Do the math on generative AI (gen AI), and the numbers can seem heady. But according to recent research, only a small percentage of companies are beginning to deliver meaningfully on its promise—and many have succumbed to “death by a thousand pilots,” in the words of Rodney Zemmel, global leader of McKinsey Digital. In this episode of The McKinsey Podcast, Rodney and fellow senior partners and coauthors Eric Lamarre and Kate Smaje join global editorial director Lucia Rahilly to talk about their book Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI (Wiley, 2023)—which, a year and a gen AI disruption postpublication, continues to offer leaders a play-by-play guide for outcompeting in a fast-changing era.
In the second segment, senior partner Wesley Walden talks about the time he went to a client meeting instead of his son’s first concert. The client’s reaction taught him an important lesson about open communication.
This transcript has been edited for clarity and length.
The McKinsey Podcast is cohosted by Roberta Fusaro and Lucia Rahilly.
Aim high—but focus
Lucia Rahilly: The three of you have probably had hundreds of client conversations in the year since Rewired was published. Talk to us about the premise of the book and about what still resonates with the executives you talk to, vis-à-vis the challenges they face.
Eric Lamarre: The themes we hear repeatedly are, “How am I going to get significant value out of this? How should I be building my talent capabilities? What am I going to do with my core systems, and can I do this at the same time, or do I need to sequence things?” Also, “How do I make data easy to consume in this world of AI?”
Rodney Zemmel: Gen AI surprised us all when it came out in the form of ChatGPT. A lot of the book was written before gen AI. So for the first six months after the book came out, we had a lot of questions about how gen AI changes what the book says: what’s relevant, what’s not relevant, and so on.
What we found is that gen AI puts an exclamation point on many of the key messages of the book—in particular, the first chapter about a business-led technology road map. One of the challenges of gen AI is that it’s so easy to apply and so easy to fire up a pilot that you can get stuck in this “death by 1,000 pilots” approach and not actually get anything to scale as business impact.
One of the challenges of gen AI is that it’s so easy to apply and so easy to fire up a pilot that you can get stuck in this ‘death by 1,000 pilots’ approach.
This first section of the book talks about the need to pick a domain, pick the right problem, set a high aspiration for how you want to change that domain with a real business target, and focus there. Don’t do everything everywhere all at once across the organization. That’s been extremely relevant in this new gen AI world.
Has gen AI changed the game?
Lucia Rahilly: Do you see organizations beginning to capture value from gen AI? Or are we still too early?
Rodney Zemmel: We do, but it’s early days. The survey we did earlier this year found about 10 percent of companies have real value in their 2024 profit and loss from pursuing gen AI.1 That’s a relatively small percentage. But we think those that have captured that value in 2024 have been following the Rewired recipe that starts with this top-down, business-led road map.
Want to subscribe to The McKinsey Podcast?
Eric Lamarre: There have been a few areas where we’re seeing gen AI hit home runs. But we also need to differentiate the gen AI world. There’s a category of application that every company on the planet is going to use—for example, to summarize a videoconference. But that application isn’t going to be competitively differentiating; a company isn’t going to make more money because of that use case.
Then there’s the other category, where a company will invest to develop specific applications that will provide competitive advantage—for example, for better maintenance of their equipment, or for better support of their sales organization to grow sales. But these applications require hard work; they don’t come overnight.
Lucia Rahilly: Gen AI has dovetailed with the framework you laid out in the book. Anything you feel is evolving that might alter the Rewired framework?
Eric Lamarre: When we constructed Rewired, it was, by design, technology agnostic. What has happened with gen AI is the toolbox has gotten bigger, but how to use the toolbox—do I have the right talent to use the toolbox? The right operating model? The right data? These are the Rewired capabilities that allow an organization to win in this world of digital and AI—and all these capabilities are the same.
Explore COP29 with McKinsey
Kate Smaje: That’s the crux. It has made the struggles with the “how” all the more interesting, intriguing, and relevant. And the notion of a handbook, of a guide to how to use those different capabilities from the toolbox, is more relevant than ever because change can be hard.
Where you start matters
Lucia Rahilly: So many variables need to come together to rewire an organization successfully. You cover a range of examples in the book. Talk to us about a few that illustrate the types of investments and behaviors that seem to be working for companies.
Eric Lamarre: You have two different types of companies out there: those early in the journey and those that have been at this for a while.
Those early in the journey tend to ask, “Why don’t we invest in ourselves, as a management team, to learn what it means to lead an organization in the world of digital and AI?” and decide when and how to start. They then begin to consider, “How do I upskill my people, develop software applications, and drive a level of capabilities that goes beyond a handful of experts and specialists? How do I create the right data platform to exploit AI?”
More mature companies focus on an area where they feel they’re being slowed down and/or can’t scale effectively—for example, a big investment in data, making data easier to consume for scaling AI models, or in the operating model, ensuring business, tech, ops, and the control functions work together. It becomes a bigger surgery on the organization to unlock more value.
