Strategy dies in the middle if managers do not know what is allowed.

Strategy dies in the middle if managers do not know what is allowed.

They sit between the slide and the work. People ask them about job loss, about the tool they already opened, about whether this email can go through a model. If they do not know the line, they stall or they guess.

Three tiers. No approval: ChatGPT for a brainstorm, Grammarly, a summary of an internal note. Manager yes: a new use of a tool you already have, a customer message, a new kind of file. Leadership yes: a new vendor, hiring, customer data. Write that on one page. The guesswork is what slows you down.

A short list beats a 40-page policy

No proprietary text into a consumer chat. No customer data without a yes. No hire or fire on a score alone. Yes: first drafts, ideas, long reports, meeting notes. Then they can answer "can I write this client email with it" without a committee.

Let them use the tools on their own pile: a week of mail, a schedule, a sheet. Talk about hallucinations and bias with examples, not slogans. Give them a way to send a new use or a scare back up. High chance and high harm: stop it. High chance and low harm: coach the prompt. Low chance and high harm: know who to call. The rest, ignore.

One or two managers first

A page or three: the tiers, the yes/no list, who to call. A flowchart for a new tool. A checklist before anything leaves the building. Refresh it quarterly with them in the room.

If someone used it badly, ask why. Often it is a skill hole, not a crime. Share the clever uses. Let them approve small experiments inside the wall. A shared sheet of requests and misses, reviewed monthly, tells you what to buy or teach. Start with one or two people. They have their own fear. Deal with that first. Then they can carry the rest.