AI Policy Generator

A policy that fits on one page is a communication tool. A policy that runs to five pages is a legal document nobody reads. The skill enforces the page limit because the whole point is that a client, prospect, or procurement reviewer can read it in ninety seconds and walk away with confidence rather than a stack of caveats.

The POLICY framework runs the build. Principles statement — two or three sentences on what you believe about AI in your work, written confidently rather than apologetically. Operations detail — specifically what AI assists with and what tools you use, named or categorised. Limitations and commitments — what you won't do, including the human-only commitments that are the section a prospect actually rereads. Intellectual property — a short statement on ownership of AI-assisted work product. Client rights — what they can ask, request, and expect. Your commitment — a forward-looking close that dates the policy and commits to updates.

The "what remains human" section does most of the trust-building. "We ensure high-quality human oversight" is vague enough to mean nothing. "Strategic recommendations, creative direction, and all client-facing communications are produced entirely by our team — AI is never used to make decisions about your business" is specific enough to be credible. The skill pushes for at least three concrete commitments here because that's the section a prospect actually reads twice.

You get the one-page policy with all six POLICY sections, written in the agency's voice rather than legal-template English. Data protection is addressed specifically — does client data enter AI tools, which ones, with what agreements — because "we take data security seriously" is the line that signals you're hiding something. Solo freelancers get a shorter "My approach to AI" variant for their website instead of the full document. Multi-department agencies get acknowledgement that different teams have different AI profiles, rather than a single policy that papers over the differences. Every output is dated and versioned because the practice will evolve and the policy needs to evolve with it.