Guide
AI pilot plan template for a team
An AI pilot that ends in a decision has seven parts: one or two tasks the team repeats weekly, three to six people who do them, a baseline measured for a week before anyone uses AI, success and stop criteria written down before the start, three to four weeks of real use, the rules that apply, and a date when someone decides to roll out, change or stop. Most pilots that fizzle skip the baseline or the criteria, so nobody can say whether it worked.
Short answer
- Narrow: one or two tasks, three to six people.
- Measured: a baseline first, the same measure at the end.
- Agreed: success and stop criteria before the start.
- Decided: a date and a person, with three possible outcomes.
Why pilots fizzle
In Microsoft and LinkedIn's 2024 survey, 59% of leaders worried about quantifying AI's productivity gains, and 60% said their leadership lacked a plan for it. A pilot without a baseline can't produce that number; one without criteria agreed up front turns into a debate about whether 20% is good. The template below fixes both before day one.
The template
| Part | What to write |
|---|---|
| Tasks | One or two, weekly, 20+ minutes, checkable, no forbidden data |
| People | 3 to 6 who do the task, one sceptic, one owner |
| Baseline | One week: minutes per task, one quality signal |
| Criteria | Continue if … · Stop if … (numbers, agreed before the start) |
| Weeks | 1: shared instructions and kick-off · 2–3: real use, logged · 4: measure |
| Rules | Company account; what never goes in; who reads before sending |
| Decision | Date, decider, and what roll out / change / stop each mean for seats and budget |
A worked example: a support team
Illustration: a made-up business. A 6-person support team answers about 300 emails a week at 6 minutes each, with 8% of tickets reopened. The support lead runs a four-week pilot of Claude Team for first drafts.
- Criteria (agreed before): continue if replies take 4 minutes or less in week 4 and reopened tickets stay at 8% or below; stop if payment details go into any AI tool or reopens pass 10%.
- Week 4 result: 3.8 minutes a reply, 7.5% reopened. One near-miss: an agent was about to paste an email with a card number's last four digits, caught by the rule check, which became a line in the shared instructions.
- Decision: the COO rolls out to the whole support team and adds the seats to next quarter's budget.
The decision took ten minutes, because everyone had agreed in advance what "worked" meant.
Run yours
- Pick the task and people, including one sceptic.
- Measure a week of baseline before anyone uses AI on the task.
- Write and agree the criteria, with numbers.
- Prepare shared instructions for the task and the rules, then kick off.
- Measure the final week the same way and decide on the set date.
Next step
If the pilot says roll out, the shared instructions you wrote are the start of your team's skills: the next rung is making them load by themselves. AI skills, explained covers the idea. For the rules and training around a rollout, see the AI usage policy template and AI training for employees.
Skills for this
- AI practiceAI pilot plan for a teamWrites a short AI pilot plan: one or two tasks, the people, a baseline measured first, success and stop criteria agreed up front, the weeks, the rules and the decision at the end.
- AI practiceAI automation opportunity scanScores a team's recurring tasks on frequency, time, rule clarity, and risk to find AI candidates, then picks the top 3 with a first experiment each.
- AI practiceCompany AI use policy draftDrafts a short, plain-language AI use policy: approved tools, data never to paste, review and disclosure rules, owners, and points flagged for legal review.
- AI practiceAI training plan for staffPlans three 45-minute AI sessions on each team's own tasks: the rules and a first task, the shared instructions, and checking AI output, with homework and a training record.
Questions
How do you run an AI pilot for a team?
Pick one or two weekly tasks, three to six people who do them, and measure how long they take for a week before starting. Agree what result means continue or stop, run it for three to four weeks on real work, measure the same way, and decide on a set date.
How long should an AI pilot last?
Three to four weeks of use after a week of baseline: long enough for people to get past the first awkward days, short enough that the decision doesn't drift.
What should an AI pilot measure?
Minutes per task, measured the same way before and after, and one quality signal the team already tracks, such as rework, error rate or reopened tickets. Opinions are useful, but they don't replace the numbers.
AI skill
Have your AI plan an AI pilot your team can judge with numbers
AI pilot plan for a team has your AI write this plan for your team: the tasks, the baseline to measure first, the success and stop criteria, the weeks and the decision.
14 days free · No card needed
Sources
- Microsoft and LinkedIn: 2024 Work Trend Index — 59% of leaders worry about quantifying AI's productivity gains; 60% say leadership lacks a plan (read 2026-11-05)
- Brynjolfsson, Li and Raymond: Generative AI at Work (NBER Working Paper 31161) — measured gains on one repeated task (read 2026-11-05)
- ChatGPT pricing — Business: no training on business data by default (read 2026-11-05)
- Anthropic Privacy Center: commercial products — Claude for Work not used for training by default (read 2026-11-05)