AI Adoption Is a Habit Formation Problem

A few months ago I finished reading Kathy Milkman's book How to Change. Milkman is a behavioral scientist at Wharton who has spent her career studying why people struggle to change their behavior, even when they want to and know it would benefit them. The book is full of research-backed tools for closing the gap between intention and action.

I wasn't reading it for AI adoption. But by the time I finished, that's mostly what I was thinking about.

The Problem with AI Rollouts

When organizations bring AI tools into the workplace, the approach tends to follow a familiar pattern. Choose a tool, purchase licenses, schedule a training session, communicate the benefits. Then wait for people to start using it.

The assumption underneath this approach is that adoption is mostly an information problem. If people understand what AI can do, and have access to the tools, behavior change will follow.

It usually doesn't.

Milkman's work offers a clear explanation for why. She argues that lasting behavior change requires more than knowledge or motivation. It requires designing around the actual psychological obstacles that get in the way: procrastination, present bias, forgetfulness, inertia. These aren't character flaws. They're predictable features of how humans operate. And they don't disappear just because someone attended a training.

The gap in most AI strategies isn't the technology. It's the absence of any real design for behavior change.

Two Tools Worth Borrowing

Milkman's book covers a range of strategies. Two of them map almost directly onto the challenge of building AI habits inside a team.

Temptation bundling is the practice of pairing something you genuinely enjoy with something you keep putting off. Milkman developed the concept from her own experience as a graduate student who struggled to make it to the gym. Her fix: she only let herself listen to audiobooks she loved while exercising. The thing she was avoiding got a hook. She started showing up for the audiobook and getting the workout.

The principle translates cleanly. If someone on your team loves a particular playlist or podcast, they only get to put it on during their 15-minute AI experimentation block each week. Or a team ritual people already enjoy, like a Friday standup or a team lunch, becomes the consistent context where someone shares one AI experiment from the week. The point is to attach the new, effortful behavior to something that already has pull.

The fresh start effect is the finding that people are significantly more motivated to pursue new behaviors after natural transition points: a new year, the start of a new month, the beginning of a new quarter, a birthday, a new job. These moments create a psychological break from the past. People feel less defined by previous failures or old habits. The slate feels cleaner, and that makes change feel more possible.

For AI adoption, this means timing matters. A new team member joining is one of the best on-ramps available. Onboarding is already a moment of learning new norms and building new routines. Introducing AI habits there, before old patterns are entrenched, is genuinely easier than trying to retrofit them later. A new quarter, a team reorg, the start of a big project: these are all real openings.

What This Changes

Neither of these strategies requires a new platform, a bigger budget, or a more sophisticated tool stack. They work with the way people already think and move through their days.

That reframe matters. AI adoption is often treated as a change management problem at the organizational level, which leads to top-down communication campaigns and company-wide mandates. Those approaches have their place. But they skip past the level where habits actually form, which is the individual person, in their daily routine, deciding whether to reach for the AI tool or do it the old way.

Milkman's research points toward a different design question. Rather than asking "how do we get people to understand the value of AI," the more useful question becomes "what would make it easier, and maybe even enjoyable, for someone to try AI in the flow of how they already work?"

That's a human-centered question. And it tends to produce better answers.

A Practical Starting Point

If you're thinking about how to build AI habits on your team, here are two concrete entry points drawn from Milkman's framework.

Find something people already look forward to and attach an AI experiment to it. Make the bundle small and low stakes. Five minutes, one use case, no pressure to report results.

Then look at your calendar for natural fresh starts. New quarter, new hire, end of a big project. Pick one and treat it as an intentional on-ramp. Make the first AI habit easy to form before the window closes.

Behavior change is hard. But it's not random. Milkman's work is a reminder that when we design for how people actually work, rather than how we wish they would, the results tend to stick.