The real AI advantage

The first wave of AI adoption has been all about efficiency gains. But what happens when everyone else, using the same tools, is experiencing the same productivity improvements? Where will competitive advantage come from then? McKinsey research suggests that the companies that outperform with AI will be those that rethink customer experiences, redesign how work gets done, and build operating models that continuously adapt to change.

In this episode of The McKinsey Podcast, McKinsey Senior Partner and McKinsey Global Institute Director Tanguy Catlin explains to Editorial Director Roberta Fusaro why CEOs should shift their focus from AI-driven productivity to AI-powered reinvention. They discuss the rise of agentic AI, how competitive dynamics are changing, and why organizational transformation—not technology—is likely to be the defining challenge of the AI era.

The McKinsey Podcast is cohosted by Lucia Rahilly and Roberta Fusaro.

The following transcript has been edited for clarity and length.

Beyond the productivity trap

Roberta Fusaro: In several recent McKinsey articles, Tanguy, you and your coauthors have been exploring the “second-day story” around AI: Once people are using it, how do they get value from it? Despite all these massive investments in AI, most companies still aren’t getting the returns they expect. Why do you think so many leaders still equate AI success with productivity, when the real value might lie somewhere else?

Tanguy Catlin: This is a very common question. Whenever we have seen the advent of general-purpose technology, which is technology that fundamentally affects the nature of competition in industries and redefines how work gets performed (for example, electricity or mobile phones), the first reflex from a leader is to try and extract efficiencies from that technology. This means they would be doing the same work they were doing before, only better, faster, and cheaper.

These technologies are ubiquitous. The challenge CEOs face is that their competitors have access to the same technology and pursue the same productivity gains. As a result, there is very little general value that accrues to any specific company. A lot of the surplus value created by these technologies is typically passed to customers or the service providers needed to implement the technology for productivity improvements.

I think the natural instinct is to say, “I have a new technology. Let me deploy it on top of my existing processes.” Instead, you should pause and ask, “How do I redesign my process and what I offer to the customers?” When that happens, we unleash exceptional levels of value. I suspect that AI will similarly unlock a significant level of innovation. What those innovations will look like remains to be seen.

Roberta Fusaro: A lot of industries are built on friction. For example, customers have to compare options and figure out which is best for them. AI is taking away some of that friction, and information is more transparent and quicker and easier to access. How can companies capitalize on this new foundation?

Tanguy Catlin: The first step is to look at what your customer wants and identify the sources of friction that either lower the customer experience or add unnecessary cost. Then, presume that someone, somewhere, will use the technology to address those concerns. Ask yourself if you will benefit from that change. If your answer is no, you’d better be the disruptor.

Roberta Fusaro: One point I found particularly interesting in our latest AI research is how we’re thinking about AI and supply chains and how AI can lower transaction coordination costs. What new business models could that unlock?

Tanguy Catlin: It’s a difficult prediction to make, but I’ll give you a few hypotheses.

Right now, we’re observing the rise of what we describe as “agentic commerce,” which means consumers have access to AI tools to help them resolve problems and access the solutions they need. Those platforms, like ChatGPT, are becoming new platforms for commerce. For example, I can ask the AI questions about what attributes I should look for in an insurance company I want to purchase coverage from, as well as what type of coverage I need. If the technology has memory and context about me as a user, it will be able to draw from our prior conversations and customize recommendations to a level that was not available before. Furthermore, if that platform is connected to a service provider, I can now use it to transact, get a quote for my insurance, and maybe purchase a policy.

On top of that, we are now entering an agentic world where we can create agents that can transact on our behalf. I can go into a model and ask an agent to track when my policy is up for renewal, to independently look at the marketplace for a better policy for my needs, and to automatically switch on my behalf.

One thing I believe will happen is that organizations will try to embed their product and service experience into their platforms, as well as into broader offerings, to try to lock in customers and avoid the risk of churn. I think you will see industries come together into one ecosystem because the friction cost to navigate across industries will be reduced.

Roberta Fusaro: A good example of this might be buying a home. You have to find the house, get a mortgage, schedule an inspection, connect your utilities, et cetera. How could AI affect the future of home buying?

