AI Governance Maturity Assessment
Twelve questions, about three minutes. Find out which of the five governance stages your AI agent program is at — and the highest-impact controls to add next.
Based on the AI Governance Maturity Model. Want the controls behind each stage? Get started with Praesidia, review your result with us or see the proof.
What the assessment measures
The twelve questions cover six control areas, two questions each:
- Inventory and visibility: whether you know which agents run, who owns them, and how you find out when one misbehaves.
- Agent identity and access: how agents authenticate and how narrowly their permissions are scoped.
- Guardrails and content control: whether inputs and outputs are checked for PII, secrets and prompt injection, and how high-risk actions are handled.
- Connection policy and cost control: whether a looping agent can run up unbounded cost, and whether agent-to-resource connections have rate limits and allowed operations.
- Audit and compliance: whether you can prove a control was applied consistently, and how data-subject and erasure requests are handled.
- Trust and automation: whether you have a trust model for agents and how the program improves over time.
How your stage is scored
Each answer scores from 1 (the first option) to 5 (the last). The two answers in an area are averaged, and your overall score is the average of the six areas. An overall score below 1.5 is Stage 1, Ad-Hoc; below 2.5 is Stage 2, Repeatable; below 3.5 is Stage 3, Defined; below 4.5 is Stage 4, Managed; anything higher is Stage 5, Optimized. The result shows a bar per area and suggests up to three of your lowest-scoring areas that sit below 3.5, each with a short next step and a guide to read.
Limits of a self-assessment
The score reflects the options you pick, not evidence. Two teams that choose the same answers can be in very different places, and an average can hide one area that is far behind. Treat the stage as a conversation starter for your security and compliance reviewers, then check each answer against what your systems actually record. The scoring runs in your browser, nothing is stored or sent, and reloading the page clears it.
The stages come from the AI governance maturity model, which describes each stage in more depth. For the controls behind the questions, read the AI agent governance guide; for the regulatory side, see how controls map to compliance frameworks.