Bill Kerr: AI is fundamentally changing the nature of the firm. To contend with digital native companies like Amazon businesses need to re-architect. The order of the day AI and industrial grade analytics wired into unified data models. It’s a stiff challenge requiring unprecedented levels of coordination. Welcome to the Managing the Future of Work podcast from Harvard Business School. I’m your host, Bill Kerr. I’m joined today by Harvard Business School professor, Marco Iansiti, coauthor with fellow HBS faculty member Karim Lakhani, of a new book Competing in the Age of AI. In place of highly specialized firms the future may hold more companies built around decision factories. The performance gaps between firms in the same industry segment may grow substantially. We may even get that chatbot to answer our real question. Marco’s going to describe the new industrial landscape and explain how businesses, policymakers and workers can navigate it. Welcome Marco.

Marco Iansiti: Thank you. It’s great to be here.

Kerr: Marco, what prompted you and Karim to write Competing in the Age of AI, what’s the audience you’re trying to reach?

Iansiti: Literally, over the last seven or eight years we’ve been working on various aspects of digital transformation.

Kerr: You lead the initiative at HBS, the digital initiative.

Iansiti: I lead the digital initiative. We started the lab to do a bunch of projects related to digital transformation, AI and the evolution of organizations around it. But the moment was really realizing that even actually relatively simple artificial intelligence applied within the operating side of the firm can really change the way the firm works in a fundamental way. We have a model that’s hundreds of years old that makes managers manage complex things, complex processes, divides up firms into a bunch of relatively autonomous units that are industrialized and do various things to kind of fit the market needs. You have generations of IT that sort of went into these individual firm units and so you have IT aimed at making customer relationship management work better. You have IT that works with the financial management of the firm and so on and so forth. There’s so much information technology around that it’s starting to fundamentally change the structure of the firm—the architecture of the economy around it—and how the firm works. So for the first time you have really business processes that are entirely enabled by software where if you look at the way that Airbnb for example matches homeowners and renters, there’s no human being that does that at all. There’s nobody that’s the bottleneck in that process. It’s fundamentally a piece of software. It’s an algorithm that figures out who is best matched with whom. And it completely changes the way that the process works. But the minute you actually enable it with an algorithm instead of a human being, it changes everything, because unlike a human organization that is really hard to scale, software-centered organizations—algorithmic processes—get better and better as they scale to a larger number of users.

Kerr: You started by talking about all the things that IT was traditionally doing in the organization, but that now we may be at the re-architecting of the company. And as you think about the technology that’s come in the past, as we developed the mainframe and then the PC and so forth, did we have similar kind of re-architecting at that point? And then as we come to today, is it about the AI and the algorithms and the digital side? Or is it just that there’s this cumulative mass of IT that’s now been achieved and put into place that we’re seeing the outcomes of that?

Iansiti: So, if you look at traditional organizations, they’ve been working in a relatively similar structure for many, many years. But as traditional industries start to become more and more software based and more data oriented and more AI centric, all of these things are sort of part of the same transformation. And then all of a sudden these very traditional industries are faced with, my God, this is an incredibly deep transformation.

Kerr: And you think about where that software is being embedded. Is it in the product or is in the operations or is it a bit of both? So, thinking about the software that goes into the car. I can think about a re-architecting operating model for General Motors—it’s just develop any car—and probably some interface between those two.

Iansiti: The software begins to connect different processes withi the firm. More and more, the firm itself has changed and that’s the essentially the sort of major transition that we’re going through right now.

Kerr: You use a term in your book that I find fascinating, “AI factory.” Can you tell us, what is in an “AI factory” and what are the components that I need and what’s it like to operate an AI factory?

Iansiti: I think what we’re seeing with data becoming a foundation for how you run things is that you have to change the model, industrialize the way that you actually assemble process, normalize data, how you clean it up. And it’s an incredibly important thing, because all of a sudden that is the thread that is shaping everything that the firm does. And any mistake that you’re making doing that or any differences between how you might do it for one business line and something else will have huge repercussions on how well the business works. You know, you can get nasty algorithmic biases in facial recognition and things like this. And, so it’s really important to think about the factory-like approach that you think about in terms of integrating, assembling, interpreting, analyzing data so that you can scale through large numbers, huge datasets, and diverse datasets that are integrated properly.

Kerr: It’s actually I think an American innovation, interchangeable parts.

Iansiti: Yes.

Kerr: And do you see a similar type of interchangeable parts in the data world? Or is it more about each individual company to come to a unified data platform?

