Here’s what nobody wants to say out loud: most leadership teams are about to make the same critical mistake.

When uncertainty rises, organizations instinctively tighten their grip. More oversight, more approvals, more governance. It feels like leadership. It looks like responsibility. But it’s precisely the move that makes adaptation slower and harder.

McKinsey’s 2025 State of AI research makes the problem stark: nearly all companies are investing in AI, but just 1 percent believe they’ve reached maturity. The primary barrier isn’t employees, who are ready, but leaders who aren’t steering fast enough.

The Pattern Everyone Keeps Misreading

A new initiative rolls out. People nod, but execution drags. Progress shows up as updates rather than outcomes. Leadership diagnoses the problem as resistance or accountability. So, the response is predictable: more check-ins, more reporting, and more enforcement.

But what looks like a performance problem is often a communication problem. What looks like resistance is often the system protecting its ability to think.

Deloitte’s 2025 Global Human Capital Trends confirms this: organizations leading in AI adoption share the common characteristics of high trust, data fluency, and agility. The firms stuck in pilot purgatory share a different pattern. They’re treating transformation as a training rollout.

The real shift is simpler than most leaders think: move deliberately away from control and toward genuine agency—not just for AI agents, but for the humans who will use those tools to take the company further than seemed possible two years ago.

If we reach back into instincts for centralized control and don’t allow smaller units and cross-functional teams to move with creativity and autonomy, the system fails. Not because the technology isn’t capable, but because humans won’t have the flexibility or decision-making authority to use it well.

Biology Has Already Solved This Problem

The most useful leadership model for this moment isn’t another business framework. It’s coming from an unexpected place.

Michael Levin, a developmental biologist at Tufts University, studies how living systems build and repair themselves without central command. His research focuses on how collectives of cells coordinate toward form and function, demonstrating that intelligence and problem-solving don’t require a brain or central controller. They require clear goals, feedback loops, and freedom for individual components to adapt.

Consider a lizard that loses its tail. The animal regenerates a tail that functions and fits, but it’s not a perfect replica. In many species, the regenerated structure isn’t segmented vertebrae, it’s a single cartilage tube. An “imperfect replicate” that nonetheless does exactly what a tail needs to do.

That imperfection is the lesson.

The system doesn’t freeze because it can’t rebuild the ideal structure. It adapts to constraints and restores what matters: function over fidelity, outcomes over orthodoxy. And critically, this regrowth isn’t directed by the brain sending detailed instructions. It’s coordinated by local cells responding to each other through signals and feedback loops, making decisions where the information actually exists.

The management translation is immediate: healthy systems don’t scale by issuing perfect instructions from the top. They scale by aligning local actors around clear outcomes and letting them solve problems with the tools available. The executive team sets goals and constraints, but problem-solving happens at the edges where reality meets execution.

Control Versus Coherence

Most companies still operate on an outdated “factory” paradigm. Standardize inputs, control the process, inspect outputs. That model works when work is predictable and roles are stable. Modern work isn’t predictable—it’s continuous problem-solving requiring judgment, creativity, and rapid adaptation.

AI agents will intensify this reality by accelerating iteration and compressing decision cycles. If humans must route every meaningful choice through approval chains, the organization becomes its own bottleneck. You can have the most advanced tools in the world, but if your people lack authority to use them when decisions need making, you’ve bought a sports car and governed it to thirty miles an hour.

The alternative is coherence—an operating model where independent actors, human and digital, can act intelligently without constant escalation because the destination and constraints are clear, and information moves quickly enough for local decision-making to be good enough.

This rests on three conditions leadership teams can build:

One outcome everyone can see. Not a stack of KPIs, but a shared definition of success vivid enough to guide tradeoffs in the moment. When people understand what success looks like and why it matters, they don’t need a manager in every decision.

Clear lanes, not constant approvals. Speed comes from freedom with boundaries. Define what teams can decide, what they can spend, and when they must escalate, then let them execute. Constant approvals feel like safety but create latency and helplessness.

Truth that travels fast. Coherence collapses when reality gets filtered on its way up. Build short feedback loops, make work visible, and remove penalties for naming what’s actually happening. You cannot adapt to what your people are trained to hide.

The Leadership Move That Matters Now

The wrong question is “How do I get buy-in?” That framing assumes you’re convincing people to follow a plan developed without them.

The right question is “What conditions let my people coordinate, tell the truth, and solve problems without me in the middle?” Because the future everyone is being promised by AI only arrives if people feel safe enough to actually use these tools—safe enough to experiment without being punished for early drafts that don’t work, safe enough to name risks without being labeled as negative or resistant, safe enough to redesign workflows without triggering a political fight over territory and headcount.

Without that environment, the most likely outcome isn’t transformation. It’s churn, dissatisfaction, and triage where leadership is always one crisis behind.

Control can create order for a moment. Coherence creates conditions for real adaptation. Autonomy is the engine that turns potential into performance.

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.