Integrating AI into industrial companies: A conversation with Arun Narayanan, former chief data officer at Anglo American
AI, Data & Analytics

Integrating AI into industrial companies: A conversation with Arun Narayanan, former chief data officer at Anglo American

Arun Narayanan discusses building teams to integrate new technologies into businesses and shares his perspectives on advances in AI.
Listen to the Heidrick & Struggles Leadership Podcast on Apple Podcasts Listen to the Heidrick & Struggles Leadership Podcast on Spotify

In this episode of The Heidrick & Struggles Leadership Podcast, Heidrick & Struggles’ Sam Burman speaks to Arun Narayanan, the former chief data officer at Anglo American, about the significance of getting culture right when it comes to driving a digital transformation. Narayanan shares how he came to be the chief data officer of mining company Anglo American and when AI became part of the role, providing some practical examples of how he applied both data and AI to solve problems within the company, navigating the path of buy, build, and partner. Finally, Narayanan offers advice to leaders who are currently wrestling with their approach to both digital transformation and AI and shares his view on the implications that come with these developments and how leaders can adapt for the future.

Below is a full transcript of the episode, which has been edited for clarity.

Welcome to The Heidrick & Struggles Leadership Podcast. Heidrick is the premier global provider of senior-level executive search and leadership consulting services. Diversity and inclusion, leading through tumultuous times, and building thriving teams and organizations are among the core issues we talk with leaders about every day, including in our podcasts. Thank you for joining the conversation.

Sam Burman: Hi, I'm Sam Burman, a partner in Heidrick & Struggles London office and the global managing partner of Artificial Intelligence, Crypto and Digital Assets, Cybersecurity, Health Tech, and Industrial Tech Specialty practices. In today's podcast, I'm excited to speak to Arun Narayanan, former chief data officer at Anglo American. Headquartered in London, Anglo American is a leading global mining company that uses the latest technology to discover new resources to mine, process, move and market future–enabling products to customers. Arun was responsible for leading Anglo American’s digital transformation by building their data analytics platform, VOXEL—the Vision of Operational Excellence in mining. Prior to Anglo American, he was vice president, data and analytics at Schlumberger Software. 

Arun, welcome and thank you for taking the time to speak with us today. 

Arun Narayanan: Thank you, Sam. Thanks for having me, it's a pleasure to be here. 

Sam Burman: Let's start off with a question around your journey in the industry. We'd love to hear a bit more about how you got to your role as chief data officer of Anglo American and why you took the role, especially given the mining industry isn't necessarily known for being technologically progressive. 

Arun Narayanan: Philosophically, it kind of makes sense to start in industries where there's a lot of need for digital transformation. If you think about what the biggest bang for your buck or the type of impact is that can be had, it can be found in industries that need these digital tools. So maybe that's one reason why I joined the mining industry. 

But how did I get started? It's a very typical journey, at the beginning at least, Sam. I did a degree in computer science and got to the United States to pursue my master's degree and got a little bit involved in data visualization and data analysis. Schlumberger approached me for a job to look at innovative ways of looking at oil and gas seismic data. Seismic data is very hard to visualize and even though I did not know it at that time, the key problem that Schlumberger was trying to solve was a collaboration problem. One person understanding something is interesting, but the whole department and the whole company has to collectively agree that the hydrocarbon resources are present and that they can be then further extracted. And that was a problem that I was asked to solve. It was very interesting. 

I was a software developer for a long time. But over time I started getting involved with business problems and trying to make the customer happy. One big area of need in the software space is the translation of the technology into the business problem space. Techies love to build technology and talk about it with all their three-letter acronyms, but sometimes it doesn't really make sense for the customer. As I started getting involved in that journey, I figured out two things. One, that I was quite good at it and, two, that there was a way in which more innovation could be brought to the table because I kind of understood both sides of the equation, so to speak. I really understood what problems customers were trying to solve and I really already had an innate understanding of the capabilities of the technology. So, I became what we called a product manager. One thing led to another, we ended up building a few innovative products, a couple of them were quite commercially successful. I got a patent for one of them. And eventually I ended up running Schlumberger’s digital program at large and built the oil and gas industry's first cloud-based platform that allowed for that kind of analysis. 

