Artificial Intelligence and Business Strategy
In collaboration withBCG
Anders Butzbach Christensen began his career in product management before landing his dream job working for the Lego Group in Denmark. Today, as head of data engineering, he’s leading Lego’s digital transformation with a specific focus on designing and building data products, including self-service applications that technology and business teams can all use to better serve their customers.
In this episode of the Me, Myself, and AI podcast, Anders joins Sam Ransbotham and Shervin Khodabandeh to describe how the Lego Group is approaching digital transformation, and how the toymaker is empowering its product teams by becoming a product-, architecture-, and engineering-led company.
Read more about our show and follow along with the series at https://sloanreview.mit.edu/aipodcast.
Give your feedback in this two-question survey.
Sam Ransbotham: Our guests often use Lego as an analogy for how organizations can build up solutions with data. But today, find out how Lego itself builds data components that connect as easily as its bricks.
Anders Butzbach Christensen: I’m Anders Butzbach Christensen from the Lego Group, and you’re listening to Me, Myself, and AI.
Sam Ransbotham: Welcome to Me, Myself, and AI, a podcast on artificial intelligence in 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.
Shervin Khodabandeh: And I’m Shervin Khodabandeh, senior partner with BCG and one of the leaders of our AI business. Together, MIT SMR 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.
Sam Ransbotham: Today, Shervin and I are excited to be joined by Anders Christensen. He’s the head of data engineering at Lego Group. Anders, thanks for taking the time to join us. Welcome.
Anders Butzbach Christensen: Thanks for having me, Sam.
Sam Ransbotham: First, tell us a little bit about what you do at Lego Group.
Anders Butzbach Christensen: I’m heading up the data engineering department within the Lego Group. We currently consist of three large global parts, each within my area. Two of the teams focus on self-service, enabling the organization to make data-driven decisions, and the last one is building a customer 360[-degree] view that allows us to build personalized experiences.
Sam Ransbotham: Let’s start with the first one. What does that mean, to be self-service?
Anders Butzbach Christensen: So a little less than two years ago, we started out our exploration of the digital transformation within the Lego Group, and for us that basically meant that we needed to do a lot of upskilling and we needed to focus on having the right competencies and teams and ways of working within the organization — so basically building the right digital foundation. And in order for us to enable the four customer groups that we have — the consumers, the shareholders, the partners, and our colleagues — we needed to make sure that they had all the right tooling to do so. And a huge part of doing that is self-service, enabling them to make data-driven decisions. So, what we did was that we took this centralized data platform, [like] almost all large companies have today, and then we made that available for everyone to use, basically. And that’s self-service.
Sam Ransbotham: So, what does that look like? If I sit down tomorrow with the Lego Group, and they won’t let me play with the bricks, how do I play with the data?
Anders Butzbach Christensen: What it basically means is that it’s super easy for the product team throughout the organization to come with their data, bring it into the platform, and then to play around with the data — transform it in whatever way they want to — and then expose it for whatever use they have. That might be for analytical purposes, but it could also be for data science purposes, etc. Making that journey as easy as possible and available to all types of skill sets within the organization is what it looks like.
Right now, it’s used for basically everything: That’s all types of data coming from our websites flowing into the platform. And then we look at how customers behave on the website and then provide the best possible recommendation experience to them.
That’s one thing, but we also use it for forecasting, for example. We have a lot of different data sets coming in from our demand planners across the globe that all get built into a beautiful data product that’s used for creating this forecasting model.
Shervin Khodabandeh: Anders, what I’m hearing is data platforms and data engineering, but I’m also hearing data science in there — recommendations and demand planning. Does your group do both?
Anders Butzbach Christensen: The way that we are organized within the organization is called the data office. We do have a data science [area]; they focus on a lot of the data science work. We also do use data science within my area, but the way that we utilize it is for enablement. So this could be, how do we build data that allows people to innovate faster? In our use case, that is enabling synthetic data on the platform. So whenever someone comes along and wants to utilize a data set that potentially contains personally identifiable information [PII], they need a legal approval, right? And that is because we need to take care of our customers’ data.