Kate Smaje: The organizations that can get some sort of human breakthrough, not just the technology breakthrough, tend to be those that are moving ahead faster. Our framework touches upon the notion of human breakthrough in different parts, whether from a talent perspective, a workflow redesign perspective, or strategic thought about the future of work, the roles, and so on.
The organizations that can get some sort of human breakthrough, not just the technology breakthrough, tend to be those that are moving ahead faster.
Rodney Zemmel: I’m going to pick an example that’s not in the book, and that’s McKinsey. We’ve had the benefit of having to take our own medicine over the past year. We chose to build our own AI capabilities rather than take something preexisting off the shelf.
We created Lilli, our gen AI platform, which we originally used for knowledge management but now use for many more things across our firm. We followed the playbook in a fairly rigorous way, from creating an agile operating model with the right delegated accountability to drive it to really thinking through how to create something amazing that combines our own proprietary data with the world’s public data in a safe way that doesn’t have a leak to the outside world.
In a way, this is going to encourage people to keep sharing and get data governance set up in the right way. We really thought about how to drive adoption and scale to an extent where we now have more than 70 percent of our firm using Lilli on a regular basis.
Lucia Rahilly: Any lessons there, Rodney, on the topic of human breakthrough—for example, on encouraging adoption or upskilling for implementation? Was it relatively easy for 70 percent of the firm to use Lilli? Or did we take special steps that other organizations might learn from?
Rodney Zemmel: It’s relatively easy to get to trial. People are curious, and they’ll try it once. But to get persistent usage, again, it’s back to the Rewired recipe of really thinking through the user journey and being front-line-centric on how it’s going to be used, so it’s not just a one-off but is embedded into how users do their work week to week.
Eric Lamarre: I’ll offer examples of two companies and the kind of investments they’re making.
The first is a large Japanese conglomerate that’s really feeling behind. It’s early in the AI journey, and the critical investments it’s making now have built up the capability of its senior management team, C-suite, and CEO. It’ll then pick a few areas, a few lighthouses, where it’s going to accelerate and build its confidence. There’s a step early on in Rewired about confidence building, and those are the investments it’s going to make.
Another example is a large New York bank that has been working on its digital transformation for years. It now has hundreds of agile teams innovating in the company. Its next move is, “How are we going to reorganize differently to really unleash quite a bit of innovation across the enterprise in a way that’s organized, not in a way that feels like everybody’s doing whatever they want?” So its theme is operating model. And that’s where it’s making the big investments to get to the next stage of rapid innovation in its business. That’s how I would contrast different investments, depending on the stage where you’re starting.
Lucia Rahilly: What kind of financial goals are appropriate to set, based on a company’s stage of investment?
Eric Lamarre: In the Rewired book, we write that if your digital transformation or AI transformation isn’t delivering 20 percent or more improvement in your EBITDA, you’re probably not creating ambitious enough road maps. And that line, in our different conversations with companies, has triggered a simple requestioning.
When companies take stock of what they’re working on, they usually find that digital transformation doesn’t make a difference to the EBITDA line. They’re seeing a bunch of pilots and a bunch of little experimentations. And then when the C-suite takes stock, they go, “You know what? We’re completely failing that test. We’re not even in the ballpark.” Other companies are actually pretty happy. They’re getting into that sweet spot, that range, and then execution becomes the next big question.
But having a quantum to test yourself against is important. I’m not saying 20 percent is the right number. Maybe it’s 15 or 25 percent. It doesn’t matter what the exact number is, but we should keep ourselves honest that senior management can’t spend time chasing stuff that’s not going to move the bottom line by 20 percent or more in the next three years. So if the digital transformation you have today doesn’t meet that bar, maybe it’s time to reset your ambition.
What are leaders doing right—and wrong?
Lucia Rahilly: What else might others learn from companies that are embarking on these pilots successfully?
Kate Smaje: The sort of delta, or movement, that I’ve seen is with those that started out by looking at AI, in particular, as a risk. “This is a risk to our business model. This is a risk to our customer data, et cetera.” And at the start of some of those journeys, my humble opinion was, “This is going to go nowhere. They’re going to tie themselves in knots at all the possible things that could go wrong.”
But what came as a nice surprise for me, through at least a couple of these, was the pivot to being conscious of the risks but not constrained by them. Organizations that aren’t looking at AI as creating all these risks but rather saying, “This is now giving us, as a leadership team, a platform to discuss the risks, understand them, figure out where our posture is on them, and really have a way forward about what responsible AI means to me.” I’ve seen a number of organizations move, night and day, and people that I thought were going to get nowhere have really unlocked transformation because of a richer—and a deeper—conversation about risk.
Eric Lamarre: One that really surprised me is the use of gen AI to modernize legacy platforms. You have an estate of software applications developed 30, 40 years ago in COBOL [Common Business Oriented Language], and they’re still running today—largely in banks, but also in other industries. It’s been a nightmare to try to modernize this old software estate. Now there are gen AI possibilities that I’ve seen work at one of my bank clients to modernize that old COBOL code into a modern Python code. It’s still not 100 percent automated, but it’s at least 50 percent automated today; all of this can be 50 percent done at the click of a button. I didn’t think that was possible.
