Staff are already using AI. The control question is whether you can show how.
govFMP is a governance control that sits in front of the models your staff already use. It scores each prompt against a rubric your agency authors, screens it against your acceptable-use and data-sensitivity rules, and writes an audit record of every decision with the reason attached. It makes no determination about any member of the public, and it should not be deployed anywhere that would create one.
Where this actually stands
There are no deployments yet and no public-sector customers. Scoring has been exercised against prompt sets built in-house, including deliberately non-compliant prompts, but never against real agency traffic. There has been no independent bias, accuracy, or disparate-impact study. Our AI FactSheet states all of this and marks six gaps on purpose. If you are evaluating vendors, read the gaps first.
What the control does
Quality scoring
Every prompt scored against a rubric, returned as a per-axis breakdown rather than one opaque number. The point is consistency of AI-assisted work across staff of very different skill levels.
Acceptable-use screening
Each prompt evaluated against the agency's own rules, including a data-sensitivity dimension mapping data classes to permitted models and roles. The decision comes back with the specific rule that produced it.
Audit record
Every decision logged with its reason, the authenticated user, and the rubric version in force at the time, so a past decision can be reconstructed against the rules that actually applied to it.
Observe mode first
Scores and clearances are recorded without refusing anything, so an agency can measure the effect on real staff traffic before any rule becomes consequential.
Take the standard, whether or not you ever talk to us
The rubric shipped as the default is PQA-1, the Prompt QA Governance Standard. It is original work, published openly, and free to adopt or fork, including by agencies that never become customers. Rubric packs are meant to be agency-authored: a rubric encodes a judgment about what a good prompt looks like, and ours was written by a small company in the United States in English.
- PQA-1 standard — the rubric itself, adoptable as a one-page internal standard.
- AI FactSheet — GovAI Coalition template, including the declared gaps.
- Two-page brief — for circulating to people who will not read the FactSheet.
- How the scoring axes work — the consumer-side write-up of the same rubric.
We would rather have the conversation than the pilot
What is genuinely useful to us right now is talking to people who own this problem inside an agency, including if your conclusion is that you do not need this. If a pilot comes out of it, it would run in observe mode.
Goes to a person, not a sales sequence. No call booking, no demo gate.
The consumer product this grew out of is live and free to try: paste a prompt and see the scoring run. It is the same rubric, without the screening gate, the identity integration, or the audit log.