AI Changed the Work. Who Changes the Organization?

AI can change a task in an afternoon. Changing who owns the outcome takes a leadership decision.

That is where many businesses get stuck. People learn the tools. Teams build useful solutions. Work moves faster. But decisions still follow the same approval chain, departments still chase separate priorities, and nobody has decided what to do with the time AI frees up.

The technology changed. The organization didn’t.

McKinsey’s 2026 State of AI survey puts a number on it: 80% of respondents say AI has improved their individual productivity, but only 37% report a positive impact on earnings before interest and taxes (EBIT).[1, p. 2] The tools are working. The business isn’t capturing all the value.

In The Real AI Gap, I wrote about why adding AI to existing processes is not the same as becoming AI-native. This is the next question: once the work can change, what else has to change around it?

Not everything. But more than the workflow.

A shared solution exposes an organizational question

Many of the teams I work with are still siloed. The people closest to the work have great ideas, but their tools, processes, and priorities pull them in different directions.

That shifts when subject matter experts build the tools they actually need. When I show teams what’s possible with Replit, they stop describing workarounds and start building toward shared outcomes, no longer boxed in by software that never quite fit their business.

That’s a strong start. It also forces harder questions. Who owns the solution across departments? Who can change the process? Which decisions can an agent support, and which need a person? What will leaders expect people to do differently?

Leave those unanswered and the team can build something new and still work the old way.

The McKinsey data shows an association: among AI high performers, nearly three-quarters have substantially redesigned their workflows. Among everyone else, it’s one-quarter.[1, p. 2] That does not establish that redesign caused higher performance.

That doesn’t mean every company needs a sweeping reorg. It means organizational choices belong in the build, not in an announcement after it. And the people the change affects should make them.

Three decisions that turn a working tool into a different business

1. Decide who owns the outcome

Here’s an example from a customer service team I worked with. An agent gathered everything needed to resolve a customer issue in minutes. Then the resolution waited for three departments to approve their pieces.

The team thought they needed a better agent. They didn’t. They needed one leader to make a clear decision and the freedom to work differently to produce the desired outcome. We found out who could resolve the issue end to end, which approvals were needed to protect the business and the customer, and which departments were involved only because no one used to have enough information to act.

This is where workflow design becomes organizational redesign. Roles, team boundaries, decision authority, and accountability have to fit the outcome.

Sometimes that means restructuring a team. Sometimes it means giving an existing team clearer authority. Sometimes it means removing a handoff while strengthening human review at the decision that matters.

This is one small example. The goal is a business that works better and is aligned on outcomes.

2. Decide which behaviors the new model requires

AI fluency matters. People need to understand what the tools can and can’t do, and which work is still theirs to own.

But knowing how to use a tool is different from being supported to work differently. McKinsey’s AI readiness research found that 70% of respondents say they’re personally ready for AI. Only 27% of leaders say their organizations are ready to make the shifts required for an agentic future.[1, p. 7]

That gap shows up in everyday signals. If leaders reward individual output while asking for cross-functional collaboration, people get conflicting instructions. If leadership encourages experimentation but treats every controlled test as a failure risk, teams learn not to experiment. If employees are told to use agents but must recreate every step manually to prove they worked, the old process stays in place.

Behavior change needs reinforcement in the daily work.

Leaders model the choices they expect. They coach people through real work. Subject matter experts test and improve solutions together. Teams share what works instead of solving the same problem separately. People can fail safely because failures become lessons.

Delegation needs boundaries, too. Handing a task to an agent doesn’t hand off accountability. People need to know what to review, when to intervene, and where human judgment is needed.

The question is not simply, “Did everyone complete the training?”

It is, “What are people doing differently?”

3. Decide where the capacity goes

Resource allocation happens before the build and after it.

Before the build, leaders decide which priorities get time, talent, and investment. A roadmap without those commitments is a list of competing intentions.

After the build, leaders decide what to do with the capacity AI releases. Faster customer response. More time for complex decisions. Room to test a new service. More sustainable workloads. The right choice depends on the business, and it’s important to make the choice deliberately.

Otherwise, saved time disappears into another queue of low-value work. The dashboard records productivity. Customers and employees notice little difference.

Measure whether the intended change happened. Did turnaround improve? Did quality improve? Did the team finally clear the customer work it had been postponing? How many handoffs disappeared? How long does a cross-functional request take now?

Hours saved are a proxy. Track the outcome.

Start with one outcome

You can’t reorganize the business around every AI project, and you shouldn’t try.

Start with a meaningful outcome and the people responsible for it. Identify the organizational choices blocking progress there. Change those alongside the solution, then learn from what happens.

I structure the work I do with customers in five connected stages: Diagnose, Align, Redesign, Implement, and Value.

Diagnose reveals the constraints. Align settles ownership, decision rights, and commitments. Redesign shapes the work and the responsibilities around it. Implement puts the solutions and new behaviors into daily use. Value tests what changed and decides where the released capacity goes next.

The three decisions are not linear. Behavior work starts before implementation. Resource allocation doesn’t end when budget is approved. Measurement doesn’t begin after launch. The stages give direction without pretending change happens in a straight line.

Leadership has to give permission to work differently

AI can give a team bandwidth it did not have before. Leadership decides whether that bandwidth goes to meaningful work.

That takes more than buying the technology or approving the roadmap. It means changing the decisions, expectations, and resource commitments that shape the work every day. Keep what works. Redesign what gets in the way. Back the people you’re asking to work differently.

Pick one AI initiative already underway. Ask its leaders three questions:

• Who owns the outcomes?
• What behavior needs to change?
• Where will the capacity go?

If nobody can answer them, that’s your next project.

Source

[1] Alexis Krivkovich, Brooke Weddle, Holly Price, and Vik Sohoni, AI is changing work. Now it has to change the organization, McKinsey & Company, October 2026, pp. 2, 7.