# Agent Workflow Evidence Kit

Measure accepted outcomes, expose missing data and make a defensible pilot decision.

## What decision will this experiment inform?

Name a specific pilot or product decision, not a vague goal to prove AI works.

[Your notes]

## What counts as one completed task?

Include the user-visible artifact and required human review.

[Your notes]

## What are the baseline and candidate?

Record sources, model, prompt, tool version and other differences.

[Your notes]

## What is your primary acceptance rule?

Use an inspectable pass condition. Critical errors should not be offset by formatting points.

[Your notes]

## Which cases will you include?

Include ordinary work and cases likely to reveal weaknesses.

[Your notes]

## Where does the clock start and stop?

Use the same boundaries and include review, corrections and interventions.

[Your notes]

## How will you handle unresolved or missing data?

Keep unknown values missing. State which denominator each metric uses.

[Your notes]

## What will happen after the run?

State the evidence needed, owner, limitations and next review date.

[Your notes]

Companion article: https://outcomedriven.xyz/blog/mcp-agent-workflow-evidence/
