Insights
How Managers Should Spend AI Time Gains
Two managers on the same floor were given the same gift this year: about six hours a week. Both lead teams of twelve. Both adopted AI in the winter. BetterUp Labs, studying what actually happens to managers after adoption, found that AI returns roughly six hours a week to the average manager. What happens next is where the two floors diverge.
What happens next
The first manager did the sensible thing. He used the time to make existing work better and to catch up on administration. In the aggregate data, that is where most of it goes — around 42% into improving work already underway, 33% into admin. Both are defensible. Neither moved his team’s performance with AI at all.
The second manager spent the hours on her people, her own thinking, and the questions nobody had time to ask. In the same research, managers who reinvested that way saw their team’s AI performance rise by 65%.
There is a harsher finding underneath. Some managers used the tool to absorb the conversations themselves — the mentor check-in routed to a chatbot, the feedback delivered as a generated summary, the career discussion politely automated. Where that happened, team coordination fell 12%, burnout rose 26%, and intent to leave rose 29%. The efficiency was real. The organization paid for it somewhere else.
So what does the second manager actually do on
She does not run the work. The work largely runs itself now. She spends the morning reading, not reviewing — going through what her team has written down about where they got stuck, what they noticed, what question they could not let go of. She is looking for the two or three observations that nobody has authority to act on yet.
She takes one of them to a person who has never been asked for an opinion on it, because different histories produce different answers.
Before the difficult conversation at three, she uses AI — not to write the message, but to rehearse. To think through how it will land. Then she has the conversation herself, in a room, badly if necessary.
She protects one hour with no output attached to it. This is the hardest thing on her calendar to defend and the only thing on it that compounds.
None of that appears in a productivity dashboard. All of it is the job.
I do not think this shift is being handled well anywhere. Gallup’s data show the average manager’s team grew from 10.9 people to 12.1 in a single year, while manager engagement fell nine points since 2022 to 22% globally — the steepest decline of any group in the workforce. Managers still spend a median of 40% of their time doing individual contributor work. Research from Harvard Business School on two consulting firms found middle managers quietly absorbing an entirely new job — validating AI output, catching its errors, coaching their teams through it — while delivery pressure stayed exactly where it was.
We handed the middle of the organization a new
The six hours are not a reward. They are a test of what an organization believes a manager is for. If the answer is still throughput, the hours will be spent on throughput, and the strange arithmetic will hold: everyone faster, nothing better.
I suspect the most valuable thing a manager will do in the coming decade is create the conditions in which other people think well — and that this will be almost impossible to measure, easy to cut, and the entire source of a company’s advantage.
The tool gave us the time. It cannot tell us what the time is for.