AI Transformation Doesn’t Just Get Stuck with Individuals; It Gets Stuck Between People.

I’ve written before that the first question in AI transformation shouldn’t be which tool to deploy. It should be: What should the work actually look like?

There is a next question that I’ve been thinking about. Once someone finds a better way to work with AI, what does it take for that to become the way the organization works?

One person can change how they work fairly quickly. They can experiment, figure out where AI helps, decide where their own judgment matters, and adjust how they work.

It gets harder when that new way of working reaches everyone around them.

Most work doesn’t happen in isolation. Change how one person works and you may change what a colleague receives, how a manager reviews the work, who makes a decision, or what another team is expected to do.

We tend to assume that if enough people discover better ways to work, those improvements will naturally spread. I’m not sure they do. What one person can change may require a team to agree to work differently. And what a team wants to change may depend on a decision they don’t have the authority to make.

I’ve been thinking about this through three questions: Where do people need agency? Where do we need agreement? And where do we need authority?

Agency: What can I change myself?

People closest to the work need room to experiment, learn what works, and use judgment about where AI adds value and where a human still needs to lead.

When someone finds a better way of doing something, managers should go beyond asking, “How are you using AI?”

Ask: What are you doing differently? What is working better? What have you learned that should change how the rest of us work?

That last question matters because what one person learns may have implications for the work around them.

Agreement: What do we have to change together?

Most work moves between people. Someone provides an input, someone reviews an output, someone makes a decision, and someone downstream depends on what came before.

Say someone starts using AI to create the first draft of a recurring analysis. What used to take hours now takes a fraction of the time, and the person has learned where they need to review, challenge, and add judgment.

Now look at what happens around that work. Does the same review process still make sense? Does the downstream team need the same output? Has the quality standard changed? If the analysis informs a decision, who remains accountable for it?

These aren’t necessarily decisions that need to go up the organization. The people who share the work may simply need to agree on a different way of working together.

A useful question is: If we worked this way from now on, what else would have to change?

Follow that question through the workflow. Look at the handoffs, reviews, expectations, and decisions around the work.

The goal isn’t to get everyone using the same tool in the same way. It is to make sure one person isn’t working in a new way while everything around them continues to operate as before.

Authority: What can’t the team change on its own?

Sometimes a team agrees on a better approach and still can’t implement it.

An approval sits somewhere else. A policy assumes the old way of working. A workflow crosses functions. Or nobody is quite sure who has the right to make the call.

At that point, someone needs to make a decision.

A team should be able to say: Here is what we want to change. Here is what is blocking us. Here is the decision we need. And here is who needs to make it.

Otherwise, good ideas can sit in limbo because no one owns the next decision.

This is also where the People function has a role beyond building AI capabilities. As the work changes, leaders need to think about expectations, accountability, manager practices, and decision rights.

And once a decision is made, managers need to translate it into everyday work. What should people do differently? What can they stop doing? What standards still apply? Where does accountability now sit?

Start with one workflow

You don’t need to map the entire organization to figure out where AI-enabled change is getting stuck.

Start with one workflow where people are already using AI and bring together the people who actually touch the work.

Ask:

Agency: What can people change themselves?

Agreement: What needs to change across the people who share the work?

Authority: What can’t they change, and who can?

Then ask: If this became the way we worked tomorrow, what else would need to change?

Push past answers like “the process” or “the system.” Is it a handoff that no longer makes sense? A review that assumes the old way of working? An approval nobody has revisited? A policy, ownership boundary, or decision right?

Finding a better way to work is only the beginning. The harder part is making it possible for that new way of working to spread beyond one person.

Start with one workflow and ask: Where are we stuck- agency, agreement, or authority?

The answer tells you what needs to happen next.

Author Bio Meghna Punhani is Chief People Officer at Eightfold AI and a business executive working at the intersection of AI, strategy, organizational design, and leadership. With over two decades of experience—including nearly 17 years at Google across People, Strategy & Operations, and Corporate Engineering, as well as executive leadership at Palo Alto Networks—she partners with CEOs and boards to help organizations scale and innovate. At Eightfold, Meghna leads the global People organization and positions the company as its own "Customer Zero," leveraging AI to redesign talent management and operating models. She views AI not merely as a tech upgrade, but as a core business transformation redefining how modern organizations adapt, make decisions, and thrive in an AI-first world.