# AI Impact Scorecard

Use this to measure whether an AI use case is mature enough to scale.

## Use Case

Task name:

Role or team:

Monthly volume:

AI solution:

Decision owner:

Measurement period:

## Scorecard

| Dimension | Baseline | AI-assisted | Change | Evidence source | Confidence |
| --- | --- | --- | --- | --- | --- |
| Cycle time: median full task time | | | | | |
| Cycle time: p90 full task time | | | | | |
| Throughput per person per week | | | | | |
| First-time-right rate | | | | | |
| Defect or correction rate | | | | | |
| Escalation rate | | | | | |
| Review time per accepted output | | | | | |
| Cost per accepted output | | | | | |
| Retained adoption after 4 weeks | | | | | |
| Employee effort score | | | | | |
| Risk or policy exceptions | | | | | |

Confidence scale:

- High: measured directly from systems or controlled sample.
- Medium: measured from structured sample or manager-reviewed logs.
- Low: self-reported, anecdotal, or incomplete.

## Unit Economics

Monthly cost:

Monthly accepted output volume:

Cost per accepted output:

Baseline cost per output:

Change:

## Evidence Notes

What was measured directly?

What was estimated?

What is still unknown?

What could make the result misleading?

## Decision

Recommended action:

- Scale
- Change
- Retest
- Stop

Reason:

Next measurement date:

What changed?

Was the change valuable?

Should this be scaled, changed, or stopped?