Right at that time, I got into a discussion with Anglo American. It's ironic, Sam, you set up that conversation to happen. And Tony O'Neill, [the group director of technical and sustainability],  and I hit it off, and from there the partnership with Anglo American began. The mining industry could benefit from these kinds of tools. They have a lot of software technology already, but what I was proposing to build and what we did end up building is a way of looking at it a little bit together. The integration and the collaboration pieces are what came to life.

Sam Burman: Casting my mind back six years ago, I remember when we had the initial conversations about the opportunity the Anglo American. The role was very much a facilitator to help drive the digital transformation agenda that Tony was looking to drive. And the chief data officer role was, at the time, much more geared around the data and analytic side of the value proposal versus AI. 

When did AI become part of the role? And we'd love to hear some examples, practical examples, of how you applied both data and AI to solve problems within Anglo, and then what net new successes that led to. 

Arun Narayanan: I love this question, Sam, because I think it really is the nub of where all these digital roles sit. AI is a tool; it's a fantastic tool and other analyses are possible. So, I think the main point for me is what type of technology or what type of solution you bring to the problem at hand. Some problems are begging for a solution like AI; for some other problems, other solutions are more apt. We started with the role and tried to understand what would make the biggest impact and what kind of investments Anglo American was trying to make, and a program began to emerge. And, in the way I like to run the program, we provided optionality to the leaders. We told them there were small, medium, and large versions of the program that they could pick from, let’s say. And they picked a medium version of the program and that's where we started. 

But, at this point, we're talking about solving business problems—we are not championing the fact that we're going to use AI or we are not going to use AI. And we really needed to make sure—and I feel that's the signature part of the role that I've done—we really need to make sure that the business problems that were being solved were relevant and that we could solve them. If we solve them, they would be able to be used and the usage would cause a positive business impact. That was the first round of the debate. When that settled, some solutions were very simple. You could build dashboards and present information to people. People are thirsty to just understand what's going on and they are intelligent and capable of doing the further analysis from that point on themselves. Maybe the analysis is not very complicated—or even if it is, they are capable of doing it themselves. You could say that's one type of solution. 

The other type of solution that we found a lot of success with is, let's say, first principles–based solutions, which means that if you know two or three variables in an equation, you can figure out the answer. Or, if you can't, then at least you can build what we call the digital twin and exercise the digital twin to understand your options. That worked out very well for us. One of our products that we built around operational planning uses the power of cloud computing and simulation to figure out the options for, let's say, a general manager on how to run their mine on a daily basis. And it's interesting because it forces a lot of questions for them. If I had one less truck, will I pollute less and produce less or will I almost be able to produce the same original target? You can begin to see that these kinds of questions now allow for sustainability outcomes to come to life. The energy intensity and water intensity can be reduced. It's possible to say that half the water can be consumed and 80% of our targets can be met—not that they would do that, but they can understand the trade-off points much more elegantly before any work has even happened. So, that's one type of answer, a simulation-based answer. 

Now, coming to the nub of your question around AI. Yes, there are many solutions for which there isn't a computational answer and, in these cases, the world has been without an answer until the evolution of technologies such as AI. What AI does is it allows us to estimate an answer, not very precisely, but at least close enough that good decisions can be made. So, one example over here is understanding the subsurface and understanding the definition of the ore body. This is work that's done by geologists; it's laborious and it's a lot of pattern-matching work. But when it's pattern-matching work, it lends itself to being automated by an artificially intelligent algorithm. That's what we did, and there was a dramatic, dramatic reduction in time. Now, the geologist is still important; they're important to build the solution, they're important to run the solution, and senior geologists are even more important because they have to accept the solution that the program has produced. But it's also possible to say, if you're being fair, that the number of geologists that are totally involved in the program in the future will be perhaps fewer than the number today, because some of their work is getting automated. 

