AI for managers: what to do when half your team is using ChatGPT.

If you manage a team in any white-collar function, some of your team is already using AI tools. They may or may not have told you. They may or may not be using approved tools. Your role is to make this work, not to pretend it is not happening.

Three things that are true at once

First, AI raises the productivity of any individual who learns to use it well. Your best people will compound their output. That is good for the team.

Second, AI raises the floor of acceptable work. Outputs that used to be "good enough" are now obviously below the bar. That changes how you evaluate the team.

Third, AI introduces new failure modes. People will quote AI hallucinations to customers. People will paste regulated data into consumer tools. People will let AI think for them and lose the skill they used to have. You are accountable for managing all three.

The three conversations to have with your team

Conversation one: what AI tools are sanctioned, for what data classes. If your organization has no policy, write one yourself for your team, even if it is provisional. The policy can be one page. The classes are: public data (any tool), internal data (approved enterprise tools only), regulated data (specific approved path or none), secret data (no AI tools).

Conversation two: what good AI-assisted output looks like. If your team uses AI to draft a memo, the memo is still your team's work. The standard for review is unchanged. The team member who hits "send" on AI output is accountable for the output. The team member who hits "send" without reading is going to embarrass themselves and you.

Conversation three: where AI is appropriate, where it is not. AI is appropriate for drafting, summarizing, brainstorming, restructuring, analyzing content you have. AI is not appropriate for legally binding final text, for confidential strategy work in consumer tools, or for decisions that require judgment your team is paid to exercise.

What to do this quarter

1. Inventory what AI tools your team is using right now. Ask, do not assume. 2. Write a one-page policy. Get your manager and your security team to ack it. 3. Pick one workflow per direct report and make it the place to practice AI use deliberately. Measure the time saved. 4. Set a standard for AI-assisted output review. It is your output. You read it. 5. Notice who is compounding. Reward them. Notice who is not adopting at all. Find out why before assuming.

This is now a normal part of managing a knowledge-worker team. The managers who do this well in 2026 will run higher-performing teams in 2027.

The LearnTrainAI for Enterprises program is built specifically for this transition.