AI adoption is a manager problem too

A company buys an AI tool.

Everyone gets access. There’s a demo, an enthusiastic announcement, maybe a channel where people share prompts.

Then a few people run with it, a few try it once, and a lot of people quietly get back to doing their jobs the old way.

It’s tempting to conclude that the team needs more training.

Sometimes they do.

But I think we underestimate how much AI adoption depends on the people managing the work.

People need more than permission

“Feel free to use AI” sounds encouraging. It also leaves an enormous amount unanswered.

What are we actually trying to improve?

Which information is safe to use?

Where is AI allowed to assist, and where is a human decision required?

How do we know the result is good?

And if somebody finds a better way of working, do they have time and support to share it with the team?

Without clear answers, experimenting becomes extra work that people fit around their real work.

The team member who’s naturally curious will probably find a way. The busy manager may never know what changed. And the organization doesn’t really learn.

Don’t make everyone reinvent the same workflow

Imagine three recruiters discover useful ways to prepare interview guides with AI.

One gets excellent results. Another keeps starting over. The third stops using it because checking the outputs takes longer than writing the guide.

You could tell them all to become better at prompting.

Or the manager could bring them together, compare approaches, agree on a quality standard, decide what data can be used, and make the useful version part of the process.

Now you’re doing something different.

You’re not just improving individual AI skills. You’re improving how the team works.

That distinction matters beyond recruiting. It applies to project planning, client work, operations and internal support.

A simple job for the manager

I don’t think every manager needs to become an AI expert.

But I do think managers need to get comfortable asking four questions:

What outcome are we improving? Not “how many prompts did you write?” but what got clearer, faster, safer or better.

What stays human? Define decisions, relationships and escalation points before the workflow spreads.

How do we check quality? Agree on examples of good work and failure, then test them.

How will the team learn? Make space to share a useful method, fix problems and stop using tools that aren’t helping.

There is a technical side to AI adoption, absolutely. There is also a management side that tends to get less attention because it’s harder to turn into a demo.

Microsoft’s 2026 Work Trend Index points to the same broader challenge: organizations need the right support, culture and operating practices for people’s AI capability to translate into value. That’s a useful reminder, but I don’t think we need a report to recognize it in daily work.

The part I’d pay attention to

The most impressive person in a team might build a brilliant shortcut.

That’s great. But if the rest of the team can’t understand it, trust it or use it, you’ve created an individual advantage, not an organizational capability.

The better question is whether the team is learning to work differently together.

A tool can help someone move faster.

Managers help make the improvement stick.

Reference for the research point: Microsoft, 2026 Work Trend Index, https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization