One page, four things
You now have the pieces: candidate tasks, an honest value estimate, ground rules, a review routine and a way to bring people with you. This lesson turns them into something you can put in front of your director without apologising for it.
The artifact is one page. Not a deck, not a strategy, not a maturity model. One page with four things on it: two pilot use cases, a named owner for each, a success measure, and a date when you decide what happens next.
Short is not lazy here, it is the whole point. A one-page plan gets read, gets challenged properly, and gets remembered by the person who has to approve it. A twenty-slide deck gets a nod and no decision.
LearnAI has no accounts and stores nothing you type. So build this plan outside the site, in whatever your team already uses: a document, a wiki page, a note. Copy the template below and fill it in there.
In plain English
- Owner:
- One named person responsible for a pilot. Not a team, not a department. One name.
- Success measure:
- The number or observable thing that tells you whether it worked. Agreed before you start.
- Review date:
- A fixed date to decide continue, change or stop. In the diary, not in principle.
- Scope creep:
- A small trial quietly growing into a programme nobody agreed to fund.
The template
Copy this, fill it in, and keep it to one page. If it runs to two, something on it is not a pilot.
TEAM AI PILOT PLAN
Team: [name] Manager: [you] Date: [today]
WHY THIS, WHY NOW
[Two sentences. The problem, not the technology.]
PILOT 1
Task: [what it is, who does it today, how often]
Owner: [one name]
What changes: [what they will do differently]
Success measure: [e.g. end-to-end time falls from 40 to under 25 minutes,
with no accuracy issues found in spot-checks]
Baseline today: [the number you measured before starting]
PILOT 2
Task: [what it is, who does it today, how often]
Owner: [one name]
What changes: [what they will do differently]
Success measure: [the number or observable outcome]
Baseline today: [the number you measured before starting]
GROUND RULES
[One line pointing at your one-page rules, plus the accountability sentence:
the person who sends it owns it.]
WHAT WE ARE NOT DOING
[Two or three things explicitly out of scope, so nobody assumes them.]
COST
Tools: [x per person per month for y people]
Time: [rough hours of setup and review across the pilot]
REVIEW DATE
[A specific date, 6 to 8 weeks out. On that date we decide: continue,
change, or stop. Stop is a real option.]
Checkpoint
Two pilots, one named owner each, a success measure agreed before you start, and a review date in the diary. Everything else is decoration.
The four things people get wrong
Owners that are not people. Operations will own this means nobody owns it. One name, and that person should know their name is on it before you write it down.
Measures invented afterwards. If you decide what success looks like after seeing the results, you will find success, because everyone does. Agree the measure first and write down today's baseline next to it.
A review date that never arrives. Without a date, pilots do not fail, they just fade, and nobody learns anything. Put it in the calendar the day you write the plan.
No stop condition. Say plainly that stopping is one of the three outcomes. It makes the whole plan more credible, and it protects you from the pilot that limps along for a year because cancelling would look like failure.
Beware of measuring adoption instead of outcomes. Percentage of team using the tool tells you nothing about whether the work got better. Measure the task, not the tool.
โ Weak prompt
Prompt
Write an AI adoption strategy for my team.
Output
Phase 1: Discovery and readiness assessment. Phase 2: Capability building. Phase 3: Scaled deployment. Phase 4: Continuous optimisation.
Four phases, no tasks, no owners, no numbers and no date. Nobody could start work on Monday from this.
โ Good prompt
Prompt
Turn my notes into a one-page pilot plan. Two pilots only. For each: the task, one named owner, what changes day to day, a success measure with a number, and the baseline. Add a review date six weeks out and a short list of what we are not doing. Plain text, under 400 words, no phases, no strategy language.
Output
Pilot 1. Task: weekly client summary, 12 per week, written by Sam. Owner: Sam. Success: end-to-end time under 25 minutes from 40, no factual errors found in spot-checks. Baseline: 40 minutes measured over five samples.
Concrete enough to start, specific enough to challenge, and short enough to be read by the person approving it.
Turn my notes into a one-page team AI pilot plan.
My notes:
Candidate tasks: [from your audit]
Estimated value: [from your calculation, including review time]
People who could own each: [names]
Constraints: [budget, policy, anything in the way]
Rules for the output:
- Exactly two pilots, no more
- One named owner per pilot
- A success measure containing a number, plus the baseline it improves on
- A review date [6] weeks from [today's date]
- A short list of what we are explicitly not doing
- Under 400 words, plain text, no strategy or phase language
If my notes are missing anything the plan needs, list the gaps at the end
rather than filling them in with assumptions.
Pressure-test it before anyone else does
Your director will ask three questions: what does it cost, what happens if it goes wrong, and how will we know. Have answers ready and the meeting takes ten minutes.
Here is my one-page AI pilot plan:
[paste it]
Act as my director, who is supportive but busy and dislikes vague plans.
Ask me the six hardest questions about this plan, hardest first.
For each, say what a weak answer would sound like and what a strong one needs.
Then flag anything in the plan that is a decoration rather than a decision.
The prompt pack to hand your owners
Your two pilot owners should not start from a blank box. Give them these three, adapted to the task, on the day you agree the plan.
-
THE DRAFTING PROMPT
You are helping with [task]. Here is the source material: [paste].
Produce [what good output looks like], in [format], under [length].
Use only the information in the source. If something needed is missing,
list it as a gap rather than inventing it.
-
THE CHECKING PROMPT
Compare this draft against the source below. List every number, date, name,
claim and commitment, and mark each as matching, contradicting, or absent
from the source. Ignore style.
DRAFT: [paste] SOURCE: [paste]
-
THE WEEKLY LOG
Task done: [x] times this week
Time before (baseline): [x] minutes
Time now, start to finish including checking: [x] minutes
Problems found in review: [what and how serious]
One thing that would make this better: [note]
That third one matters more than it looks. Five weeks of a two-line log gives you a real answer on the review date, instead of a conversation about how everybody feels it went.
On the review date
Sit down with the logs and decide one of three things. Continue as is. Change something specific: a different task, a different prompt, a different owner. Or stop, write down what you learned, and pick a different candidate from your audit.
All three are good outcomes. The only bad outcome is arriving at the date with no measurements and extending it by another six weeks.
Whatever you decide, tell your team the actual numbers, including the disappointing ones. Teams that see honest results from the first pilot volunteer for the second. Teams that only ever hear success stories stop believing any of them.
Finish the quiz below and the Team Navigator badge is yours. More usefully, so is a plan you can send to your director this afternoon, which is a considerably rarer thing than a badge.
๐ Quiz
Question 1 of 4What are the four essential elements of the one-page plan?