Iansiti: When the internet exploded on the scene during the 90s and everybody started to use it. All of a sudden you can see a glimpse of the fact that you can take information from one place, transmit it somewhere else and have a perfect replica of that information, all kinds of different applications. And it’s a very, very powerful thing.

Kerr: You’re really used to framing that bottlenecks as places where corporations struggle to get beyond a certain skies, or scope or something. Walk us through what happens when those bottlenecks are relaxed, and then how does the organization as a whole build around that?

Iansiti: I teach operations management and so we have this whole tradition of talking about bottlenecks. We’re talking about National Cranberries and figuring out what the bottlenecks are in cranberry production. Because when one tries to manufacture a car or deliver your perfect latte at Starbucks, there’s always some resource that is providing the constraint on how quickly and how efficiently you can actually do that. Now, when you begin to digitize the operation, right? Think about like what is the bottleneck in Google search? Or what is the bottleneck in Amazon setting a price. So, what is the bottleneck in financial granting alone? And you realize that traditional bottlenecks are removed. There’s no organizational constraint. The only constraint essentially is computing power.

Kerr: As you’re highlighting through the cloud and through other mechanisms. Even if you’re not a direct computing provider, you can access that.

Iansiti: You can access that through the cloud. And it’s very powerful. And that’s a really great thing if you want to grow fast. It’s also a big challenge because it means that…

Kerr: And many of our startup companies are coming out of the school, are using AWS or some similar platform but they also try to have that unlimited upside. If the scaling process starts, they can grow and it’s even all at a variable cost rate. It’s not a heavy fixed cost in advance of it.

Iansiti: You start to see an organization that can scale to massive numbers of users with a very, very small number of employees.

Kerr: Would it be a reasonable litmus test then to think about if I want to make a bucket of the AI kind of ready companies or the digitally native companies or whatever and those in part to use this presence of bottlenecks as something that kind of helps me separate the two piles?

Iansiti: I think it does. I think it’s who is driving whom in other words. So that in a traditional organization you have human-centered processes that drive the work. Even if a lot of it is automated, there’s always a manager somewhere doing things, influencing how the work gets done. If you go into an Amazon warehouse, you still have humans in the organization, but you have algorithms that are specifying the optimal routing that some person needs to go through in order to pick out a specific part, or what it should take. And for better and for worse, that algorithm is fundamentally managing the human, and at whatever scale you want. And, so you’re kind of marginalizing the human labor in some fashion and then adding sort of designers and managers and data scientists around it, but not in the critical path of the work actually being done.

Kerr: Yep, and do you think that the example in the warehouse, is that an interim step to a fully automated warehouse? So, we’re going to use that person for a while and the software is directing it, but ultimately, we want to install autonomous vehicles or something that accomplish that task. Whereas the designers and others would still be necessary or needed.

Iansiti: If you look at, for example, a company called Ocado that does grocery delivery in the UK and has been licensing its technology to a number of different organizations in the US and Australia and elsewhere. Ocado is almost entirely automated within its warehouses.

Kerr: Other than putting the data first and trying to think through the algorithms, how else does managing an AI-centric company, you know, take root? Like what are the different learning processes, like what else changes about that organization?

Iansiti: In some ways the whole architecture of the organization needs to change. So, if data is the core of what it is that you do, that data that you might have about a certain consumer is useful to almost every department, or every use case, every function. And so designing that properly means that you kind of have to turn the organization around and try to break down the traditional silos and build at least a common structure underneath that. That, of course as you can imagine, leads to all kinds of other challenges—from sort of organizational challenges, culture challenges, leadership challenges. There’s a shift in the skill sets that are required, where you see some workers that are in some ways doing less skillful parts that robots are just not equipped to do at this point in time. And then there’s a bunch of other workers that are more in the design stage, that are out there trying to figure out how to get these systems deployed and understand them. So the IT around this looks very different, and the management is real different, right? It’s something we also have to think about. And it’s part of the MBA and the executive programs that what managers manage is increasingly technology. We still manage people, but it’s a smaller number of people, typically, in these kinds of organizations, and requires much greater familiarities with data analytics and all that stuff.

Kerr: I’ve heard an expression in this context that the fastest learner wins and that’s often then used as a follower, or as a justification for “I’ve got to get big fast. I got to be the first mover” and so forth. Does that align with your experiences for trying to manage and build one of these AI first digital native companies?