And this is the double-edged sword that you get from digitization and automation. The level of the work increases, but the number of people that are needed can shrink. So, maybe that's the answer to your question, Sam, as to how we started out with a little bit of AI but really pivoted the program to business impact and then picked the correct technology solution to solve business and bad problems. 

Sam Burman: When we were talking about this recently, what I found interesting is that, like all organizations and leaders in your position, there is a combination of knowing how much to buy, how much to build, and how much to partner. And VOXEL, which I mentioned at the top of the conversation, is a software platform that I believe you built internally. I would love to hear a bit more about how you navigated the path of buy, build, and partner, because that's a really important topic, I think.

Arun Narayanan: So, VOXEL, it has a couple of expansions. The one that you mentioned earlier, the vision of operational excellence, is definitely one way to expand that. For me, it's the mining industry's first end-to-end digital transformation platform and you could say it's a bit optimal in many ways. It is a bit of a build, because Anglo built it, and it's a bit of a buy, because there are components in it, and there are partnerships in it as well. 

Now, let's start with how it looks, because at the end of the day when you're trying to get the mining industry to use complicated computer science–based software, it has to have an elegant experience. And all of the capabilities that VOXEL supports have a consistent look and feel and they have the look and feel of the Anglo American brand and the VOXEL flavors. So, from an end-user standpoint, it looks like one thing and it looks like Anglo American has built it. So let's park that. Now, behind the scenes there are intelligent components that the industry has built that would not make sense for us to go out there and rewrite. So, those are the pieces that we would end up buying. We try to put these in in an embedded manner, which means that they run behind the scenes; they take the problem and then they answer the question back, but the visual experience is still the VOXEL experience. So, that's one other way in which we can do it. And then let's talk about partnerships, because we're not going to go reinvent something like a cloud platform again. So, VOXEL runs on Microsoft Azure and we have very competent and effective partnerships with companies like Microsoft. And yes, there is a mix of general awareness and competency in the marketplace from a variety of consulting and software development houses that we have used to further accelerate the speed at which we have built. But we've really taken a look at differentiated IP and that's what we have sat down and written in our offices. We have looked at the industry’s capabilities and what is already available, and we have embedded that. And we have sort of tried to hold on to a single visual experience, so that the end users are not switching between different experiences. And that's how we managed to build that. And I'd say, from the totality of the vision, about 30% to 40% of the platform is already built out as of this point.

Sam Burman: We've been talking about the technical side of AI and data, but let's talk about leadership capabilities because this is the soft side that is equally important. As we both know, when you're driving a digital transformation, there needs to be significant sponsorship within the organization and beyond the individual and their team who are executing, and you need followership as well to get a successful transformation involving something like AI. What leadership capabilities have been most important to you, Arun, in developing both of those soft skills, and what challenges have you had to overcome? 

Arun Narayanan: The whole program is underpinned by that sponsorship element. We got some of it out of the box, because the leaders of Anglo American trusted us and wanted to use digital technologies to reimagine mining to improve people's lives, which is the purpose of the company itself. And they saw a pathway to get there, so some of the sponsorship was there from the get-go, but other sponsorship had to be earned. You could say the shortest version of the answer would be relationship and trust and building all of that step by step, slowly, one person at a time. Executive leaders could be convinced about the future of digital and there were many conversations had. Everybody has a slightly different lens. Some of them wanted to know the impact, some of them want to know the rate at which investment happens, and others wanted to sort of see how it would work for someone else before they would adopt it. So, you had the whole gamut of experiences. But the relationships have to be worked at least at the executive level. 

