.

Thank you for having me. It's a pleasure.

Heineken is such a big brand, and I think most people listening to us know Heineken, and I know you've been there as chief technology and digital officer, previously as president and CEO of Heineken USA. How would you describe the company?

I've been here now for 30 years. And look, I would jokingly say Heineken is a hundred and sixty-year-old startup. We started here in Amsterdam, so born in Amsterdam and raised by the world. Our brands are sold in over 190 countries in the world, and we have large production facilities, breweries or big sales organization in 80 markets in the world. So we're basically omnipresent. And yes, we are, uh, the Heineken Company, but in many markets we have local brands like Birra Moretti in Italy, Cruzcampo in Spain, Tiger in Southeast Asia, Tecate in Mexico.

So we are a large global company with a global footprint from North America to Europe, but also available in South Africa, Nigeria, Haiti, Brazil, Indonesia and many other markets in the world.

Why did you say a startup?

Well, we are still a family controlled company, and I think in the DNA of Heineken and all our management across the world, many more people, you can feel that. You can feel this entrepreneurship, you feel the can-do mentality. We have quite senior leadership in our operating markets across the globe, and that entrepreneurship, that can-do mentality, that agility to react very quickly to local circumstances, yeah, is part of our DNA. And I find that entrepreneurship something of a startup companies.

That, by the way, has downsides as well, eh? Because it feels like we have 80 startups and balancing that with global scale, that is one of the challenges we have.

You've been a big proponent of AI and analytics in general, I know. Tell us what you're excited about.

Why I like AI is, I believe it or not, once graduated in ninety five, uh, trying to predict the stock market with neural networks. And I was not very successful. But there's always been something in me that loves data, that loves this forecast and this loves analytics, that loves the structure of data. And also my current job, which is much broader than just AI, I really find these exciting times.

What are some examples of ways that you're using artificial intelligence?

There's, I think, a broad array in Heineken where we are using, sometimes with big value AI, and sometimes more proof of concepts. We now live in an era where 95 of what we talk about is gen AI. But truth be told, 95 of where Heineken gets its value from, and also a large part of where we invest in, is classical analytical AI, a machine learning model, predictive analytics, et cetera. But if I split the two, then we are doing quite a bit on these more predictive models around, of course, revenue management, optimization of promotions, consumer promotions in grocery stores, business to business trade promotions.

We do quite a bit, as you can imagine, a company like Heineken. There's a lot of advertising and promotion, so advertising, marketing mix, commercial mix optimization, sales execution. So what next best actions, product recommendation. So a lot on the commerce side, because that is where we need to win and want to win and need to be competitive. There's of course a large cost base in manufacturing and logistics. And by the way, machine learning is a fairly good tool to optimize logistic movements of trucks and even truck loading, truck movements and truckload, but also manufacturing, where more and more we use data with IoT.

We take data out of our machines in every brewery we have across the globe and try to analyze this and predict behavior of machines and of course optimize the output. So a lot, I would say, across the entire value chain, from consumers to customer engagements to machine optimization to truck. That's where really the value is that we can measure, we can track and we can drive.

Which, by the way, is something we see across industries and particularly in consumer packaged goods, is that there's just so much value on the good. I would don't want to even say old fashioned, because a lot of these techniques aren't really old fashioned, but basically good optimization and prediction and stuff that I would say the predictive AI is really, really good at. Sometimes companies even forget about that and focus so much on gen AI. Do I, I you see that?

agree, and by the way, I can even make it worse. I think there's even a lot of money in BI, just, uh, understanding descriptive, how it works and what happened in the past. We underestimate something what's under our nose. We all like the new shiny object, but BI, classical AI and now gen AI, I think it will very fast take off. We can talk about that in a second, but I agree with you. So for us, it's really important to still keep a good focus on these classical AI models.

When you said a startup, a 160-year-old startup, a part of me thinks a startup is not so bad for AI because there's so much experimentation and so much learning new things, and you want to be agile and learn. On the other hand, you also are running scale. So tell us how you think about the startup culture and AI and how that's been.

Well, I would say a one-year startup would be great because you're digital native and you start from scratch. A hundred and sixty-year startup has a lot of legacy. And I think, like many other large companies like Heineken, we are still quite fragmented in our tech landscape, in our process landscape, in our data definitions. And that was for a long time not an issue, right? Being very local, close to your communities is a strength of the company.

How do we keep this local strength, local understanding, but balance that with global scale? And big data becomes quite valuable when you can access not only the data, but also make it comparable. This startup mentality has for a long time helped it to be agile, fast, adaptive in the marketplace, but in the last couple of years became more of a hindrance because the lack of standard data definitions and standard data access was inhibiting faster scaling. So we really had to go back and say, across the globe, we need to make an intervention.

We really need to work on data quality. And I have to say, I have never seen before that our general managers, our managing directors in markets, start understanding that they need to invest in the quality of data, that they start to understand that unique data, so first party data, is relevant to compete in the marketplace. We have very senior people talking about not only the importance of, but the value of data. That's a huge win.

That's game changing.

