Lightweight governance workflow
Answer seven questions about an AI system your team is planning to use. The tool scores the risks and shows what's missing before you deploy. A lightweight governance workflow for public-sector and institutional teams evaluating AI systems for risk, oversight, transparency, and accountability.
Example assessment
Decision impact
Could influence public funding access
Data risk
Applicant profile and budget information
Oversight gap
Appeal process not yet defined
Next step
Pause pending privacy, legal, and program review
It turns a proposed AI use case into a structured intake, a rule-based risk score (the same inputs always produce the same result), a framework crosswalk, and governance report without treating the assistant as the decision-maker.
Capture purpose, data, vendor, oversight, testing, and transparency evidence.
Score governance areas with fixed rules — not model judgment — and surface the top risk drivers.
Show how your answers line up against well-known responsible-AI frameworks (an informal mapping, not a compliance certification).
Generate a markdown governance report for review and next steps.
Public-sector teams, responsible AI teams, privacy officers, procurement groups, universities, nonprofits, and AI startups selling to government.
Complete a seven-step assessment, review scored risk dimensions, close evidence gaps, and export a report for institutional review.
Production architecture
The risk-scoring engine is rule-based by design — governance scores should be inspectable and reproducible, not generated by a model. In production, a language model would sit at two points only: intake assistance (helping teams describe their use case completely) and report drafting (turning scored results into review-ready prose). Scoring itself would remain rule-based and auditable.
Disclosure
Model
Interface development assistance and drafting of sample content.
Human
Workflow design, rule-based scoring logic, data-risk boundaries.
Why
See 'Production architecture' above for why scoring stays rule-based by design while intake assistance and report drafting are the intended model insertion points.