The AI Rollout Nobody Warns You About Is a People Problem

A client called me this week in the middle of a six-week trial for a new AI tool inside his legal department. The CEO wants it. Adoption looks likely. And he's the one running the trainings, building the relationships, onboarding a new hire so he has time to manage the thing.

Here's what stood out. He wasn't worried about the software. He was worried about the two deputies who don't want it, the boss whose job he's doing without the title, and the fact that nobody has officially said who owns this.

That's the real work of an AI rollout. The tool is the easy part.

Why do most AI tools get bought and then ignored?

Because someone treats the launch like a training session. Sit everyone down, show them the buttons, send a recording, call it done.

Then Monday comes and everyone goes back to the way they already work. The tool sits there. Six months later someone asks why nobody uses the thing they paid for.

The reason is simple. You didn't change how the work happens. You added a tool on top of a workflow and hoped people would rearrange their day around it. Most won't. They have deadlines and habits and no reason to trust that the new way is faster.

The ones who get adoption right redesign the workflow first. Where does the contract start, who touches it, where does it slow down. Then they put the tool at the exact point where it removes a step people already hate.

Who has to want this for it to work?

Not everyone. And they don't all want the same thing.

My client mapped his three decision-makers before he built anything. The top lawyer cares about risk and keeping the peace. The operations side cares about speed. The litigation group is scared of being left out. Same tool, three different reasons to say yes.

So you don't sell one pitch. You build a version of the case for each person that speaks to what keeps them up at night. The risk person hears about fewer mistakes. The speed person hears about a 20% cut in contract cycle time. The left-out group hears they're first in line.

Most people running a rollout skip this. They send one email to the whole department and wonder why half the room is cold. You're not talking to a department. You're talking to five people with five sets of incentives, and you need most of them nodding before the group moves.

How do you get put in charge when nobody named you the owner?

You don't wait for the title. You ship something.

My client's instinct was to push for a formal role first, get the org chart sorted, then start. I told him that's backwards. The authority follows the work, not the other way around.

Pick one small team. Build a quick prototype that solves one thing they complain about every week. Let it work in front of people. Now you've got proof. You already fixed something they watched break for months, and asking for room to do it again is a much easier conversation.

He's got a boss sitting on the AI steering committee who isn't doing much with the seat. So the plan is to ask for that seat directly and start bringing prototypes instead of slides. Same politics, different position. You stop being a name on a request and become the person the room already ties to progress.

What breaks a rollout that looked like it was going fine?

Politics nobody put on the calendar.

The trial had CEO support and a real shot at full adoption. On paper, done deal. In the room, two deputy general counsels and my client's own manager were stepping on each other behind the scenes over who runs the change. That kind of friction doesn't show up in a status update. It shows up as a launch that keeps slipping for reasons nobody will say out loud.

The move there isn't to fight it head-on. Make the legal team the visible face of the rollout and use his manager's backing to smooth the parts he can't control directly. Give the credit away where it buys you cover. Keep the CEO sponsorship warm so when the friction gets loud, someone above it all still wants this to happen.

The part most people get wrong

They think buying the tool was the decision. It was the easy 10%.

The other 90% is figuring out whose day changes, who feels threatened, who needs to look good, and how you get twelve people with different jobs to agree on a new way of working. No vendor demo covers that.

My client is spending his next month offloading his old tasks to a new hire so he can chase complaints across the department and turn each one into a quick win. Not a launch plan. A list of small, specific things that make one person's Tuesday easier, built one at a time until the tool stops feeling like something IT installed and starts feeling like how the place runs.

He's calling the top lawyer right after the vendor session ends to pin down what the man wants out of this. Everything else waits on that answer.