You're bringing up some very interesting trade-offs that seem honestly quite difficult to me. On the one hand, it'd be great to have everything all cleaned up and beautiful and pretty. But while When you're doing that, you're not making progress on some other model. When you're spending money on traditional AI, you're not necessarily working on generative AI. As follows, Shervin and I were looking at the same issue in a report that we published. And, you know, one of the things we asked people was, is the organization's focus on generative AI taking attention from traditional AI initiatives?

And in our research, we found that only 26 of the people thought that gen AI was distracting. And I think that's important here. People are not forgetting the idea of traditional AI. But I wonder about that. How do you make these decisions? How do you decide, hey, we should spend one dollar cleaning up data or one dollar building new model, one dollar on traditional or one dollar on generative? How do you make that decision?

Yeah, that is not easy, and I think that's the one where we make, of course, most mistakes. And I don't think we live in a time where you can choose to do either or you need to balance that. So it's a bit about the benefit you can get in the short term while investing in the long term. It's about cleaning up historical legacy systems and at the same time building for the future. In Heineken, we even have a strategy that's balanced, the digital strategy balanced across those two big streams.

I would love to say, look, let's wait with all the new stuff and for the next five years really sort out the back end, harmonize, standardize, clean it up, centralize, go to the cloud, everything, do it, make it perfect. But we are gonna be out of business. At the same time, if I only invest on the front end of the market, highly innovative AI model, and be it classical AI or gen AI. If you only do that, then we get stuck in a spaghetti, a labyrinth of all kind of mapping of historical data sets.

So you need to do both, and you need to try to do that intentionally, very deliberate. And so one of the things we did in Heineken is creating what we call a program board, which is a subcommittee of our executive team where we try to decide because everything is nowadays interdependent. We try to decide where to focus, what to do first and what to do next to understand how all these things mingle together. And if you do this smartly and intentionally, actually all little bits can lead to a future goal.

You also said making mistakes, and I couldn't help but think of that as actually a positive thing, whereas maybe 1520 years ago in technology, that wouldn't have been such a positive thing. How do you think about the necessity of making mistakes and the culture and the mindset that some mistakes have to be made and that the learning from those are far more valuable than the potential damages of a mistake.

Yeah, I think I wouldn't be where I am today if I didn't make a lot of mistakes in my life. And culturally, we find it difficult to acknowledge mistakes. And actually, uh, we launched something we called an F it award. Sorry to say it like that. I hope that's allowed on the podcast because we want to stimulate a culture of we need to try, eh? You can only win if you're willing sometimes to lose. I think the good news is that the cost of failure is coming down tremendously.

You can do much smaller experiments and many more at the same time. So I see it as my role to minimize the cost of failure. I see it as my role to as quickly as possible understand. Yeah, we need to stop and then say, let's take a new decision and do it again. I also see it as my role to stimulate a level of risk appetite. A and not just risk awareness. If people are really afraid of making mistakes because they think they cannot make any career, that would be wrong.

So we need to protect the people that are willing to make the mistakes. But I truly believe it's easier now than it was ten years ago, because the cost of failure does not have to be as high as it was ten years ago. And that has really changed.

Yeah, and that's actually quite interesting way to look at it. Not to avoid or minimize mistakes, but actually minimize the cost of failure, which is a different mindset. When it comes to digital and AI, what is your portfolio of buy versus build versus partner? How do you think about the landscape? I know it's a very general question, but I've seen companies index entirely one way or the other with sort of very hard and fast rules. I've seen companies be a bit more, I would say, nuanced, with some guiding principles as to how they might make those decisions.

How do you think about that?

Also, this has changed. I think we have come to the conclusion in the last probably four or five years or so that certain competencies suddenly became critical or of competitive importance for Heineken. And it was always, of course, we want to be the best marketeers, we want to have the best consumer insights, we want to brew perfect Heineken. And now also machine learning engineers, data scientists, software engineers, and cybersecurity specialists become very important in the company.

So one, I would say, is that more and more we start to value these critical roles in the company. Two is we try to adhere to a bit of a process when we do AI models. One is we try to ask the business, and I should say digital and technology is an essential part of the business. But we try to ask the commerce people or supply chain people or finance people, where do you think the big value could be?

And we very often use external parties to help us to identify that. If a partner can help to say, look, this is how you could look at revenue. Have you considered this? We start to design that, then we build it, we said we want to build in house, we want to own the code. If we have coded it ourselves, it's much easier over time to maintain. And we, per definition, don't build one product in one go.

You start adding releases, new features, new functionality over time. So design external, build ourselves, and then when we deploy, we get into the adoption. Because it's not about here is an AI model or an application or piece of software. No, the real change and the real effort and the thing we always, always underestimate is changing the behavior of the users. And that can be customers, consumers, often also our own employees. And that is where we still use a lot of external help.

After that, the whole cycle, we run the project, we do continuous process, and we get to DevOps. We can do that better, but I think it's essential that you do that together. So there's a part of the process we look outside, then a part we want to do internal.

We like to close down our episodes with a series of small questions, just rapid-fire questions. Just answer with the first thing that comes to mind. What do you think is the biggest opportunity for AI right now?

