In 1965, Paul McCartney played George Martin a new song, “Yesterday,” and asked what it needed. Martin heard something McCartney had not: a string quartet on a Beatles rock record, at a time when nothing quite like that had been attempted. McCartney resisted the idea, but Martin pressed for it. The result became one of the most recorded songs in history.

That episode is worth recalling as executives move quickly to bring AI into every meeting. The most valuable person in the room is not always the one with the most answers. Sometimes it is the person who can hear a possibility that no one else has recognized.

Today, I often tell people something that can sound almost absurd: the Age of Answers is over. Not because answers have lost importance. The opposite is true. Humanity has access to more answers than at any previous point in history. Artificial intelligence can write business plans, diagnose diseases, analyze markets, draft legal briefs, build software, and produce strategic recommendations in seconds. Work that once required teams of highly trained experts is increasingly available to anyone with a laptop.

That is exactly the point. When answers become abundant, they cease to be a durable competitive advantage. George Martin demonstrated this decades before AI existed. He was not a better songwriter than Lennon or McCartney, nor a better musician than Harrison or Starr. What he contributed was a different way of hearing possibility: orchestration, tape experimentation, and the willingness to tell four young musicians hard truths about their own material. That is the promise of AI as well — not as a replacement for human creativity, but as an amplifier of it, if leaders know how to use it that way.

That differs sharply from how many organizations use AI today. In minutes, whiteboards fill with timelines, milestones, decision trees, and Gantt charts. Every team identifies the “X” on the map, and every team designs the fastest route to get there. Almost no one asks whether they are using the right map.

The thinking may be disciplined, the analysis rigorous, and the execution flawless. But innovation rarely begins with an “X” on a map. It begins by asking whether the map accurately describes the territory in the first place.

For years, helping accomplished leaders make that shift was extraordinarily difficult. Success rewards certainty, and organizations reward predictability. Yet every meaningful breakthrough begins with an uncomfortable question: What if we are solving the wrong problem?

Today, leaders no longer need to debate that question in the abstract. They can ask AI to generate the solution they were about to build, usually in less than a minute. That changes the conversation completely. The issue is no longer whether AI can produce better answers, but what leaders should do when everyone has access to the same ones. In that environment, the leadership task moves upstream: defining the question, testing the premise, and deciding which answer deserves trust, attention, and action.

Never Put the New into the Old

The lesson reaches far beyond any single industry. Many organizations see AI as a technology challenge. It is not. It is a leadership challenge. Every technological revolution eventually exposes an outdated operating system. The constraint stops being the technology itself and becomes the assumptions, structures, and habits built for the world the technology is replacing.

Martin never tried to make the Beatles sound like a classical ensemble because that was where he felt most comfortable. He did not force a new sound into the framework of his old training. Many organizations do the opposite with AI: they force a revolutionary capability into structures designed for a world in which answers were scarce.

That mistake is visible today. Organizations create AI task forces, appoint Chief AI Officers, mandate AI training, and rewrite company policies. Those initiatives may have value, but many are attempts to bolt a revolutionary capability onto organizations built for an era when information was scarce, expertise was expensive, and answers were difficult to obtain. That world no longer exists.

The Three Levels of AI Leadership

The central question is not “How do we use AI?” A better question is, “What kind of organization should we become because AI exists?” The answer begins with three levels.

Level One: Optimize. This means doing yesterday’s work better — summarizing meetings, writing reports, automating customer service, and generating code. The gains are real, but efficiency stops being a competitive advantage when everyone has access to the same tools.

Level Two: Simulate. This means challenging today’s assumptions. I once gave a cohort of military fellows an innovation challenge and watched them build the fastest possible route to a solution without ever asking whether it was the right solution. Today, I skip the debate: I ask AI to generate the answer they were about to spend a week building. It usually does so in under a minute. That is Level Two — using AI not to confirm a plan, but to stress-test whether the plan should exist at all. Leaders can use AI as a skeptical customer, an aggressive competitor, or an impossible board member, war-gaming scenarios and searching for blind spots.

Level Three: Create. This means inventing tomorrow’s possibilities. At this level, AI stops being merely an assistant and becomes a creative partner. Most leaders ask AI, “How can you help me do my job better?” The more consequential question is, “What can we create together that neither of us could create alone?” This is the level on which George Martin operated with the Beatles — not optimizing their sound, not merely stress-testing it, but helping create something none of the five of them could have made alone.

That is why the real revolution is not AI itself, but what AI reveals about leadership. For decades, intelligence was the scarce resource. It is not anymore. AI democratizes intelligence, but it does not democratize judgment.

Intelligence produces answers. Judgment decides which questions are worth asking. Judgment recognizes when a problem has been framed too narrowly, when everyone — including the machine — is converging on the same obvious solution, and when it is time to abandon the map before someone else draws a better one.

George Martin never out-wrote Lennon and McCartney. He did not need to. He heard what they could not yet hear, and he told them the truth about it. That is the job now: not to out-answer the machine, but to hear the question it cannot ask.

Before the next AI strategy meeting, leaders should ask:

Where are we merely optimizing yesterday?

Where should we be using AI to challenge our assumptions instead of confirming them?

What could we create that has never existed before?

The age of answers is ending. The Age of Judgment has just begun.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com