AI Policy Launch Pack

Example: from an AI request to a recorded decision

This fictional example shows how to use the included documents together. It is not a customer testimonial, a vendor assessment, or a compliance determination.

A 35-person consulting firm wants to let five employees draft internal meeting agendas with an AI assistant. Maya, the operations lead, owns the rollout; Alex, the IT lead, reviews the tool.

1. Record the request. In the tool approval form, Maya records request AI-001, the proposed business plan, the five pilot users, the purpose, and the proposed inputs: public information and invented examples. Client files, personal data, and credentials are outside the request.
2. Review before deciding. Alex records the vendor documents examined, their dates, the relevant data-use and retention terms, access settings, and any unanswered questions. Missing evidence keeps the request pending; a policy template does not resolve it.
3. Log the decision and restrictions. Assuming the reviewers resolve the required checks, Maya records a 30-day restricted pilot in the approved tools register, linked to AI-001. Only the named users and reviewed plan are in scope; inputs stay within the permitted categories and a person checks every output before use.
4. Publish the rules and train users. Adapt the employee policy to match those restrictions. Use the literacy material to explain safe inputs, output review, and incident reporting; record the five users’ training and acknowledgements in the rollout records.
5. Set the review. In the launch tracker, assign Maya a review 30 days after the pilot starts. Check incidents, actual usage, vendor changes, and open questions, then record whether the pilot continues, changes, or stops. A new data category requires another review.

The resulting record connects one request, its evidence, an accountable decision, employee instructions, and a review date. Your organization supplies its own vendor findings and approval judgment.