·
Copy the blank fields into your existing document or work system, then complete them with the people responsible for the workflow. This page does not save your answers.
Use this card to compare a current process with one proposed AI-assisted process. Complete it with the people who perform, review, and receive the work.
This worksheet provides general operational education. It is not legal advice or a compliance determination. No attorney-client relationship exists, and none of its protections apply.
1. Define the completed unit
Workflow: ________________________________________________
Trigger: __________________________________________________
Comparable unit of completed work: _________________________
Accepted output and next recipient or system: _______________
Process owner: _____________________________________________
Qualified reviewer: ________________________________________
Measurement window: ________________________________________
2. Set the quality gate
Write the conditions that every acceptable unit must meet. Mark any condition whose failure stops the pilot.
| Acceptance condition | How it is checked | Critical? | Baseline pass / total | Pilot pass / total |
|---|---|---|---|---|
| Yes / No | / | / | ||
| Yes / No | / | / | ||
| Yes / No | / | / |
Critical misses and affected units: _________________________
Uncertain outputs routed to review: __________________________
3. Compare end-to-end work
Use the same unit, output requirement, and comparable input mix for both columns. Record stage times across all started units, including rejected attempts. Count each minute once; do not count the same review again as correction.
| Measure | Current process | Proposed process | Evidence source |
|---|---|---|---|
| Units started | |||
| Units accepted | |||
| Total intake time (minutes) | |||
| Total generation or drafting time (minutes) | |||
| Total review time (minutes) | |||
| Total correction time (minutes) | |||
| Total handoff time (minutes) | |||
| Total active effort (minutes, all started units, including rejected attempts) | |||
| Accepted units per total active hour | |||
| Median active time for individually accepted units (minutes) | |||
| Elapsed delivery time, including queues and waiting (minutes) | |||
| Units reopened downstream | |||
| Exceptions outside scope |
Keep individual observations so the team can inspect the spread. Calculate each unit's total before finding its median; do not add stage medians.
Include rejected attempts when comparing total effort and cost. If no units are accepted, report that result without dividing by zero or claiming savings.
For each process, calculate:
- Accepted units per total active hour: units accepted ÷ (total active minutes ÷ 60).
- Active minutes per accepted unit: total active minutes ÷ units accepted.
- Capacity difference per accepted unit: current-process minutes per accepted unit minus proposed-process minutes per accepted unit. A negative result means the proposed process used more effort.
These are batch measures. They include rejected attempts and are different from the median of accepted units alone. If recorded active time is zero, investigate the measurement before calculating a rate.
4. Record cost and realized value
Loaded labor rate used, if any: ______________________________
Tool cost allocated to this workflow and period: _____________
Implementation, training, support, and oversight cost: _______
Capacity difference per accepted unit (minutes): _______________________
How saved capacity will be used: _____________________________
Real expense avoided or additional accepted units completed: __
To express capacity in money, multiply the difference in minutes by the loaded hourly labor rate and divide by 60. Allocate other incremental costs to the same accepted volume. Do not charge again for effort already included in the timed work.
Do not call capacity cash savings unless the organization avoided a real expense. Record the assumption and the evidence needed to confirm it.
5. Explain adoption and exceptions
Eligible users: __________ Active users: __________
Eligible units: __________ Units using the pilot: _________
Reasons for nonuse or workarounds: ___________________________
Most common exception: ______________________________________
Where work increased instead of decreased: ___________________
6. Choose the route
Select one route and record why.
- [ ] Continue: quality passed and the measured result justifies ongoing use.
- [ ] Redesign: a bounded change could fix a source, review, exception, adoption, or cost problem.
- [ ] Stop: a critical condition failed, review burden erased the value, or no credible route remains.
Decision and evidence: ______________________________________
Owner: __________________________ Decision date: __________
Next review date or close date: ______________________________
Evidence that would reverse this decision: ____________________
7. Check the decision
Before closing the card, confirm:
- [ ] The current and proposed processes handled comparable work.
- [ ] Review, correction, rework, and handoff time are included.
- [ ] Critical misses were examined before averages.
- [ ] Synthetic, estimated, and observed numbers are labeled.
- [ ] Capacity and cash claims are separate.
- [ ] A named owner can continue monitoring or close the pilot.
The Testing & Success Metrics framework in Ortaire's AI Governance Toolkit is the companion record for this card, in editable Word and Excel.
