What you can still do when the AI policy says no
Most company AI policies do not ban thinking. They ban sending project files, drawings, and contract language to a tool the company does not control.
Read the policy as a data rule, not a work rule. If you are not sure, ask before you paste.
What you can still do
Use the tool your company already pays for, inside its own account. That is usually Microsoft Copilot on the company tenant, or ChatGPT Enterprise or Team if they bought it.
Those keep the files in the company's environment and, on the business plans, do not train on your chats. The free chatbot in a browser tab is the one the policy is written against.
Use AI on material that is not the project. A blank agenda, a generic checklist, the wording of a lesson learned with the names and the job taken out. You can get a clean first draft without handing over a live file.
Use it to structure your own notes after you have removed the identifying parts. "Turn these bullets into an action list with an owner, an action, and a date" is allowed on almost every policy, as long as the bullets are yours and not the owner's confidential set.
What you do not do
Do not paste a drawing, a contract, an RFI log, or a pay app into a personal ChatGPT, Gemini, or Claude account to just summarize it. That is the violation, even if the summary is good.
Do not trust a tool to decide. It can draft the line. You still name the owner, the action, and the date, and you still send it.
One thing to ignore
Ignore any post that says a new model can review your whole project folder. If the folder cannot sit inside a tool your company has approved, the demo does not apply to you.