The 30-day AI checklist

Before you start

• Name one owner for the project
• Block two hours a week in their calendar
• Agree that one process may change

Week 1: Pick and measure

• List the team's repetitive weekly work
• Score each: often, rule-based, real hours, cheap to check, measurable
• Pick one process
• Write down today's number (hours, days or errors)

Week 2: Map and prepare

• Write every step as it really happens
• Mark each step: rule, judgment or hand-off
• Choose one source of truth for the data
• Collect 20–50 real examples to test with
• Decide what AI does and what a person approves

Week 3: Pilot

• Pick the lightest option: assistant, automation or custom tool
• Run it on your test examples
• Run it on live work with a person checking every result
• Log what went in, what came out, and whether it was right

Week 4: Roll out and decide

• Write a one-page how-to for the team
• Compare results with the week 1 number
• Fix the cases it got wrong
• Decide: keep, build properly, or move to the next process