So, what we do is that we give them a synthetic data set that’s generated based on the schema, and then they can get started right away. It has zero similarities [with] the original data set, but it gives them the same output. That’s a way for us to make that data science team innovate way faster than they originally would, because it can take months before you get that legal approval.
Shervin Khodabandeh: Yes, yes. Synthetic data’s really quite brilliant. Sam, if you remember, when we had the conversation with …
Sam Ransbotham: Humana.
Shervin Khodabandeh: Humana, yes. As well as, I think, with Moderna.
Sam Ransbotham: Oh, mm-hmm.
Shervin Khodabandeh: Synthetic data for the purpose of experimentation. It’s quite fascinating. Anders, you mentioned “data product” a few times. It’s been a pretty big buzzword for about a decade, at least that I’ve heard. A lot of organizations we speak to are building data products. How would you define “data product”? What would constitute a data product? What are components of a data product?
Anders Butzbach Christensen: If we first of all look at the definition of the data product, then for us, a data product is a way of thinking. It’s a mindset, right? So we used to think of data as a byproduct — something that is a part of a product out in the business, or something like that. But to us, it’s really important to focus on data as a product, and that’s why we call it a data product.
If you are very technical, it could be a database that consists of a lot of data sets with a lot of different data attributes within it. In theory, it’s a lot of different inputs that go into one product, right? But for us, it’s really the mindset. It is the fact that that data is no longer just something we use as a part of our daily work. It is how we make decisions. It’s how we create value.
Shervin Khodabandeh: When I think about a product, a product would have a product road map, and it would evolve over time, and you would innovate on a product. Is that also what’s going on with your data products?
Anders Butzbach Christensen: Definitely. We view it as any other product. It’s no different than another software product we build. A good example is if we look at the 360[-degree] view that we’re building within my area, there are a lot of different data sets that go into that data product. I think we have four or five different IDs across the Lego Group that we need to stitch together. In order to build this 360 view, we need to bring in those IDs, and in order for us to provide value to … let’s say that we want to build a personalized email campaign or something like that. The way that we evolved this product is by looking at what is the minimum viable product that we need to build. What data sets do we need to bring in, attributes, etc.? And then we can basically evolve from there on. You don’t need all five IDs in order to deliver that value. You could actually bring in maybe one or two and then start building those personalized campaigns.
Shervin Khodabandeh: The beauty of what you’re saying is, in addition to building those intelligence products on top of your data product, you’re not just building it one time, because the rest of your organization can also use a lot of those assets in the future. And this is quite important, because a lot of the time you see in an organization, maybe they build a personalization engine specifically for what they want, and all of that data pipeline and data engineering then go to waste. And somebody in the next room would do the same thing, and they’d go all the way back to scratch from the raw data again.
And I think what you’re saying is, you’ve created the commonality, or the common layers of data, and then folks just use what they need when they need it, but it’s not duplicative.
Anders Butzbach Christensen: You’re spot-on, and that actually [speaks] a lot to the reusability of data products across the organization. And then, when you have onboarded all this data into the platform and built all these data products, it’s also really important that you make it discoverable to the organization so that others can utilize it for other purposes and create other types of value.
What we do is that we have a discoverability tool, where you can go in and look at the different data sets. Of course, there are private data sets, because if it contains, let’s say, personally identifiable data, then it has to be private and you need to request access. But that could also be data products that don’t need to be really private and can be publicly available for people to tap into.
And it’s really important for us to make it discoverable on the platform so that you can start stitching the data together and create new types of value, and you don’t have to reuse or re-ingest or things like that, as we have seen across the different organizations for years.
Instead of being very restrictive about how you build these types of platforms, you need to think of it as empowering the product teams to get as much freedom as humanly possible.
