When a team starts using AI tools, the leader's first instinct is usually to do more. More check-ins, more reviews, more dashboards, because so much more is happening.
That instinct is understandable, and it's usually wrong. When the work speeds up, the habits that made someone a good manager start to slow the team down. In The 5X Effect I put it this way: the hardest part of building a high-performing team often isn't what the team does. It's what the leader stops doing.
Here are five habits that cost more at AI speed than they ever did before.
1. Keeping every option open
Most leaders can tell you what the team should build. Very few can tell you what the team is allowed not to build (i.e., what's not in scope).
That gap was expensive before. With AI, it's worse. An AI tool builds whatever the description implies, and if nobody has ruled something out, it often ends up in scope. A team without a clear "not this" list hedges against every possible request, and now it can hedge much faster.
Instead: cut your goal down to a sentence short enough to remember, and write down what's out of scope. Expect everything you leave out to feel important. Leave it out anyway. Its easy to add stuff later. Its nearly impossible to take it out.
2. Taking problems back
A developer brings you a problem. You know the answer, and it's quicker to just solve it. So you do.
Every time, the team learns that problems belong to you. At AI speed, problems come up many times a day, and a leader who takes them back becomes the place where all of them wait.
Instead: when someone brings you a problem, ask what they think should happen. If their answer is reasonable, let them do it, even if you'd have done it differently.
3. Filling the silence
You ask the team a question. Nobody answers right away. After a few seconds, you answer it yourself.
That silence is where people decide whether their opinion matters. Fill it, and they learn to wait for yours. A team that waits for the leader's opinion can't keep up with work that moves this fast.
Instead: ask questions you don't already know the answer to, and wait. Count to ten if you have to. The answers you get after the silence are usually better than the ones you would have given.
4. Counting work as done when it's handed off
In most organizations, "done" means "I handed it to someone else." The developer is done when it goes to review. The reviewer is done when it goes to test.
With AI writing the code, there's far more work to hand off, and far more of it piles up waiting for the next person. A team only finishes what its leader checks. If you count handoffs as progress, the team will produce handoffs.
Instead: check, every day, that finished work is really finished: tested, reviewed, and ready to use. Ask about what was finished, not what was started.
5. Judging progress from reports
When there's more going on, it's tempting to manage from a distance: more status reports, more dashboards, more metrics.
Reports tell you what people decided to put in them. They rarely show you where the work is actually waiting, or why. You find that by sitting in the team's meetings, watching the work move and asking questions.
Instead: spend time where the work happens. A few hours a week watching the team work will tell you more than any dashboard.
What you keep
None of this means the leader steps away. You still decide where the team is going and why. You still decide what gets built and what doesn't. You still protect the team from changing priorities and reshuffled membership, so it has time to get good at working this way.
What you give up is deciding how the work gets done. That's the trade. And it's the one most leaders find hardest to make, because for years, deciding how was the job.
If you're not sure which of these habits you rely on most, the Leadership Self-Check takes three minutes and shows which part of leading is holding the rest back. It's free, and you'll see your result right away.