Tanguy Catlin: That’s all one experience, and it includes five or six different industries. That friction cost will be removed, and business models that allow customers to have a unified experience with one service provider across multiple touchpoints will emerge.

The second thing I think will emerge is that this technology will unlock significant levels of productivity. Mature economies need productivity gains because the population is aging. To maintain the growth in GDP per capita, we need to have a smaller workforce contribute more on behalf of the entire population.

Those productivity gains will come from technology. When you gain significant levels of productivity, you typically drive consumption. Consumption, in turn, creates jobs, which is a good thing. The economic surplus allows the consumer to spend on things other than potatoes and poultry. They want to consume entertainment, healthcare, and wellness. There will be a rise of sectors and gains in productivity that will lead to wealth benefiting the consumer.

The operating-model advantage

Roberta Fusaro: You’re talking about a brand-new world for customers. For companies, it’s great for efficiency, partnerships, profit opportunities, and more creativity and innovation.

If AI models are becoming increasingly accessible, then what actually differentiates companies from one another? Say, five years from now, will it be data, distribution, customer relationships, or something else entirely?

Tanguy Catlin: All of the above, probably. You’re correct that the value creation for an enterprise will be because of innovations with product services and business models more than productivity. The second point you are addressing now is that the sources of competitive advantage will be affected by this AI technology.

For example, AI provides access to predictions at a really low cost. So we’re going to use predictions for everything, because the cost of this will be extremely low. To make predictions, you need quality data, and the cost of your prediction is cheap. The value of quality data will go up significantly. One source of competitive advantage for companies will be any proprietary data that they can keep for themselves.

When you’re bombarded with predictions, you’re going to need to use your judgment to choose wisely and place the right value behind outcomes associated with those predictions. There will be significant value for the portion of the labor force that is trained and demonstrates the skill set associated with judgment.

I think data is a competitive advantage that will be more valuable in the future. I think another one we discussed earlier is whether you can embed your product and services into the habits of your customer.

To answer your question as to what will differentiate the winners from the losers—I believe it will be the metabolic rate of learning, as well as the ability to create a self-reinforcing system where technology allows you to experiment better.

Experimentation gives you greater insight into what allows you to win. Therefore, you must have offerings that you can enhance with technology for testing. I believe that it offers good insight into how we differentiate winners from losers in the world, because we are currently overwhelmed by the pace of change. People need to realize that advancement will never be as slow as it is today. A big part of winning the future will be having an adaptable operating model that allows you to learn fast and keep up with change.

Roberta Fusaro: How should organizations think about organizing differently, to achieve that speed?

Tanguy Catlin: When you’re building agentic systems, you’re tasking agents with achieving certain outcomes, which means you’re pivoting from a knowledge-based organization to an outcome-based organization.

To deliver those outcomes, you’re going to need to combine human labor and agents, or skills that work together, to produce that outcome. You will automate several of the entry-level tasks and activities, as you will probably assign more value to the people who have the capacity to oversee the system. So the shape of the labor force will change. It will become flat and horizontal toward your outcome, with a different mix of talent in the organization.

How do you get there? Today, the single most important challenge to the adoption of the technology we’ve discussed is change management and upskilling of the workforce. It turns out that if you provide employees access to AI tools, and you don’t train them to use those tools, your outcomes worsen.

It is necessary to increase the upskilling of your workforce to evolve. Leaders will also need to demonstrate that they are growth oriented, have a certain level of humility, and have a set of incentives and systems that reward experimentation rather than punish failure. That will be critical for organizations to make that pivot.

The hardest part is not AI

Roberta Fusaro: Tanguy, you’ve talked about how CEOs need to simultaneously manage a strategy lens, a technology lens, and a people lens. We’ve talked a little bit about some of what’s required for the organization to prepare itself for AI success. Which of those three lenses is proving the hardest for leaders today, and why?

Tanguy Catlin: The organizational lens by far is the hardest. While it is not difficult, people are not knowledgeable enough to realize that productivity is not the right strategy to pursue right now.

You will need to pivot to differentiation, and I don’t think people fully understand how AI will affect the value of different strategic sources of advantage in competitive moats. The system adaptation is a human challenge. Think of all the general-purpose technologies that have been deployed over time. They have taken years, if not decades, to be able to move from improvement you can measure at the micro individual level to the format where you can see them at the P&L or societal level.