Iansiti: So as long as you don’t just learn once, but to keep on learning and adapting and changing along the way, I think that’s fair enough. I think one of the things that we’re seeing with a lot of the digital technology going into organizations these days, is that in some ways traditional models of competition are not really working anymore. In the sense, that thinking of your own space as limited in your own industry, and your competitors as being the old competitors, doesn’t really work anymore. You know the old industry analysis stuff that we used to teach is not really effective anymore because so many threats are coming out of spaces that are connected to your industry through digital network, but maybe completely outside.

Kerr: I think now everyone’s scared to death of Amazon because Amazon appears to know no industry boundary that it can’t cross along the way. Do you envision that being evermore pervasive in the future, or are we getting to some kind of new normal and from here on we’ll have redefined some industry boundaries?

Iansiti: I think that especially on the consumer facing industries, we’re seeing a transition that I think is in many ways irreversible, which is based around the fact that data about consumer preferences, your ability to personalize what you provide to a certain user is useful across a broad variety of different services. So, from a consumer side, I see quite a bit of a sort of a certain number of companies that increasingly cutting across traditional sectors, and it doesn’t have to be Amazon, right? Telecommunications firms, for example, is traditional players. I’ve seen very similar kinds of changes in the sense that they’re seeing some of their core businesses threatened, but they also have opportunities to sell financial services, to sell [the “internet of things”] IOT.

Kerr: Move into content creation.

Iansiti: Move into content creation, right. So we are seeing firms that are going in some ways away from competition within a well-defined industry, and much more competition in a network of connected spaces, or connected domains, that…where I can take the assets that I have and the capabilities that I have in the data that I have and use it as well as I can across as many different things as I can possibly manage.

Kerr: As I think about the industry landscape and these digital natives, then also Wall Street and everything else, you kind of come into some that wow, they seem to earn an amazing amount of profit. And then others go public but then really struggle with losses, or even billed to be something and then it doesn’t achieve that kind of promise. If I gave you a basket of these digitally native goods, you know, is the best basket to invest in? Or are we under appreciating the incumbents?

Iansiti: As you get to more digital models, your ability to make money is almost all or nothing, right? That in some sectors it’s incredibly competitive, and will always remain competitive. And in some other sectors, it’s winner take all, winner take most. In traditional industries usually has some sort of a oligopoly, have a few firms competing. They have capabilities that are a little bit differentiated, like the old sort of cell phones. Right. In the old time of the old Nokia Motorola battles and we had seven or eight players that were, you know, all had, you know, 5 percent to 12 percent market share, or something like that. They were competing on designs and unique capabilities. As you switch from that to much more of a platform era, where you have heavily digital providers like Google with Android, then that industry structure has shifted, and all of a sudden you have one player in Android and another player in Apple, which you can argue in is in similar or different industry, that is extracting all of the profits and everybody else essentially is at zero profitability. And so we’ve seen that kind of pattern in a few different industries where either you can actually extract these extraordinary values because you have strong network effects, so you have strong advantages from data. Or in other environments the advantages are relatively weak, as for example with ride sharing. And so that is a zero profit environment.

Kerr: Yeah, and so those examples were, especially the Apple or with two companies we consider us to probably be on the digitally native side. Take us to like an AI firm is entering into a space with non-AI, or not yet as AI as they would like to be, firms. You used the phrase “collision” in your book. So, tell us what is the collision happen, and is this disruption and that we can pretty much bet on who’s going to win here, or does the incumbent sometimes prevail?

Iansiti: In “collision” we mean something that is not disruption, because you have a kind of firm, say Marriott—that has competed in a certain environment for a hundred years, offering hotel rooms to business travelers and to vacation travelers—that is facing and colliding with a fundamentally different type of company in the Airbnb. Airbnb in a few years has accumulated inventory, or three times as many rooms as Marriott offers, but does so in a fundamentally different fashion. If you look inside the firm, they’re built in fundamentally different ways. Marriott’s got a large organization, lots of management, lots of people everywhere thinking about sort of how can they optimize individual parts of the business. Airbnb is a big data platform that accumulates data about the users, accumulates data about the homeowners, and does what best job it can do to match the two together to optimize the work, and essentially the human organization is pushed out into the homeowners. They all have responsibility for serving up the service. And so, you can kind of see—you have two fundamentally different beasts. I mean, these firms have nothing in common in some ways except for the fact that they’re competing for the same user. And that’s a collision, right? In disruption it’s much more like you have one firm that has a business. That business is being threatened by some kind of technological change, and somebody else’s coming along to threaten it. Here it’s not a technological change that’s threatening the firm. It’s a fundamentally different kind of company.

