Decision criteria

Use this checklist to decide if a documentation task is a good fit for LLM automation.

1) Source of truth

• Do we have a trusted source (spec, code, API contract, approved docs)?
• Can we enforce citation or traceability back to that source?

2) Risk and blast radius

• What happens if the output is wrong?
• Who is harmed (customer, support, compliance, security)?

3) Human review lane

• Who reviews (writer, engineer, PM)?
• What is the definition of “done” for that output?

4) Evaluation readiness

• Can we define a rubric and test cases?
• Can we measure improvements over time?