# AI pilot decision memo
Use fictional or approved information. Separate observed results from assumptions and forecasts.

## Decision scope
- Workflow, users and allowed actions:
- Proposed next commitment:
- Accountable owner and decision date:
- Acceptance gates defined before testing:

## Comparable evidence
- Task selection, batch size and completion standard:
- Baseline / pilot configuration and dates:
- Possible differences or confounders:

| Effort component | Existing process | Observed pilot | Forecast only | Evidence |
|---|---|---|---|---|
| Drafting / preparation | | | | |
| Review / correction | | | | |
| Allocated upkeep | | | | |
| Total, without double counting | | | | |

## Calculation and assumptions
- Pilot total minus baseline:
- Percentage difference, with stated denominator:
- Break-even review effort under explicit fixed assumptions:
- Setup, fees and other excluded costs:
- Why this is not automatically a causal or ROI estimate:

## Quality and readiness
- Case-level gate results and unresolved cases:
- Observed correction effort:
- Source, monitoring, incident and fallback owners:
- Rollback rehearsal and unfinished-work handling:

## Decision
- Stop / improve / scale, with rationale:
- Approved boundary and excluded expansion:
- Next experiment, effort limit and deadline:
- Observation that would change this decision:
- Stop rule, fallback and retest owner:

Source guide: https://learn.aicompetence.org/knowledge/stop-improve-or-scale-ai/