Start AI adoption with the work people dread

August 1, 2026 · Restorative · 2 min read

A university professor described his department: a few AI evangelists, and everyone else dreading it. Mandates from the top don't fix that. Start with the boring work people hate.

I was talking to a university professor this week about why AI keeps stalling in his department.

His picture of it: a handful of evangelists who use AI in genuinely clever ways, and then everybody else. Most of that everybody else has no clue where to start. Some of them dread it, because the version of AI they've been sold is the one that takes their job.

Into that room lands a mandate from the top. Use AI. No target, no starting point, no permission to be bad at it for a while. The evangelists carry on doing what they were doing anyway. Everyone else waits for it to blow over.

Lower the task granularity

A big ambitious AI project asks people to commit before they trust anything. So make the first ask small enough that trust isn't required.

Start with the work you hate and that is boring. Meeting notes nobody writes up. The same three paragraphs of grant boilerplate every quarter. Reformatting a spreadsheet somebody else exported wrong. Nobody defends that work, nobody's identity is built on it, and everyone can tell within ten minutes whether the output was any good.

Look for evangelists at the bottom

The clever users are already there and they're rarely the people with a mandate in their job title. Find them, name what they're doing, and give them room to show a colleague. Usage spreads sideways far better than it comes down. A peer showing you the thing they built to skip an hour of dread is a different conversation than a directive from the top.

Connect what already exists

Most organisations have more AI in use than they think, scattered across people who never mentioned it. Map that first. Two people automating adjacent parts of the same process is where compound effects start, and it costs nothing to find.

Where to start

Pick the task in your week you actively dread, the small boring one, and hand that to AI first. Check what comes back yourself before you trust it with anything bigger.