The one-page AI playbook every small team can write

When everyone experiments separately, useful lessons can disappear into private chat histories. A short shared playbook gives your team somewhere to record what works, where it fails, and who makes the final call. It should be a working document grounded in your own process.

Describe approved uses in everyday language

Write down tasks people can recognise: restructuring an internal outline, summarising approved notes, or generating alternatives for a headline. Name the source material each task may use. Avoid a vague instruction to “use AI responsibly” when you can describe an actual boundary.

Make the review visible

For each task, identify who checks the output and what they check. A writing draft might need a facts-and-tone review. A spreadsheet formula needs to be tested against known inputs. Keep anything that sends, publishes, or changes records behind the review your team already uses.

Use this structure

• Task: What is AI helping us produce?
• Inputs: What information can be used?
• Reviewer: Who is responsible for the finished result?
• Checks: What mistakes must we look for?
• Learning: What changed after the last attempt?

Add one concrete example

Include a sample brief and a reviewed output with sensitive information removed. Show one correction that mattered. This makes the playbook easier to learn from than an abstract set of rules and gives new team members a realistic picture of the review effort.

Keep it short enough to maintain

Review the page when the tool, task, or team changes. Remove advice that no longer matches your process. A useful playbook is not a guarantee that every output is safe or correct; it is a shared starting point for making decisions and learning from mistakes.