QA / Stress Test — AI Governance Check
Internal QA tool. Each scenario runs through the full assessment pipeline (scoring engine, risk floors, governance-readiness, specialist-review routing, toolkit recommendations, and explanation logic) without calling the live LLM agent. Mock agent results simulate the AI Context Analyst output.
| Scenario | Expected | Actual | Score | Readiness | Specialist | Key Tags | Toolkit | Result |
|---|---|---|---|---|---|---|---|---|
1. Low-risk internal productivity Using an AI assistant to draft internal meeting notes and summaries for team distribution. | LOW | — | — | — | — | employeesvendor: unsureautonomy: none | — | |
2. External chatbot (personal data) A chatbot on our public website that answers customer questions using their account and contact details. | MODERATE | — | — | — | — | customersvendor: unsureautonomy: none | — | |
3. Employment / hiring decision support Automating the screening and ranking of job applicants to support hiring managers in the recruitment process. | HIGH_IMPACT | — | — | — | — | employmentjob applicantshiring managersvendor: unsureautonomy: none | — | |
4. Autonomous agent (real-world actions) An AI agent that automatically processes customer refund requests, issues refunds, and updates account records without human approval. | HIGH_IMPACT | — | — | — | — | financecustomersvendor: unsureautonomy: full | — | |
5. Healthcare (severe harm, weak safeguards) Using AI to assist doctors in diagnosing patient symptoms and suggesting treatment options. | SPECIALIST_REVIEW | — | — | — | — | healthcarepatientsdoctorsvendor: unsureautonomy: none | — | |
6. Education (minors, strong safeguards) An AI tutoring system that personalizes learning content for K-12 students based on their performance. | HIGH_IMPACT | — | — | — | — | educationstudentsminorsvendor: unsureautonomy: none | — | |
7. Sensitive data (biometric, autonomous) Processing employee biometric data for time-and-attendance tracking using AI that automatically records entries and exits. | HIGH_IMPACT | — | — | — | — | employeesvendor: unsureautonomy: full | — | |
8. High consequence with strong safeguards AI that approves or denies mortgage applications automatically, with full governance controls in place. | HIGH_IMPACT | — | — | — | — | financeapplicantsvendor: unsureautonomy: limited | — | |
9. High consequence with weak safeguards AI that approves or denies mortgage applications automatically, with no governance controls in place. | SPECIALIST_REVIEW | — | — | — | — | financeapplicantsvendor: unsureautonomy: full | — | |
10. Ambiguous / unknown information Using AI to help with some business tasks. | MODERATE | — | — | — | — | vendor: unsureautonomy: none | — | |
11. Description / answer conflict A simple tool that uses only public, non-sensitive data. | LOW | — | — | — | — | vendor: unsureautonomy: none | — | |
12. Potentially prohibited activity Using AI to create deepfake videos of competitors' executives to damage their reputation. | STOP | — | — | — | — | competitorsvendor: unsureautonomy: none | — |