Home

Testing an AI knowledge assistant

Testing an AI knowledge assistant

Use a fixed set of representative questions and record the existing process as a baseline. Evaluate whether the answer is useful, whether its reference supports it and how much work a person needs to correct or verify it.

Include missing information, old versions and restricted documents. Test source updates and removal. Review repeated attempts and several users where relevant. Agree what would justify proceeding and which failures must be resolved first.

A useful first step

What does your team
keep searching for?

Bring one recurring question. We will work through what a useful test would involve.

Find your first use case