AI pilot charter and decision record

Print this sheet using your browser’s Print command. Complete the charter before collecting results, then attach the measurement log and decision. Blank fields are for your own evidence; no data is sent or saved by this page.

1. Define the test

Decision and owner
Workflow start and finish
Allowed inputs and exclusions
Operator / reviewer
Product / plan / model
Prompt and configuration ID
Sample, calibration exclusions and repeated cases
Start and review dates

2. Fix the rules before measurement

Metric or conditionDefinition, threshold and evidence location
End-to-end time
Accept / correct / discard
Critical failure / stop rule
Who stops and restarts the pilot
Approved data and access

3. Record the result

Assisted total = preparation/generation + review + correction or manual replacement. Saving = manual baseline − assisted total. Keep rejected attempts and negative savings. Use the same timing boundaries and final output standard for both workflows.

Matched case count
Manual / assisted total minutes
Accept / correct / discard counts
Critical failures and responses
Software, setup and administration costs
Observed adoption and eligible volume
Evidence references

4. Decide and state the limits

Approve, approve with enforceable conditions, extend to resolve a specific gap, or reject. Explain which rules passed or failed. If changing configuration, retain the old results and test fresh cases.

What was not tested, and what remains an assumption?

Decision owner, date, conditions, monitoring owner and reassessment trigger:

ScoutChoice · Methods worksheet · September 13, 2026. Completion is not certification or vendor performance evidence.