Kate Smaje: What I love about that example, Eric, is an organization that said, “Here are my universal truths. Here are the underlying assumptions of what I can do, what I can’t do, and where the ceiling is on certain things.” And then it used technology to break them.
It’s really interesting to see organizations consider their jigsaw pieces: “If I move that a little bit to the left, move it up or down, then what can technology do to change where that piece sits?”
Lucia Rahilly: Are there any use cases of gen AI that, at the outset of this process, when gen AI sort of exploded onto the scene, you thought, “OK, this is going to be a slam dunk,” but then have been far more complex to implement or fallen short in terms of results? And if so, what can others learn from those?
Kate Smaje: For me, this may be a controversial one. I’ve been surprised at how hard it has been to deliver the efficiency benefit. I see a lot of organizations do the math, and the math is relatively simple. You can get to some pretty heady numbers about the level of productivity improvement technology can unlock. But realizing that is still super hard. It’s hard because it tends to occur task by task. And it’s hard because we all, frankly, have long backlogs that we don’t get to during the workdays. And it’s hard because the value map of really understanding where and how it’s achieved is still a bit fuzzy in most organizations.
Eric Lamarre: Kate is so right. We fall in love with the potential and the big numbers, and we forget how hard it is to get it done. I’ll give you an example. I’m excited about the potential of gen AI to automate contact centers. But it doesn’t happen with, “OK, let’s load up the gen AI software. Everybody can go home. It’s done.”
Let’s go through the steps. We’ve got to figure out, what are the types of calls we could automate? A change of address—that’s perfect for automation. Getting a change in your mortgage terms and negotiating new fees maybe can’t go through gen AI. And then, for the ones we’re routing to gen AI, how do we manage risks and potential consequences? There’s a whole bunch of work that needs to happen on that side as well. Once all that is in place, then for the agents still handling calls, how do we upskill them to receive different types of calls? That’s hard work.
Rodney Zemmel: There’s no question that it’s proven harder than the hype to capture value. You can trace it back to one of the failure modes we talk about in Rewired. You’re starting on the wrong problem, or you’re doing it with the wrong level of aspiration, or you’re doing it with the wrong operating model.
What’s next
Lucia Rahilly: I hear you saying that the framework mapped out in this book is timeless and that it continues to resonate in important ways with your clients and with leaders of a variety of organizations. Looking ahead, what do you see surfacing, either as challenges or as opportunities, in the next horizon for tech-enabled transformations?
Kate Smaje: The next horizon for technology is how you build for scale. How, at the start of doing something, you look at it and say, “What happens when we’re wildly successful around this? What then?” And I think that will hold true for how we build the technology. It will hold true for how we rewire the organization. It will hold true for how we raise the bar around value creation potential. But for me, there’s a big theme in here about scaling.
Eric Lamarre: Scaling is the name of the game. We’re nowhere near the finish line because we’re now entering the phase of, how do we build organizations that can scale digital and AI innovation across all their processes? The next ten years will be about a product-and-platform operating model that allows many agile teams to innovate in a way that’s coherent.
The next ten years will be about a product-and-platform operating model that allows many agile teams to innovate in a way that’s coherent.
The next ten years will be about something Kate and Rodney coined in an article from earlier this year,2 which I love: IT as a service. IT will become easy to consume by the rest of the organization, so it can also drive its own innovation.
Rodney Zemmel: I think it’s about the evolution of leadership. We’re at a turning point where, in the future—maybe this is in five years, maybe a little more or maybe a little less—there aren’t going to be business leaders and technology leaders. Every business leader is going to need to be a technology leader. It’s an “and,” not an “or.”
And I think where the book has struck a chord is it’s given business leaders a manual to understand what it means to be a technology leader. And it’s given technology leaders and organizations a way of communicating with the business leadership to say, “Look, this is what needs to happen about really driving impact.” Of course, it’s about technology, but it’s never just tech. It’s this six-part recipe that we think is needed to really make a difference.
Human to human
Roberta Fusaro: Next up, senior partner Wesley Walden talks about a client’s reaction when he prioritized work instead of his family.
Wesley Walden: It was one of the first times I ever worked with this particular CEO, and I flew about five hours across the country to be there for this big meeting on a Friday afternoon. I was very nervous. I sat in his office, and partway through our conversation, my phone buzzed with a message. A picture of my six-year-old son at his first school concert came up. The CEO saw it, and he said to me, “Oh what’s that?” I said, “It’s my son’s first school concert. And my wife just sent me a picture.”
He looked at me in complete shock and said, “What are you doing here?” I said, “I wanted to make sure I was here in person for this meeting because it’s an important conversation.” And he said, “You're crazy. If my daughter or my son had a school concert this afternoon, there’s no way I would be here in this meeting with you.”
It really taught me a great lesson about balancing family and work priorities and about the fact that our clients are real people. They have families, and they’re all balancing their busy jobs with their family priorities as well. It also encouraged me to have open conversations with my clients about whether or not I have a major family event or a major personal event that I needed to prioritize differently.
In fact, it makes you more human and relatable as a person when you share how you think about your priorities.