Kerr: Yeah, and you go back to the Marriott example. One thing that of course that can give them some comfort is that if I were to send you on an HBS trip to Cincinnati, you’re probably not going to use Airbnb. But so my next question would be, in 2030 will Airbnb and Marriott look a lot more like each other than they do today?

Iansiti: I don’t think they’ll look like each other. I think they’ll go increasingly at the same users, and in the same use cases around that. With collision, you essentially have to re-architect the firm. Satya Nadella talks about rethinking the core…

Kerr: That’s the Microsoft CEO.

Iansiti: Microsoft CEO Satya Nadella. He talks about changing the core. But to give you a sense, I mean Fidelity’s an interesting example. Here’s an organization, Fidelity Investments, traditional financial investment company. And they had been investing a lot in artificial intelligence, building on many years of experience around data and analytics. And so really, they’d been at it for 10 years in some fashion. Building these assets and capabilities part of this long journey. And they have a bunch of new competitors in fintech and new, different options for trading and so on that are coming after them. And so they’re threatened in a very collision-like fashion. But they’ve taken their time and they’ve invested systematically over the years, and they’re in a much better position now.

Kerr: I think part of what you’re highlighting here is that it’s not that you want to just layer some AI on top of whatever you’re doing right now. You need to rethink how you’re going to do the business, and then build it around that technology set.

Iansiti: Absolutely. There’s a phase that one has to go through, which really is much more playing with the AI, doing a bunch of pilots, understanding what the potential is. But then you need to get to a stage where you begin to think about how that AI is going to serve to integrate things that you do across traditional businesses, across different kinds of things, and you begin to build this horizontal structure that becomes, in some ways the new nervous system of the firm. Look at how Microsoft manages its internal processes—not the sexy, new product stuff, but really how to manage the supply chains or how he manages the financial connections between different organizations and things like that. Integrating the data that used to be disparate and fragmented across many different silos.

Kerr: Do you have examples of companies that successfully do more of a big bang approach? Or do you recommend against the big bang approach?

Iansiti: The reality is the process unfolds over years. Right? I mean, Disney made a huge commitment to do this. Comcast, 10 years ago, actually made a huge commitment to do some serious digital transformation, create a platform called X1, and it was a huge bang in the beginning. But it’s still unfolding 10 years later. The challenging thing with this, is that in order to rethink how the organization really works, the CEO really needs to make a real commitment. I was just talking on the phone, literally this morning, with someone who leads a $2 billion manufacturing company, manufacturing glass for the auto industry. An amazing organization. And you look at all the different initiatives that they had going on in digital transformation, probably 15 different major things. And the big challenge is not the individual pieces, but it’s really connecting them all with this clear vision.

Kerr: Yeah. Any best practices for, if you’re the CEO, how you think that future state through?

Iansiti: I think there really needs to be, first of all, really one strategy around this is, "What we’re going to do, it might take a while, this is where we’re going." Secondly, that strategy needs to have an actual grounding in architecture. So, you need to actually begin to paint the blueprint of the future organization. Figure out how you’re going to get there. That’s partly organization, partly it’s technology. It’s like, "What are we going to do? How much is going to go to the Cloud? What kind of data are we going to integrate? What are some of the clear primary objectives that is going to put meat on the bones of the actual strategy statement?" Third piece, you’re going to have to start incubating this thing somehow. You’ve got to plot the trajectory of how that’s going to unfold, and where are you going to go first, and how are you going to expand from there.

Iansiti: And the final thing is almost around portfolio and governance. So, as you get to much more software-centric model and data-centric model, there’s a lot of things that change. And there are some real process that need to be put around how are you going to govern and all those changes, for two reasons. One of them is the number of different simultaneous things that a large company is going to be involved with. It’s just huge. You’re going to have hundreds of initiatives, if not thousands, eventually that go through this. And so there has to be a really clear process for how these things connect. And then secondly, governance also means understand the risk in all of this. And this is where, I think, people spend less time really absorbing what their repercussions are going to be. When you’re data-centric business, that data has huge value, but it’s also associated with huge risks. And the number of people that have heard, "Oh, we have all this data, it’s great. We keep it all. We don’t know what we’re going to do with it." Right? "But it’s all there, in case we ever want to do something with it." And that’s the worst possible scenario, because then you expose yourself to all the risks, and you get none of the value. And so really understanding how all the projects are connected, where the data is, making a catalog of the data, figuring out how you’re going to protect it, make sure you have the right privacy strategy around this, all this governance, is really the fourth best practice thing that I can say. I’ve tried to convince people to think of that.

