Writing a Policy People Will Follow
Most AI policies are unread and unfollowable. What a usable one contains, and what it leaves out.
Policy
Rules about tools are obsolete within a quarter. Rules about information survive.
7 notes
Most AI policies are unread and unfollowable. What a usable one contains, and what it leaves out.
A short list people can remember beats a comprehensive one they cannot. How to decide what goes on it.
Writing the rule so that it still works when the tools change, which they will, every few months.
The short list of things that should be flatly forbidden, and why keeping it short is what makes it effective.
A policy with no workable exception route is a policy people route around. What a good one looks like.
Coding assistants raise questions the general policy does not answer, and developers will have adopted them already.
This policy goes stale faster than any other you maintain. What triggers a review and what a review actually examines.