Short answer: Start with simple approved-tool, data and ownership rules that a small team can follow. Start by making the decision explicit, naming the accountable person and keeping sensitive information out of unapproved tools.
What this helps you decide
Adapt the editable policy starter for the work and data this team handles. The useful outcome is not a generic statement; it is a documented rule that someone on the team can apply to the next real task.
Practical steps
- List the approved AI tools and their owners.
- Name the data that must never enter an unapproved tool.
- Require a human to review external, financial, legal or employment-impacting output.
- Set one person and one date for the next policy review.
Keep the boundary clear
AI can assist with drafts, summaries and structured thinking. It should not quietly become the final decision-maker for matters involving people, confidential information, legal commitments, finance, eligibility, employment or public claims. Escalate when the impact or data sensitivity increases.
Related resources
Updated 2026-08-29. Educational resource only. Adapt with the appropriate owner or adviser for your context.