Shervin Khodabandeh: It’s sort of like the actual physical Lego blocks, right? Quick digression: We probably have 300 Lego sets that my kids have built. And they serve their original purpose, following the instructions, and then they end up creating other monstrosities or cool creations, so I could totally see that analogy.
Sam Ransbotham: I was just thinking about that. We have so many people who come on our podcast and mention, “Oh, it’s like a building block,” and they always make an analogy to Lego, of course. And Lego itself is using these building blocks and their own analogy.
Shervin Khodabandeh: Yes.
Anders Butzbach Christensen: Couldn’t agree more, but I think it’s also really important to say that what we do is that Lego is a physical toy brand, right? We build physical toys, and what we do in the digital space is that we enable better experiences for those physical products. It’s just to enhance that experience for our users, and to make the experience of buying products and interacting with products way more fun for our consumers. And I also think that that [speaks] very much to how you can use things like machine learning or AI to then make sure that we do it in a responsible way.
So, what we are also looking at is, how can we utilize machine learning to audit the platform and make sure that we don’t have data lying around forever, but it actually gets cleaned up, and at the same time, also look at PII detection? How do we make sure that our builds and product teams actually know if they have PII data within the platform or not?
Sam Ransbotham: People can’t see this, but in the background, Anders’s office is filled with these toys. I’m quite jealous, because my background is pretty plain here. What kinds of things are people building with your data bricks?
Anders Butzbach Christensen: We actually do have a lot of cool experiences that got published not too long ago. We did the launch of Lego Super Mario, I think last year, and that was a huge success. That is a toy that is not only physical, but it’s actually also interactable through these digital experiences, and that’s something that was a huge success to us.
Shervin Khodabandeh: When I hear [about] the data product and various usages of it … I think within the enterprise, you mentioned planning and optimization, and I could imagine, as a customer, interacting with Lego websites or instructions or purchasing recommendation engines and things like that. But you also mentioned something around the actual experience of building, like during play. Are there also products that enable better play or different play experiences?
Anders Butzbach Christensen: Around three or four years ago, we released a mobile application where you could use AI to do different play experiences. But … you needed to have a phone or an iPad in order to interact with the experience. I think for us, it is combining it, like we did in Super Mario, where it was within the toy, that made the difference.
Shervin Khodabandeh: Anders, this is quite fascinating, what you’ve shared with us, and I’m sure many of our listeners and folks who wear your hat in other organizations are wondering if there is a secret sauce or if there’s a recipe. Do you want to share with them and us, what does it take to actually build what you guys are building?
Anders Butzbach Christensen: I think one of the really important things for us has been to empower our product teams to actually build these products and take the ownership of them, and the way that we did that was by establishing three different pillars within the company. We wanted to be product-led, architecture-led, and then engineering-led. So, what it means is that being product-led … we actually recognize that the different teams are products themselves. We no longer have these projects that have a start and end date, but it’s products that are evolving.
Then the second one coming up is actually looking at the architecture itself and making sure that the products we build, like data platforms — it’s not only something that solves a short-term need but also a long-term need, and that we make sure that we don’t need to redo things over time. I think that way of thinking has been really important for us, and it also sets the direction for our teams.
Then the engineering-led for us is a lot around the way that we deliver our technologies and make them available across the company but also outside the company. We actually strongly believe that the engineers are the specialists. So when a product manager or management comes along, they, of course, set the scene by talking about the “why” and “what” we do things, but the team is solely responsible for the “how,” and they need to figure out, you know, when we talk about synthetic data, “OK, if we need to work with synthetic data, then how do we do that in the best way? What does a POC [proof of concept] look like? What does a minimum viable product look like, and how do we evolve that over time?”
They need to figure that out and tell us, and then we, of course, look at things like, how can we deliver value as soon as possible? And that’s getting those POCs out there, tested, and making sure that they deliver value to the organization. For us, the engineering-led is really important and one of the key reasons why we have had the success we have today.
