AI adoption is everywhere. Token costs have become a board-level topic. Pilots are proliferating in companies, across every function and geography. Yet relatively few CEOs point to enterprise-level financial impact from AI: McKinsey research shows that 89 percent of organizations are reporting regular use of AI in at least one business function. Yet the proportion of AI high performers, or those who attribute at least 5 percent of EBIT to their use of AI and describe the technology’s impact as “significant,” has remained flat at about 6 percent of all respondents to McKinsey’s State of AI survey.1
There is still a notable distance between AI’s promise and the value companies are realizing from it. Technology isn’t the sole reason for this gap, however. There’s also a gap in leadership—rooted in business fundamentals such as strategy, organization, culture, and risk, but magnified by the rise of AI.
CEOs’ AI moment is shaped by a series of paradoxes: Move fast and commit boldly to AI despite the absence of benchmarks or best practices. Be decisive about where to place your bets but retain some optionality; AI models and tools will evolve, and any commitment will need to be revisited. Build proprietary advantage using AI tools that are available to everyone, but recognize that product and process differentiation will come from how quickly your organization learns, not the model or provider you use. Emphasize productivity enhancements through AI and use those efficiencies to help fund the rewiring journey, but remember to invest just as much in growth—efficiency alone is not a survival strategy. Embrace AI deeply, but don’t cede your organizational identity to it. The strongest companies will use AI to double down on what makes them distinctive now and what will set them apart in the future rather than reflexively pivoting to a generic “AI-first” playbook.
These imperatives are complicated and interrelated, presenting CEOs with a unique business challenge: AI is changing how companies approach strategy and planning, talent development, capital allocation, technology systems, corporate culture, and the value proposition to customers. No other technology trend of this leadership generation has so thoroughly affected all six of these domains at once, which means no single function can completely own this challenge. This is not just another technology cycle. It is a metamorphosis of the organization, with choices that can only be made by the CEO.
In conversations with hundreds of CEOs across industries and geographies, we’ve observed that they take one of three approaches to AI:
- Technology supporters frame AI primarily as a technical program and look to chief information officers (CIOs) and chief technology officers (CTOs) to take the lead on decisions and pilot programs.
- Productivity champions emphasize domains with high efficiency in an attempt to bend the cost curve.
- AI visionaries see AI as central to redefining their company’s competitive advantage; these leaders believe their organizations need to learn rapidly—and reinvent themselves to stay alive.
A year ago, the CEOs we spoke with were evenly distributed among these groups. Today, the visionaries, who are not delegating AI to others but treating it as their personal responsibility amid uncertainty, are becoming more common. They recognize that AI is not just a program to implement or even a transformation to undergo. It is fundamentally reshaping how organizations create value.
The CEO must directly manage the decisions that follow. The choices cluster around three areas: raising the organization’s strategic ambitions, rearchitecting the organization, and resetting the culture. Visionary CEOs understand that, across all three, AI removes the constraints around which today’s organizations have been designed—forcing senior leaders to reconsider how the organization itself works.
Raise strategic ambitions—even when there’s no precedent
The introduction of AI is already redrawing value pools, collapsing marginal costs, and creating new competitors and new sources of competitive advantage for organizations. It raises for CEOs questions that go well beyond “AI strategy.” Among them: How fast and in which direction should we move? Where will growth come from? How can productivity gains made today be reinvested into tomorrow’s growth engine?
CEOs cannot delegate this conversation. The introduction of game-changing technologies requires more than resetting targets. It means recalibrating expectations about the business’s full potential, rethinking employee- and customer-value propositions, and repositioning the company’s competitive equation in an AI-shaped future—even when there is no precedent for what any of those might look like.
In our experience, the highest-leverage strategic decisions for CEOs are choosing where to place their bets and how to build, own, and protect proprietary intelligence.
Place big bets, pursue bigger ambitions
The CEO’s most critical role is selecting the two or three domains where AI-native redesign will have the greatest impact on the business. For an insurer, that may be underwriting and claims. For a pharma company, it may be R&D. For a consumer business, it could be an end-to-end customer journey.