Now, I think technical partnerships are quite critical because the computer science departments that we have are not aware of the domain knowledge. We do not have subject matter expertise, we really don't understand geology or metallurgy, so we had to partner very closely with those departments to bring those skills into the team. And, for us to do that, again, the relationships have to be built, programs of work had to be defined, milestones had to be set, and we had to make sure that we were tracking and progressing through those milestones, building trust, building the relationships one at a time. 

And maybe last but not least, the same sort of—you used the word followership—the same concept had to be executed at the end-user level to make sure that there was excitement in the community in which we were going to deploy this, that they were wanting this and that they were ready for it. There were a couple of tools that we used. One, obviously, would be just the standard communication and relationship building, but the other thing we did that was very unique was run a digital literacy program that allowed them to step into the program without being fearful. It allowed them to learn about all these technologies without thinking that it was going to come and take their jobs away. In fact, their own improved skills allowed them to apply for or get higher level digital jobs, still in their space but move one or two notches higher up in their pay grade because they're skilled both digitally and in the original domain as well. The digital literacy program was received quite well. Seven or eight thousand people went through the basic training. Two or three people went to such an advanced level of the nano-degrees we provided that they effectively had come over from their old departments and were almost working in the data analytics team at that point. 

Sam Burman: Sticking with the theme of people, then let's talk about your team. While you were the head of the team, obviously you had to rely heavily on partnering with team members with all sorts of different skill sets. So, my question is around how you go about building a high-performance team when the talent is, should we say, more nascent or more scarce than perhaps other areas of the IT ecosystem?

Arun Narayanan: Yes, it's my full-time job. I love my team and I want to think my team loves me back, too. And it took a lot of effort to build that and you could say that the way I approached my job was almost toward the idea that if I built the right kind of team, the program would succeed. So that was really the main focus of my management style. Some of the concepts for me: one is collaboration, the team had to be one that couldn't be split. Not necessarily backstabbing but even divergence of opinions after a certain point. It's OK for people to have different viewpoints but they have to come together, agree as a team we're doing A, B, and C, and we're not doing the rest. We all had to be on the same page. And it took a while for us to get through the standard norming and forming stages of the team lifecycle to get to the performing stage. The other aspect, for me, especially when you look at the data analytics team in Anglo American, is that it's not an established team. We don't have established processes; we're sort of choosing new things to do. I’d almost argue that we are like a start-up inside of a large company; it creates a dual cultural basis. And what's really needed is people to solve the problems that are at hand, so we came up with concepts called an “I-shaped person” and a “T-shaped person.” Have you heard of that before? 

Sam Burman: Yes. 

Arun Narayanan: So, I-shaped people are people who are very capable or skilled in one thing and they typically don't care about what else is happening. In the beginning, we were looking for people who were T-shaped, who had an idea of everything and then brought their own knowledge in one part of that value chain. But, over time, I've progressed my thinking from T-shaped people and you can argue my current team right now is full of V-shaped people, where they have deep expertise in one area and decent knowledge of the adjacent area and less knowledge but still in the adjacency space. And because of their ability to play in the adjacencies, it's possible for them to take care of problems that their department may create in the adjacent space and think through the impact on other people. And, as they do that, the team becomes more robust and cohesive, and you could say that was the basis of how the team was built. Of course, we did look for really good talent, the organization was very supportive in us getting such high-quality talent. But, in my mind, having the cohesive nature is really what helped the team be a high performing team. 

Sam Burman: What is the significance of getting the culture right when it comes to driving a digital transformation like the one you've been driving in Anglo American?