Automation and very quick decision-making. That's the biggest opportunity for AI.

What's the biggest misconception about AI right now?

That AI is only for the very few, and you can wait and sit it out. I think it is highly urgent. Everybody should be curious, learn as much as they can, because it will impact everyone's life.

What was the first career that you wanted?

I wanted to work for an investment bank.

I guess if your stock market prowess had been better, you might have taken a completely different path then.

We we started this podcast by talking about failures. This was one of my first ones. I did not succeed.

When are people using too But you did minimize much AI?

the cost of failure, for sure.

Yeah, you minimize the cost of failure, so that ties back in.

I work in beer. I like it when people think the glass is half full, so thank you for that.

So when is there too much artificial intelligence? When are people using these tools too much?

When the human interaction is gone. We have a purpose in the company, which is the joy of true togetherness. I really hope I will never talk to AI and drink a beer with AI. If real, true connections are out of our life, then it has gone really too far.

Yep, well said.

What's one thing do you wish that AI could do that it currently can't do? What's limiting us?

The limit is our own understanding and capability to deal with it. I really think the biggest thing today is the speed of society, the understanding what we can do, and we need to do that carefully. We need to bring people along. If change is too quick, that will lead to polarization. I think actually that is one of the things we see in society today, that some things are going so fast, it's very difficult to oversee the consequences.

So we need to take people on the journey, and this is sometimes difficult to understand. AI can be a black box.

I think that's really interesting because that ties into your third step of deploy and adoption and how difficult that is and why you're working with external people for for exactly those things.

We have now learned that when we deploy an AI model, we make three pieces of the model. One is we call it decomposer. Basically, that is looking back what has really happened. And if you give that to somebody with 30 years of Heineken experience, yeah, that kind of makes sense. I can see that. Then the second piece we make, we call a simulator. So what if I would put more money on Heineken or more money on Dos Equis or less money on Walmart and more money on Kroger?

And then, you know, yeah, I can understand it will work like that. And then the last part is the optimizer, which is the black box. And the first two, hopefully, they will not use a lot, but they will help adoption. Because it helps them to say, I trust the model without really knowing what's happening under the hood. So adoption is key.

Well, it's been a very refreshing Yeah, exactly. conversation. Lots of people want to talk Yeah, about how amazing everything exactly. is going. It's been refreshing. You're willing to talk Yeah. about some of the things that that you've learned from in doing things poorly. Yeah. Thank you for joining us today.

Thank you again for having me and hope we can make a lot of failures to learn very fast and then really make impact on society and help our customers to grow.

We just finished our conversation with Ronald and I found it really illuminating and inspiring. He actually gave pretty, I would say, valuable ideas on playbook that I would expect many CIOs or chief technology digital officer CEOs are struggling with. This is a constant balancing act, the willingness and in fact the desire to make mistakes, but manage them by minimizing the cost of mistakes or the cost of failures versus avoidance. I think that ties very nicely into what we've also been saying about the importance of learning and organizational learning and experimentation.

I thought it was interesting the idea of his program board with all the people talking about all the different projects. You know, it's easy to say that on a podcast, but it just seems really hard to get all that information in one place to be able to look at it all at the same time. I bet that's a challenge.

Well, I mean, one of the other things that here is it's a complex company, right? I mean, it's so many different brands, so many different geographies. There is such local importance, local knowledge, local expertise, local context, yet it's a global company, so it needs to run at scale, yet it needs to adopt locally. And so this idea that, look, life is not just as simple and things are complex, and it's okay for them to be complex.

We should just understand the complexity. I also liked when we were talking about AI and gen AI, he also went to BI. Good old-fashioned business intelligence, like, unless you appreciate the nuances of what is going on, how are you going to design or develop anything that's going to create value?

Yeah, he broke that down well with the decomposition. That was a First step, I bet that's pretty highly BI oriented.

And the other thing I found very refreshing was this complete openness to call a mistake for what it was.

Our listeners may not appreciate this, but that most people obviously come on and they want to tell good stories about things that have gone well, and we do like to pay attention to those stories of things going well. I mean, that's that's enjoyable to hear about. It's nice to not have the impression that every project goes well for every company perfectly all the time. And I thought it was really interesting how he broke down the steps of projects to design, build, deploy, and capture value.

But then he had lined those up with where they're getting internal and external knowledge from. He clearly put some thought into that and organized the way that they're

that's very important. that they're trying Yeah,

to combine information from the best sources, and they've thought that through. I thought that was interesting. agreed. He clearly was a glass half full kind of person, and I think that's a good theme for our episode.

Thanks for listening, everyone. If this episode inspired you to buy a Heineken product and maybe think about AI while you're enjoying it, well, that's not a bad idea. We're back in early February to talk with Jeff Miller, executive vice president of communications for the National Football League, just in time for the Super Bowl. Please join us.

Thanks for listening to me, myself and AI. Our show is able to continue in large part due to listener support. Your streams and downloads make a big difference. If you have a moment, please consider leaving us an Apple Podcast review or a rating on Spotify and share our show with others you think might find it interesting and helpful.