# Automation candidate worksheet

Planning aid. This is not a permission grant, an ROI guarantee or a recommendation to automate a role. Compare bounded tasks, not job titles.

## Candidate task

Name:
Purpose:
Duty supported:
Accountable owner:
Current performer:
Starts when:
Inputs and source:
Current actions:
Output and receiving team:
Software used:
Frequency and volume (source/date):
Current effort or delay (measured / reported / estimated):

## Evidence and boundaries

Which inputs are reliable enough to use?
Which information is sensitive or restricted?
Which errors would matter most?
Which cases should be excluded?
What judgment or approval remains human?
Who is authorized to approve a pilot?
Is needed access available and legitimate?
What is unknown?

## Compare the next action

Choose one: clarify process / improve inputs / document responsibility / integrate existing software / test bounded AI assistance / leave human-led / collect more evidence.

Reason:
Supporting evidence:
Alternative explanation:

## Define the smallest responsible test

Test data approved by:
Baseline and source:
Expected output:
Acceptance checks:
Human reviewer:
Failure/stop conditions:
Fallback:
Evaluation date:
Ongoing owner:

## Keep economic claims honest

Potential capacity released = frequency x time potentially saved per task x realistic adoption. List all assumptions. This is not cash saved, and implementation, review, software and maintenance costs must be considered separately. Use actual pilot measurements before publishing a result.

More guidance: https://dutygraph.com/learn/identify-automation-opportunities/