01

Build a source register

List services, FAQs, policies, scripts, forms and representative interactions. Record an owner, version and approval status.

02

Resolve contradictions

Highlight conflicting hours, service areas, prices or promises. A human owner must resolve the conflict before the content is used.

03

Write boundaries beside answers

For every approved answer, record when it no longer applies and which person handles the exception.

04

Test retrieval, not just wording

Ask realistic questions and check whether the correct source is found. A polished answer from the wrong policy is still a failure.

Technical evaluator notes

Evaluation should confirm architecture, data handling, knowledge ownership, permissions, human authority, QA/UAT criteria, integration boundaries, monitoring, audit evidence, release gates and rollback for the specific workflow. The guide does not claim that external systems are connected.

Find where AI could helpBack to resources