What’s important here is being bold and systematic: Broad adoption of AI across the organization enables greater diffuse productivity and raises AI fluency—and that matters. But a real metamorphosis comes from fundamentally reinventing priority domains end to end, not from spreading AI thinly across the enterprise. The best CEOs make a clear distinction between investing in the broad adoption of AI for learning and fluency and making concentrated investments in two or three domains to achieve larger-scale transformation and profit-and-loss (P&L) impact. The ambition here is not incremental efficiency; it is growth.
Some companies successfully enforce their AI commitments by publicly setting the bar high—for instance, using investor days to clearly articulate their aspirations for growth through AI transformation—and embedding AI- and transformation-oriented KPIs into every performance discussion, from the CEO on down.
Own what makes you distinctive
Once CEOs have identified priority domains, they must make strategic choices about which AI capabilities the company will own and which it will source from third parties. These decisions are more complicated than the traditional “buy versus build” choices.
Cost is one factor. AI costs behave differently from traditional software costs. An AI-enabled task can consume up to 30 times more tokens at one company than at another, depending on how it is orchestrated. CEOs therefore need the flexibility to buy, build, host, route, and switch between models as economics, capabilities, and business needs evolve. The company’s proprietary data, knowledge, decision logic, semantic layer, and learning capabilities should be permanent assets, while the underlying models should become increasingly interchangeable. As David Tepper, CEO of Pay-i, frames it: Tokens are not value; tokens are the bill. The focus, instead, needs to be squarely on how to use AI agents to create value.2
Competition is another factor. When a business routes trade secrets, workflows, customer data, and other intellectual property through a third-party model provider, it may be transferring its means of production to a vendor that could later choose to compete in the same space. The CEO must therefore work with technology leaders to set explicit controls for intellectual property, denoting what can leave the building and under what conditions while rigorously protecting everything that should be kept in-house. Such controls are especially important in an evolving geopolitical and technological landscape, where ownership of AI capabilities is becoming central to companies’ ability to compete on the global stage and build strategic resilience and societal trust.3
As access to frontier intelligence becomes increasingly ubiquitous, competitive advantage will come from the learning loops built around those models, rather than from the models themselves. Each new piece of information fed into the system or correction to the existing model will compound into proprietary expertise that a competitor using the same model cannot replicate. This form of competitive distance cannot be bought; it must be built. To that end, the CEO has to treat institutional learning as a strategic asset.
To be clear, the CEO should not own enterprise IT architecture—that would be a category error. But to ensure that AI scales effectively across the organization, the CEO must proactively engage in the following key decisions involving enterprise IT architecture: Which domains should receive priority investment? Where should capital be allocated? Who owns the data? And as the threat of AI-enabled cyberattacks continues to accelerate, what role should CEOs play in identifying emerging risks and implementing appropriate protections? What level of risk will the company tolerate, and how can the organization ensure responsible usage?
The CEO of a large global airline recognized this early. In the initial phases of the company’s AI transformation, he spent several hours over multiple weeks with the CIO to understand what each element of the IT architecture enabled and why it mattered—not to become a technologist, but to make informed capital decisions. That is the CEO’s job.
Rearchitect the organization around human–agent teams
In a world in which both AI and humans have agency, operating models need to be redesigned to account for this new “hybrid intelligence,” the combination of human judgment and AI capability required to run and grow the business. As human–agent systems become the building block of work, “individual in a role” and “expertise in a function” can no longer be relied on as the core units of organizational design.
Many of the industrial-era principles of organizational design are changing. Under a human–agent operating model, for example, meetings will become more event driven rather than calendar driven, planning cycles will be continuous, and managers will spend less time gathering and transmitting information and more time exercising judgment, coaching, and resolving exceptions. Some decision cycles could collapse from weeks to hours. Some decisions that don’t require human intervention will disappear entirely. It’s a significant shift—one that the CEO must manage without benchmarks or blueprints to turn to.