Arun Narayanan: It's very important, Sam, because at one level we don't know what we're doing; we're trying experiments. So, for me, culture is responsible for making sure that the experiments succeed, right? At one level, you could say that the culture is a culture of a little bit of humility to understand the problem that we need to solve. If computer scientists are going to define the problem, it's possible they've picked a problem that's not relevant to the business. So how do you talk to the business, how do you have the humility to go talk to people and say, “Hey, I can do these things, but what is the relevant problem that we could solve for you?” I think that's one part of the culture. I think the other part of the culture is fear of failure and risk-taking. We don't know if these things are going to work. Even if you did all the Q&A at the beginning, we may not have all the data or the data may not be convergent enough to provide a relevant answer. So, I think there needs to be some confidence that we're going to try a few experiments; some are going to fail and some are going to work, and that is OK. I would say these are maybe the two broadest pieces of the culture that we have installed in the team. And we did get a lot of support from management to have that kind of culture where we were going to fail but fail fast and move quickly through the lifecycle of projects. 

Sam Burman: Apologies to any listeners who can hear some background noise. Arun and I are speaking in our offices and a very enthusiastic Zoom meeting has just started next door. 

Let me get to my next question, Arun. So, we've talked about leadership, sponsorship, high-performing teams, and now culture. You've been living and breathing this challenge of transformation over the last five or six years. What's your advice to leaders who are currently wrestling with their approach to both digital transformation and the sharp end of AI? 

Arun Narayanan: I think the first step would be to understand that they have to embrace innovation at the top. It's risky, and it's not necessarily going to work, so people who are able to provide the innovation need a mandate. It has to start from the top by saying, “I want to put this team together. I want you all to try these things. This is the scope, this is the budget, this is the time.” And also give some boundaries and framing for them to be able to proceed with their work. 

The other thing is to understand whether the executives have thought through what will happen if the innovation works, because it's one thing to say, “We tried to do something, and we solved it.” Is that just going to be a science project that you can check off? Something you have a patent on? Are you able to talk about it at a conference? Or have you as a leader really thought through new ways of working that are underpinned by the availability of these digital technologies? And, if you haven't thought about the new ways of working and you're not able to be bold enough to make those decisions, I think that's where you should start. Because that doesn't involve the technologists, that involves the board, that involves the executive leadership team, and understanding if they can answer questions like, Are we OK with remote work? What is the degree of automation this industry will take? Are we OK for decisions to be made on a piece of glass thousands of miles away? Are we losing out because we can't tap into the expert who was sitting in the other side of the planet? And if you have thought through new ways of working and you're able to sponsor innovation departments, I think you've gotten started. After that, it's a simpler journey. I think you need to find talent, give them the space to try out the experiments, provide them active and engaged feedback on what they're doing, and then I think you're on the way. One thing that I learned about at Anglo American was the democratization of things. So, with democratized data, people can go look up what they wanted to look up on dashboards. That was very empowering. And we also democratized AI. And what I mean by that is, rather than saying there are like 12 clever people and they are the only people who are going to be doing anything to do with machine learning or artificial intelligence, we actually found an AI toolkit, like a canvas where people can pull data in, ask different questions, and quickly build basic models that are quite intelligent. Anybody could do it. We deployed the license to people who did some of the early digital literacy work and that allowed them to become sort of citizen data scientists by themselves. So that then fosters a culture of innovation everywhere in your organization. So, if you're a leader trying to say. “I want to change the fabric of my company,” that's how I would approach it. 

Sam Burman: I think it's a really great point. Basically, what we're saying is that AI is so much more than the technology itself; it's everything around people, culture, transformations, and different ways of working. So that’s really interesting answer. Much of the mood music around AI at the moment, in my opinion, is quite negative, unfortunately. I'm very much an optimist on the application of AI, but where do you land in terms of this next frontier and its impact on organizations and talent?

Arun Narayanan: Yes, we are looking at, you could say, the scary end of AI, with generative AI being in the news. And some people are using it for obviously practical and good reasons. But other people are not, like high school students using it to avoid homework, or others using the image-based generative AI tools to produce completely unethical and random images. So, yes, it's a little bit weird. But I have a slightly different take on it. I feel that, at the end of the day, these are just tools and tools can be used for anything. 

