Checklist
Rules for validating AI-drafted requirements
Read-back validation before baseline sign-off
A checklist for business analysts who need to check requirements drafted with AI help against the people who will use them. It covers how to package requirements, how many verdicts each needs and when to take a disputed requirement to a workshop. The thresholds are rules of thumb from practice, not a standard. Koji can run the interviews that produce the verdicts.
Where this appears
What's in this checklist
- Freeze the draft and tag each requirement with its source
- Split it into sets of six, grouped by who is affected
- Give every requirement four answers, including "can't judge"
- Plant one altered requirement in every set
- Get three verdicts per requirement, from at least two roles
- Call it validated at 80% "right as written"
- Take any requirement two people call wrong to a workshop
Use this image
Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:
html
<a href="https://www.koji.so/visuals/requirements-validation-sign-off-rules-checklist"><img src="https://www.koji.so/visuals/requirements-validation-sign-off-rules-checklist/requirements-validation-sign-off-rules-checklist.png" alt="Checklist of seven rules for validating AI-drafted requirements with stakeholders, from sets of six to a workshop for disputes." width="1080" height="1080" /></a>
<p>Source: <a href="https://www.koji.so/visuals/requirements-validation-sign-off-rules-checklist">Rules for validating AI-drafted requirements</a>, Koji</p>