If the early take on AI was all about job loss and replacement, the prevailing view of AI now is more about reinvention and redefinition. Research from the McKinsey Global Institute suggests that by 2030, nearly every occupation will experience significant skill shifts and the partial replacement of current activities because of AI. These data suggest that, unlike with prior technological waves, most jobs will require a fundamental redesign rather than redeployment alone.4
This will not be a frictionless transition. Organizations will need to invest materially in skilling, apprenticeship, role redesign, and workforce mobility, while recognizing that not every individual will make the transition at the same pace. CEOs will need to ensure that workers develop AI-oriented skills while strengthening and elevating traditional human capabilities, such as judgment, creativity, adaptability, and complex problem-solving. In fact, these traditional capabilities will only become more valuable. As AI makes information increasingly accessible, employees will turn to leaders not for answers, but for judgment. They will seek context rather than content, direction rather than data, and meaning rather than information.
Taking all this into account, the CEO must answer the organizational design questions that no one else can—while not treating “design” too rigidly as a concept. Depending on the company’s value creation objectives, the CEO will need to consider, among other things, where to shift resources once productivity gains are achieved (employee productivity is not the same thing as enterprise productivity); which parts of the business require more (or less) investment in people and capabilities; how to balance investments in both technology and people; and how the organization might change its shape and composition in future, which will have implications on everything from early career strategies to incentives.
None of these decisions can be made function by function. In the AI-era organization, leadership is a team sport, and the decisions that matter most cut across strategy, technology, finance, talent, and operations. The best decisions are collective ones that benefit the entire organization rather than any single function. The CEO’s distinctive contribution is to impose and embody this enterprise view, insist that leaders solve for the whole, and explicitly surface tensions and help the team come to a common view on each of them.
The rest of the senior team remains vital in this process: The CIO must build the data foundations, establish a flexible tech stack, and set vendor strategy. The CFO must rebalance the company’s investments in human and technological capabilities to fuel growth rather than efficiency alone. The chief human resources officer must codevelop the human–agent workforce, establish the environment of trust that holds it together, and guide the workforce transformation. The COO must sequence and deliver domain transformations.
But the CEO’s job is the one no one else can do: to set an explicit, shared mandate, break the silos that push leaders to optimize locally, and hold the team collectively accountable for enterprise outcomes rather than functional wins.
Build a culture that strengthens and preserves competitive moats
For all the discussion of technology decisions associated with AI, the largest barrier to AI adoption is not technical, but human. Culture, trust, fear, inertia, and organizational politics are the primary constraints, as they are in any major transformation. Lack of clarity, collective-action problems, loss aversion, murky communication, and other cross-cutting dynamics come into play, meaning the CEO is the only person who can address them at the enterprise level: As the CEO goes, so goes the culture of the organization.
There are several cultural characteristics associated with those organizations that have had an easier time adapting to the AI era. Hands-on learning and AI-fluent leadership are critical. Previous McKinsey research demonstrates that active engagement and role modeling by senior leadership is the single strongest predictor of successful AI implementation.5 The data show that 70 percent of employees say they feel personally ready to use AI, but only 27 percent of leaders say their organization is ready to make the changes required to use AI at scale. And in many organizations, senior teams are significantly less fluent in AI than middle managers and frontline employees. The gap between stated priority and demonstrated behavior erodes trust and slows adoption throughout the enterprise.
To counter that, the CEO must be a visible, genuine learner. The CEO does not need to master every tool or coding technique, but they do need to set the right example—asking sharp questions, setting expectations that teams use AI regularly, and showing enthusiasm for trying something new. For instance, one chief executive we worked with participated in a company-wide contest to program virtual self-driving cars for a racing league.
Several other CEOs have told us they spend one to two hours each week with more AI-fluent colleagues, simply learning how to use the tools more effectively—a form of reverse apprenticeship. The CEO at a global retail bank identified for himself and the entire executive team a set of mentors from the company’s technology and AI teams to answer detailed questions about new technology, such as “How does a transformer work?”
