When it comes to AI, investors are starting to distinguish between the serious and the dabblers. Although almost every large organization is testing the use of generative AI somewhere, fewer than half of companies represented at recent Bain CEO forums have moved from pilots to real scale.

In Bain’s CEO forums, we see companies traveling a predictable, four‑stage path to AI implementation:

• Stage 1: Understand the shift
• Stage 2: Run disarticulated tests
• Stage 3: Evolve the business
• Stage 4: Lead the transformation

Most companies are marooned in Stage 2, running isolated pilots on different platforms, without a shared vision, operating model, or talent plan.

Escaping this trap is a leadership act, not a technical one. Most CEOs believe they have a clear strategy, but that it’s in execution, particularly the ability to translate that ambition into scaled outcomes with speed and consistency, that they falter. In our experience, companies that are succeeding typically have one thing in common: a CEO who has a clear conviction about the way AI can transform the business.

Does AI adoption require a fundamentally new model of leadership, or does it simply raise the bar on behaviors CEOs have long known matter—clarity of purpose, focus, and the ability to drive change at scale? The answer is: both.

Five leadership strategies—anchor in purpose, lead with questions, own the AI agenda, protect experimentation, and build an AI-first mindset—help CEOs bridge these demands, ensuring they don’t lose sight of their traditional responsibilities while building a model of leadership for the AI era.

1. Anchor AI in purpose and customer impact

The most essential question leaders can ask today isn’t “What can this tool do?” but rather, “How does AI advance our mission?” And its logical follow on, “How, when, and where do I start?”

Yet most CEOs are still approaching AI primarily through the lens of cost reduction and productivity, instead of using the technology to reshape customer outcomes or competitive positioning. Rather than starting with tools, Walmart’s CEOs have framed AI against the outcomes they want to deliver: value, assortment, convenience, and trust. Former CEO Doug McMillon repeatedly emphasized those priorities as the company’s North Star. Under current CEO John Furner, that stance has become operational. He has reorganized Walmart to centralize AI platforms and shared capabilities across the enterprise, enabling its operating segments to be “closer to our customers and members,” and is rolling out new customer-facing experiences, including integrations with tools such as Google’s Gemini to make shopping more seamless, intuitive, and personalized.

Walmart has simultaneously built a coordinated portfolio of AI tools to achieve those outcomes. For example, Trend-to-Product uses AI to sense trends and tighten apparel development timelines by roughly 18 weeks, while tools such as Sparky help customers discover products and compare products more easily. These are not isolated use cases but rather part of Walmart’s broader customer-first push to use AI to accelerate responsiveness and relevance across the business.

Illuminating questions CEOs should ask:

• Where are we using AI not just to reduce costs but also to fundamentally change customer outcomes? Why aren’t we moving faster on those efforts?
• What are we doing with AI that doesn’t clearly advance our strategy? Why haven’t we stopped?

2. Lead with better questions, not answers

Once AI’s purpose begins to become clear, CEOs must shift how they lead the conversation around the technology. Traditional models of leadership reward certainty, whereas AI rewards curiosity. Curious CEOs are asking: “Where can AI elevate—not replace—our people? If we designed this business today as AI‑native, what would we do differently?”

Sal Khan, founder and CEO of Khan Academy, personifies what it means to lead with questions in the age of AI. When he first engaged with early generative AI models, Khan did not start with a roadmap. Instead, he and his team started by asking questions about how their students learn and leveraged the principles of learning science in their research. They designed experiments to explore what technology could do. They held internal hackathons and tested use cases, eventually creating Khanmigo, an AI-powered tutor and teaching assistant that now supports learners on a global scale. Rather than start with answers, Khan treats AI as something to interrogate and learn from, using the technology personally and across the organization to test ideas quickly. He is explicit about the risks of AI but focuses on designing guardrails that channel the technology toward positive outcomes, what he calls “turning fears into features.”

Illuminating questions CEOs should ask:

• What do we do that an AI-native competitor would do completely differently or would never start in the first place?
• Where are we defaulting to humans for work that AI could either do better or meaningfully augment? What’s stopping us from changing?
• What training and support can we offer to make the work that’s best done by humans also the best work for humans to do?

