Future of My Business — 14-day AI pilot checklist
Free editable template | Version 3 October 2026
Guide: https://futureofmybusiness.com/resources/14-day-ai-pilot

ONE EXPERIMENT
Workflow:
Question we are testing:
Owner:
Reviewer:
Start date:
Review date:
Time and spending limit:
Approved tool and account:
Approved source information:
Excluded cases and data:
Quality standard:
Stop conditions:
Manual fallback:

DAYS 1–3: RECORD THE CURRENT PROCESS
[ ] Define the trigger, input, output and common exceptions.
[ ] Record at least five current-process examples.
[ ] Include active checking and correction time.
[ ] Note case difficulty and operator experience.
[ ] Keep quality criteria the same throughout the comparison.

DAYS 4–7: PREPARE AND TEST IN DRAFT
[ ] Check authorization for the tool and the information it will receive.
[ ] Start with invented or appropriately de-identified examples.
[ ] Record setup effort once, including preparation and training.
[ ] Test an ordinary case, missing information and an excluded case.
[ ] Confirm the reviewer can verify the output.
[ ] Keep live sending and consequential actions under human control.

DAYS 8–12: RUN A BOUNDED, APPROVED TRIAL
[ ] Record at least five comparable trial examples.
[ ] Count work, review and corrections without double-counting.
[ ] Include failed attempts and manual completion.
[ ] Record quality and follow-up issues separately.
[ ] Mark any change in tool, prompt, data or process.
[ ] Pause when a stop condition is reached.

DAYS 13–14: COMPARE AND DECIDE
Average baseline effort per case:
Number of trial cases:
Baseline equivalent (average baseline x trial count):
Trial work + review + corrections:
One-time setup effort:
Total trial effort (handling + setup):
Net time difference (baseline equivalent - total trial effort):
Subscription/usage costs (money, separate from minutes):
Quality comparison:
Important limitations:

Decision: KEEP / ADJUST / STOP
Reason:
If adjusting, one change to test:
Where any released time will go:
Next review date:
Responsible person:

KEEP requires lower comparable routine handling effort and at least the same
quality. State separately whether setup has been recovered. ADJUST or STOP
remain valid outcomes. Extend the trial if there are too few comparable cases.
Five examples are a practical starting sample, not statistical proof.
Time released is not automatically cash saved or revenue gained.
This small before-and-after comparison does not prove causality.

Effort log: https://futureofmybusiness.com/resources/ai-effort-log.csv
Measurement guide: https://futureofmybusiness.com/resources/measure-ai-time-savings

CSV instructions: record minutes in separate work, review and correction
columns. Baseline total includes all three. In the single SETUP row, put all
one-time setup effort in work_minutes and leave review/correction blank.
Do not count that setup row as a baseline or trial case. Keep missing values
blank; zero means a measured zero. Use yes/no for quality_pass.
No formulas or automatic calculations are included in the CSV.
Do not put personal or confidential details into the notes field.