Kerr: A lot of our concern on this podcast, and people we speak with, is about making sure that the future looks brighter for the worker than some scenarios could hold. Are you optimistic about the worker of the future? Will we have jobs for them? Will they be meaningful jobs? Will they be subservient to some software program that’s directing them to do this and that, and they’ve become more mundane tasks? How do you envision that future?

Iansiti: Yeah, I know. It’s a great question. The evidence that I’ve seen—both things that I have researched myself as well as evidence that I’ve seen, for example, from Eric Brynjolfsson’s group at MIT—shows that basically every job is going to change. Some occupations will be more affected than others, but some elements of every job will be affected and will be transformed. There will be more machine-learning and artificial intelligence enabling certain aspects of what every one of us does. The second piece, obviously, is that there’s going to be a lot of changes in what people need to know in order to do various jobs. Right? So as these jobs change themselves, there’s going to be retraining, there’s going to be transformation needed, there’s going to be dislocations. And I think that the argument that people sometime make around these changes not being all that major because, in the past, workers have always been reabsorbed. What is happening now is that you have dislocation across virtually every sector happening almost simultaneously. Digital transformations are sweeping across every industry almost simultaneously. And that’s where I think it gets a little bit more a subject of concern.

Kerr: We talked about all the ways the roles and workers need to change. You’re talking to the CEO—maybe of the glass manufacturer you just mentioned—and the conversation is something like, “Yes, I know we need to look like that in the future, but my workforce has this set of skills right now. What’s the pathway? Is there a way that we can get the workforce of my firm from here to there, or is it one where I have to go out on the open market and look for talent or some other mechanism to do that?”

Iansiti: There’s definitely an element of both. There’s huge demand for a generation of workers with deep expertise in data analytics, software engineering. The retraining question is an important question, and one I think we should invest as much as we possibly can in, and it’s clearly one for us to actually drive an enormous amount of change and help people through the transformation.

Kerr: There’s lots of conversations right now about breaking up certain big tech firms and that inequality is skyrocketing on the firm level due to these digitally native companies. How do you think about that future policy environment?

Iansiti: I think regulation of technology companies over the next 10-, 20-, 30 years is going to be an enormously important and super interesting topic. I also think that simple solutions are often not great solutions. I’ve worked on behalf of the European Commission on some cases that they have gone through. The appropriate remedies for having abusive market power, if you like, in a technology-centric company—which now describes almost any company, right?—is a subtle thing, because you’re going to have to have new regulatory guidelines and guideposts, because these firms look and behave very differently than traditional firms might have. For example, breaking up a network-effects intensive business doesn’t really make a lot of sense. You can break it up into two, but the networks are going to compete and one is going to win, and then you’re going to be in the same spot again. The kinds of remedies that are useful are going to really about figuring out, fundamentally, how you want these organizations to behave, and what are the fundamental values and fundamental principles that they’re going to need to observe in certain particularly important industries and certain categories. For example, news is a good one that comes to mind, right? Also, I don’t think the regulatory agency is going to be the entire picture, right? We need some quick-response ways of driving regulation. One thing that I’m excited about is the whole impact of communities, on…one could think of almost as an active regulator. If you think about the open source community, they’ve done so much in terms of creating alternative technology, alternative platforms, organizations that run themselves, police themselves, calling out when something is wrong very quickly. Some combination of rethinking traditional regulatory agencies and then really driving communities in the open source model to go off and improve the kind of inequality and the kind of challenges, abuses, if you want to call them, in the sector, I think will be the way to go.

Kerr: Marco, I’ve got one final question. If you were in a room with a junior in college, is there one kind of extra bit of advice that you would provide her about the future?

Iansiti: Understand and play within the environment that is unfolding around you, right? In some ways it’s almost the same advice that I would give a CEO and a junior in college. Play your cards in a way that’s in sync with what the trajectory is telling you to. The digital transformation is here to stay, AI is here to stay. You may not like it. There are a lot of challenges that it brings with it, along with some opportunities.

Kerr: Marco Iansiti and Karim Lakhani have the new book Competing in the Age of AI. We thank Marco for joining us today.

Iansiti: Thank you very much. It’s been a pleasure.

Kerr: Thank you for listening to this special episode of the Managing the Future of Work podcast. To find out more about our project on the future of work and for more information on the coronavirus’s impact, visit our website at hbs.edu/managing-the-future-of-work and sign up for our newsletter.