I think we sometimes also need to remember that we hire these skilled people because they are the specialists and they are the best of the best, right? So empowering the teams is just a key thing in order to achieve great success.
Sam Ransbotham: How did you end up in this role? I mean, our show is Me, Myself, and AI. I’m curious how you actually ended up involved in all these things.
Anders Butzbach Christensen: I think that’s a longer story. I have an educational background within computer science and in web development, and I actually started out as a consultant seven years ago. I did mobile applications and websites and moved into project management of the clients that we’d built those products for. And then, I think as so many other people in Denmark maybe dream about working for Lego … we’ve all played with the bricks, and we dream about working for them.
Sam Ransbotham: It’s not just Denmark.
Anders Butzbach Christensen: You’re absolutely right, and we also see that. But totally by coincidence I saw this job ad that said something about “senior product owner, big data,” and I probably didn’t know what big data was, but the “senior product owner” I knew, because we were also working with agile within the consultancy agency. So I applied for the position without knowing too much about what the domain was. And then, after four or five rounds of interviews I, by some coincidence, managed to get the role.
I started out in the Lego Group. I had our recommendations team for Lego.com and Lego Life, which is our social app. Then I did imagery moderation and tagging, to make sure that we have safe content within our platforms. Then, after half a year, I got the job to digitalize our supply chain, building demand forecasts, etc., which was a huge project, and I didn’t know anything about supply chain, forecasting, or anything. And then we decided to accelerate our digital transformation, and then I became head of data engineering.
When we started out, there were not a lot of companies that had tried that before, so we needed to do a lot of discovery and research to see how other companies do that and then try to stitch everything together. And that basically leads us to where we are today — to where I am. And I think the curiosity around data is what’s kept me here.
Sam Ransbotham: That’s great. OK, so you’ve set me up for this: Many of the things you’re talking about with your digital transformation seem like you’re fairly mature in them at this point. You mentioned, for example, the synthetic data that you’re working on, and the PII identification. What’s next? What kinds of things are you and the Lego Group headed toward?
Anders Butzbach Christensen: There is a huge journey ahead of us. There is, of course, a lot of product teams that use the platform today. I think we have 600-plus data products on our platform. But there are still a lot of digital product teams that are not using it, especially through self-service. There is a long journey ahead of us to build up the ambassadors around the organization and make it as easy as humanly possible to build those data products that create a lot of value.
Shervin Khodabandeh: You’ve been on this journey for quite some time and have a lot of lessons learned and experiences. If you were to maybe fast-forward, what would be ideal? If you sit down and say, “Wouldn’t it be cool if, 10 years from now, when organizations are talking about data products, they could have X, Y, Z?” What is “X, Y, Z” 10 years from now?
Anders Butzbach Christensen: What I dream about is that all data producers expose their data into this platform so that it’s available in the discovery tool, and I dream about a one-stop shop, and then it’s really easy, no matter what skill set you have, to stitch that together into a data product.
And for me, that goes for technical teams building transformation scripts in Python or another programming language, to businesspeople who go into a drag-and-drop tool and then suddenly it’s available in a reporting tool. So, what I dream about is this one-stop shop for everyone across the organization, enabling them to become data-driven. And if I look ahead, that’s where we are in hopefully before 10 years.
Sam Ransbotham: You know, Shervin, as we step back, we’ve heard from a lot of people who are expressing some similar types of things but in different domains. Obviously, not everyone is making children’s toys … or toys for grown-ups; as Shervin just mentioned, he still plays with Lego. But we’ve seen people in health care, we’ve seen … Ziad Obermeyer … at Nightingale [Open Science], they’re building an infrastructure to allow people to get access to the data and use things. I think we’re seeing this everywhere. That seems like it’s kind of exciting.
Is that where we’re headed? Is every person then going to become able to do all these things on their own?