First of all, the technology landscape itself will keep evolving. We’re already seeing emerging technologies appear around the bend that are beyond the generative AI tools, like the Vision Pro announcement from Apple. So that's going to promise that, just like you have people walking around with Air Pods in London, people will be wearing these ski glasses and walking around cities, right? So, technology changes will keep happening all the time, and they will manage to fall in line to practical industrial uses. That's what people like me do. Every company has a department like that, to bring technology in line so that it solves real problems. You know, the commonly used example would be to think about a calculation department like 50 or 60 years ago. You’d send them some math questions and they’d come back and tell you what the answer was. And then a calculator showed up. But that kind of technology did not shut down accounting, it just allowed accounting to move one level up in the hierarchy, right? So you're able to do more what-if scenarios to understand whether you should borrow more money or less money, how you can run another project, or how you can do something else, because the calculator lets you progress with the math much more quickly. And I think this is the bend that will happen with the current generation of AI tools. As it is, we're already seeing technologies like ChatGPT being offered in a safer manner by Microsoft. So, you can plan to run it inside your company firewall with just your company context, without having unauthorized users or without having the data leaking into the public general knowledge pool. So, I think people need to also understand there are terms and conditions that come with some of these technologies and as we sort of evolve, it will all stabilize. I'm more optimistic on this, on this front, Sam, compared to you. 

Sam Burman: My favorite anecdote is that with every technological revolution that has occurred, there's been more jobs that have come as a result of that, rather than fewer jobs. 

Arun Narayanan: I can totally see that, I believe it.

Sam Burman: My final question, Arun, is: keeping in mind what we've just spoke about, what are the leadership implications that come with these developments in AI and data and how can leaders adapt for this in the future? 

Arun Narayanan: I think that people have to be brave, that they have to understand things will change and that they have to tap into it by leaning into that space. I think it starts with some of the points that we spoke about, like, are you sure you know what problem you want to solve? Do you have the funding? Are you willing to go that far with the commitment to making it happen? Without that, I think everything will wobble in your team. Let's say your firm, in your mind, has a well-defined problem space and you have a well-funded team. I think, from there on, it goes as to how you give confidence to the team, how you ask them to not be bound to one technology but to look at blockchain or to look at the ability to solve problems using just standard computer or AI or some of the new generative AI tools—whatever is the right technology to solve the problems. And then a little bit around who is going to use it and how will it work? Like, did you think about the new ways of working? Are people confident they are literate in these technologies? And as they are literate and confident, are you able to keep track and measure and say, “Hey, we were supposed to have this kind of impact. A year later, are we having that kind of impact?” And it's a feedback loop. If technology A or B or solution A and B are working, let's double down on those. And if something else is not approved for the industry or it's a really hard problem to solve and cannot be solved, it’s time to put your energy toward some of the next frontiers of things. 

But it's a dramatically exciting space, Sam, when you look at video analytics and the ability to do generative AI and how so much acceleration work can happen because of pattern-matching. These kinds of technologies are just begging to be used in industrial use cases to reduce waste and reimagine how these industries work.

Sam Burman: It’s absolutely a fascinating moment in time and an exciting journey ahead, I hope. So, Arun, thank you for joining us today, for sharing your story, your insights, and your advice with our listeners. 

Arun Narayanan: Thank you very much, Sam, it's always a pleasure to see you.

Thanks for listening to The Heidrick & Struggles Leadership Podcast. To make sure you don’t miss more future-shaping ideas and conversations, please subscribe to our channel on the podcast app. And if you’re listening via LinkedIn, Twitter, or YouTube, why not share this with your connections? Until next time. 

About the interviewer

Sam Burman ( is a partner in Heidrick & Struggles’ London office and the global managing partner of the Artificial Intelligence, Crypto & Digital Assets, Cybersecurity, Health Tech, and Industrial specialty practices.

Stay connected

Stay connected to our expert insights, thought leadership, and event information.

Leadership Podcast

Explore the latest episodes of The Heidrick & Struggles Leadership Podcast