This change requires adaptation, and adaptation comes with a workforce that is confident in the value of pursuing change, that has been upskilled to be able to pursue that change, and that has a set of management systems that are evolving to be congruent with the change that is required. That is by far the hardest adaptation to make.

Roberta Fusaro: I’m thinking of that movie where they say, “Soylent Green is people.” AI is people. It’s the hardest input to manage. If you were leading a company through an AI transformation, what metrics could you use to understand whether AI is creating long-term value versus just short-term productivity gains here or there?

Tanguy Catlin: A number of those metrics are not changing. Can you measure whether you’re gaining market share? Can you measure whether you’re extending your margins? Can you measure whether your customers are more satisfied with your offerings? Do you measure that your employees feel excited about working for you? Those are the metrics that do not change.

I would be very explicit up front—you cannot deploy AI everywhere in your organization at once. It’s impossible. You will need to define where AI is going to create the most value and focus on those areas. In the McKinsey book Rewired, we say, “Find a domain and then do not deploy the technology on top of existing processes. Redesign those processes to capture value from the technology. And do it in such a way that will be scalable so that you can expand to other domains over time.”

Roberta Fusaro: Which industries do you think are most vulnerable to disruption from AI?

Tanguy Catlin: Which one is not? You have two sides to AI. You have cognitive AI and what I would call robotics. We’ve seen faster development in cognitive AI right now, so I think industries that rely heavily on white-collar labor will be exposed to significant changes in the short term. You can see professional-services firms having to rethink how they will provide a service and price it appropriately. Simultaneously, we see continued development in robotics, which will affect more blue-collar industries. For example, you can see the rise of robo-taxis having an impact on the transportation industry.

There are some industries where regulation will affect the level of structural disruptions that we can anticipate. For example, the healthcare space will face more challenges. At the same time, the methods we use to pursue R&D in pharma right now are being transformed by AI, and hopefully that will one day lead to the development of a cure for cancer.

It’s a great question and a complicated one to answer, because I suspect most industries will be affected by AI one way or another. I think the primary question is whether it will be cognitive AI or robotics AI that will drive different levels of disruption.

As we mentioned earlier, there is a lot of unnecessary friction in those industries that can be taken away, and/or will have AI offering unbelievable innovation that would unlock sectors we don’t even think about today.

Roberta Fusaro: What about individual jobs? How could those be affected by AI?

Tanguy Catlin: At the macro level, about 53 percent of activities performed by workers could be automated if we were able to use today’s available technology. It’s a significant number. However, the reality is that very few workers can automate the majority of their job tasks.

That leaves you with jobs where the activities are material enough that you can explore how to enhance productivity with automation, but not to the point that you can simply remove the job altogether.

There will be massive job reconfiguration associated with AI. You will be able to regroup activities to extract productivity from them alongside massive upskilling. For those who remain, their jobs need to be augmented with AI.

Move early, learn faster

Roberta Fusaro: You’ve spent much of your career helping organizations transform through technology. As you look back at previous digital trends, what lessons do you think leaders need to remember most as they navigate the world of AI?

Tanguy Catlin: I think most people have been thinking about technology or innovation as a zero-sum game. I do believe that when you look at the evidence, every new general-purpose technology has actually significantly increased the size of market and value creation.

I think that’s very important because many people are fearful of the impact that technology will have without actually recognizing that it could be a significant unlock. To me, you must be thoughtful about the risk but embrace the fact that a productivity tool will significantly expand the size of many markets.

As the markets expand, the value typically accrues to a smaller number of participants as a “winners take all” dynamic takes hold. Those winners are emerging very early in those disruptive moments. Place some bets, evolve your pivoting model to move fast, and make sure that you are learning along the way so you can pivot when you need to.

Another thing I come back to is this notion that people are obsessed with technology. I think you should obsess about it with your people. Take a step back and try to figure out how you can move faster than you have before. As you do that, think through the change management and workforce transformations that will be necessary to go through that journey. That’s what I’ve seen over and over again, and I think that will be even more important this time around.