Anders Butzbach Christensen: I truly hope so, and I think we’re seeing it in a lot of places right now. AI is moving forward, and Google’s not Google anymore. Now we have ChatGPT coming up, and there’s just a lot of technologies that we couldn’t even imagine five years ago. They’re now publicly available in the hands of normal people.
Shervin Khodabandeh: First of all, this is quite fascinating, and thank you for sharing. We’re going to move to another segment, where we just ask you five rapid-fire questions and you give us the first thing that comes to your mind.
Anders Butzbach Christensen: Go ahead.
Shervin Khodabandeh: All right; let me actually get the questions. What is your proudest AI moment? Or, let me generalize — data moment?
Anders Butzbach Christensen: One of my proudest moments with data was when we actually moved from having a lot of different types of data coming in with different types of data qualities that we couldn’t stitch together. It didn’t create the value we needed, because they just didn’t do things in the right way. So, what we did was that we built a tool that educated the data producers on how to actually create good, high-quality data products. And that was a huge success and, I think, one of the key reasons behind that data product becoming a success.
Shervin Khodabandeh: What worries you about AI?
Anders Butzbach Christensen: The worry is not that it’s going to take over our jobs. I think the worry is that we unfortunately never can stop learning, right?
Shervin Khodabandeh: That’s very good. That’s very true. It’s also the same thing you said about, you’re never done with the data. What’s your favorite activity that involves no technology?
Anders Butzbach Christensen: For me, it’s physical activity. When I get off work after many hours, I need to do something that takes my mind off data and tech.
Shervin Khodabandeh: What was the first career that you wanted? What did you want to be when you grew up?
Anders Butzbach Christensen: First off, I actually wanted to be a lawyer, which I found out wasn’t for me. I have an aunt that is a lawyer, and I borrowed one of her schoolbooks, and I looked at it and I found out that there’s way too many laws. I gave up on that fairly quickly. And then I wanted to be a trader, and then I found out that that requires looking at stocks on a screen for a long time, and that wasn’t for me either.
Then I moved into tech and building different types of applications, and I think what I really found cool about IT is that you can build products that create a lot of value, a lot of revenue, without having to buy a lot of things. You don’t need a physical product, and I think that was something that really got me into IT.
Shervin Khodabandeh: What’s your greatest wish for AI in the future?
Anders Butzbach Christensen: My greatest wish is that it will make the world a better place, and I’ll leave it there because that means … that can be done in different ways.
Sam Ransbotham: You’re not going to tell us how. You’re going to hold that for the next time we talk to you, I guess.
Anders Butzbach Christensen: Exactly.
Sam Ransbotham: I think what’s particularly interesting about today’s discussion is, a lot of the people that Shervin and I talk to talk about Lego as building blocks, and they make an analogy of the things that they’re doing in their organization: “Oh, we’re building these Legos so that people can build data.” And what they don’t realize is … they think they’re talking about Lego bricks, but they’re actually talking about the way that Lego approaches data. I think that’s pretty fascinating. I think that’s the kind of thing that a lot of people can learn from. Thanks for taking the time to talk with us. [We] appreciate your sharing this.
Shervin Khodabandeh: Yeah, thanks, Anders. This has been great.
Sam Ransbotham: Thanks for listening. Next time, we’ll talk with Rathi Murthy, CTO and president of Expedia product and technology. Please join us.
Allison Ryder: Thanks for listening to Me, Myself, and AI. We believe, like you, that the conversation about AI implementation doesn’t start and stop with this podcast. That’s why we’ve created a group on LinkedIn specifically for listeners like you. It’s called AI for Leaders, and if you join us, you can chat with show creators and hosts, ask your own questions, share your insights, and gain access to valuable resources about AI implementation from MIT SMR and BCG. You can access it by visiting mitsmr.com/AIforLeaders. We’ll put that link in the show notes, and we hope to see you there.