The most effective CEOs are willing to challenge and even upend their own routines to role-model new ways of working. They are candid about organizational successes, but also about situations in which the organization perhaps launched too many pilots or failed to address resistance. Indeed, the best CEOs actively own the AI narrative in their organizations; they understand that in doing so, they will be better equipped to guide the business through changes that are both driven and enabled by AI.
Beyond the CEO’s own fluency, CEOs in some leading organizations are setting an explicit standard for the entire senior leadership team—and beyond. They are ensuring that executives know how to use AI tools themselves. They are also empowering executives to ask questions, encourage their employees, and model the behaviors that drive broad adoption. Moderna CEO Stephane Bancel has launched AI Academy, training employees at all levels in AI tools, workflows, and governance—democratizing the technology and signaling that AI is a permanent, central feature of how the company operates. In other companies, CEOs are taking senior teams on “go and sees”—visiting companies inside and outside their industries—to accelerate adoption of AI-driven approaches to sales, marketing, and other critical business activities.
Early research on how executives themselves use AI—as a reasoning partner, a research accelerator, a decision support system—suggests that leaders who are personally fluent create a fundamentally different quality of strategic conversation with their teams. AI can increasingly provide analysis, recommendations, and transparency—theoretically giving leaders the ability to intervene everywhere. The best CEOs, however, use this newfound transparency not to create the world’s greatest micromanagement machine but to delegate more responsibility to others and create conditions for trust and accountability. They actively seek to construct a trust architecture alongside a technology architecture.
In every AI-ready culture, there must also be a willingness to experiment at every level and to look beyond the status quo. In fact, this may be the most urgent need for organizations looking to create value from AI. It also presents a critical leadership task for CEOs—to shift the culture from expertise based to learning based and to unleash courageous unlearning and inquisitive relearning among individuals and teams.
Consider what that looks like: Salesforce launched early agentic copilots internally before public release, letting engineers propose entirely new use journeys and break workflows. And Citi piloted a new workflow with 5,000 employees, then collected feedback before rolling out the changes more broadly, starting with a minimally viable solution, monitoring performance, and iterating rapidly.
These and other leading organizations are building feedback mechanisms, emphasizing psychological safety, and creating working norms that allow people to innovate and clearly signal to markets, regulators, and communities that the organization is leading responsibly with AI.
At the center of it all, the CEO’s overarching role is to normalize a new culture of AI, both its positives and negatives. It’s not enough to just encourage open discussion and collaboration—the CEO must demand it. This is harder than it sounds, of course.
In our conversations with CEOs, some have shared that they have never felt more alone in their decisions about AI. Boards have ramped up pressure on CEOs but often do not understand AI themselves, creating stress without providing clarity, and failing to connect the AI opportunities they hear about in the boardroom with specific business objectives on the ground. We’ve also observed that technology vendors are under pressure themselves, selling solutions that increasingly demonstrate that they do not understand how to capture value for companies. Finally, management teams reporting to the CEO often fear what AI will mean for their own domains and become resistant to change and collaboration.
Isolating as it may be, it’s critical that CEOs accept the cultural challenge: As recent McKinsey–World Economic Forum research notes, culture follows the CEO. Regardless of what policies are issued, behaviors change only when leaders signal what really matters. Employees watch what CEOs measure, reward, talk about, invest in, and role-model.6
In a relatively short period of time, we’ve gone from AI to generative AI to agentic systems. AI models and tools keep changing, as do their applications and cost curves.
The one constant is the CEO’s obligation to make the choices only they can: to raise the organization’s ambition, rearchitect how it works, and reset its culture. These are top leadership decisions, and they cannot be delegated. Perhaps the biggest challenge for solution-driven CEOs may be the realization that this is not a transformation with a finish line; the AI metamorphosis is ongoing, which means we’ll also be returning to, reviewing, and revising this playbook.
But the ever-evolving state of AI will also create generational leadership opportunities for CEOs: Those who treat it as their personal, nondelegable responsibility can meaningfully impact and even define the next era of their organizations and industries. In that sense, the role of the CEO has never mattered more.