3. Own the AI agenda and reinvent the core

Fewer than half of CEOs recently surveyed by Bain feel confident they can build the necessary capabilities—including AI—at the pace required. Sharp questions inevitably expose that scattered pilots are not enough. CEOs must own the AI agenda and commit to reinventing the core. This includes actively removing structural, funding, and governance roadblocks that slow progress. CEOs set the pace and must move at that pace themselves. Stage 2 organizations treat the implementation of AI as a portfolio of experiments, while Stage 4 organizations treat it as a CEO‑owned transformation of the business model and customer experience.

In comments to investors in early 2026, Jamie Dimon, CEO of JPMorgan Chase, candidly addressed a topic many leaders struggle with: AI’s impact on workers. JPMorgan is spending more than any other company in its industry on technology this year—$19.8 billion—including a dedicated multibillion-dollar tranche for rewiring core workflows across every business line for AI. Workers have been displaced because of AI, he acknowledged. The company has active redeployment plans and headcount has roughly remained flat, but the composition of the bank’s workforce has shifted. Client-facing roles are expanding, while operations and support are shrinking.

Operational benefits are quickly adding up. With more than 450 agentic AI use cases already deployed, operations teams are now handling 6% more accounts per employee, fraud costs per unit have fallen 11%, and software engineer productivity has climbed 10%. One division replaced a controls review process that previously required 200 people with an AI-enabled workflow—then identified 3,000 to 5,000 additional employees across the firm in similar roles. This scale of change requires CEO commitment to AI not as a technology initiative, but as a mandate to fundamentally reinvent the business from the inside out.

Illuminating questions CEOs should ask:

• What core processes are we protecting that we should be willing to break in order to fully capture AI’s value?
• Where is fragmentation of data, tech, and ownership actively preventing scale? Who is accountable for fixing that?

4. Give permission and protection to experiment

Of course, bold bets on the core require a culture in which people can try, learn, and adapt. Many leadership teams continue to send the implicit message, “Innovate—just don’t fail.” So, CEOs need to give teams permission and protection to experiment in service of the bolder mission and actively intervene to clear the technical, legal, and organizational barriers that teams cannot remove on their own. Siemens CEO Roland Busch has put industrial AI at the center of the company’s public story and product roadmap, including from major keynote presentations at CES. That level of sponsorship gives engineers, plant leaders, and product teams cover to run bolder experiments in factories, infrastructure, and healthcare—places where the stakes are high and old, and the outdated processes are deeply entrenched.

Illuminating questions CEOs should ask:

• Where are we asking teams to innovate while implicitly penalizing failure? What am I doing that might be contributing to that?
• In which high-stakes areas of our business are we avoiding experimentation? What risks are we afraid of?

5. Hold teams accountable for adoption by leading with an AI‑first mindset

Experiments only matter if they change how the organization works every day. It is not enough for AI to be “available” somewhere in the enterprise; people need to see that the technology is becoming the default way leaders think, decide, and work. At Bain, Worldwide Managing Partner Christophe De Vusser is a role model of using AI to those ends. Rather than delegating AI to IT, he devotes more than 20% of his time to the firm’s modernization and AI agenda, resolving trade-offs (in resource allocation or vendor selection, for example), helping accelerate adoption of new processes and technology, and maintaining the pace of individual strategic initiatives across the program. He was also a top user during the pilot of Bain’s internal agentic platform—a visible signal that leaders are expected to use the tools, not just endorse them. And he continues to learn alongside clients and leaders across the AI ecosystem, testing Bain’s direction as the technology evolves. The message to the organization is that AI-first is not a slogan but rather a behavior modeled from the top.

Illuminating questions CEOs should ask:

• Does my team see AI in my calendar and my habits, not just my speeches? Am I spending 15% to 25% of my time learning, using, and advancing AI?
• Who on my leadership team is visibly modeling AI adoption? Who is quietly resisting it?
• In what day-to-day work is AI still considered optional? What would it take to make it the default?

CEOs who break out of Stage 2 understand that AI is not a set of experiments to manage but rather a transformation to lead. They anchor AI in purpose, stay relentlessly curious, take ownership of the agenda, create space for experimentation, and hold their organizations accountable for real adoption. The question is no longer whether AI will reshape your industry but whether you will shape how it does. Those who lead decisively will redefine their businesses; those who hesitate will find themselves competing with companies that already have.