# Automation decision record

## Candidate check
- Scenario, business consequence and expected result:
- Repetition frequency and time horizon:
- Current manual procedure and observed effort:
- Stable interfaces, test data and dependencies:
- Reasons to keep any part manual:

## Effort assumptions
- Setup and integration hours:
- Automated review hours per run:
- Maintenance hours per month:
- Manual hours per run:
- Runs per month / months compared:
- Hourly labour rate and additional tool costs:
- Range of uncertainty for each estimate:

Manual effort = manual hours per run × runs per month × months.
Automated effort = setup hours + automated review hours per run × runs per month × months + maintenance hours per month × months.
Compare equal scope. Include any tool charges separately from labour hours.

## Quality and ownership checks
- [ ] Expected behaviour is agreed independently of the implementation.
- [ ] Data can be created and cleaned without affecting other tests.
- [ ] Failures point to useful evidence, not only a red status.
- [ ] A named team maintains selectors, fixtures and requirements.
- [ ] The pipeline gate matches the consequence of failure.

## Decision
- Automate / keep manual / run a limited pilot:
- Assumptions most likely to change the conclusion:
- Scope of the first implementation:
- Evidence the pilot must produce:
- Owner and review date:

Example — fictional: a check takes 30 minutes and runs twice a month. A 20-hour setup plus one hour of monthly maintenance exceeds six months of the current manual effort. Unless there is a separate risk or feedback-speed reason, keeping it manual is a reasonable decision. Different repetition and maintenance estimates can reverse that conclusion; keep those assumptions visible.