Governed Autonomy Organizational Maturity Evaluation v4

Version 4 | Released 1 November 2026 | Governed Autonomy Institute

This instrument evaluates how consistently an organization establishes, operates and improves agent governance. It is separate from the runtime baseline, the permitted effects of any use case and deployment conformance. A maturity rating shall not license an action, waive a Law or convert incomplete deployment evidence into conformance.

Evaluation method

Declare the organizational boundary, evaluation date, owners, evidence period, representative deployments and exclusions. Rate each dimension independently using observed operating evidence. The names Identified, Governed and Continuous are retained for continuity, but v3.6 narrative ratings and RFC revision 2 level-to-tier scores do not automatically convert to these ratings.

Dimension Identified Governed Continuous
Inventory and accountability Agent/workload and use-case inventory has owners and explicit gaps; authoritative sources and lifecycle responsibilities are identified. Issuance, change, expiry and withdrawal are routinely reconciled against operated deployments; sampled attribution is dependable. Discovery and change signals expose drift; effectiveness is measured and exceptions are resolved through an operating feedback loop.
Missions and profiles Missions, effect boundaries and policy owners are documented for sampled deployments; unassessed cases remain explicit. Deployment owners approve profiles, standing bounds and change rules; permission is based on profiles rather than maturity labels. Profile changes and observed failures systematically improve mission design and risk decisions without granting silent scope growth.
Enforcement and dependencies Enforcement points, trusted functions, exclusions and known bypasses are inventoried. Applicable controls, cross-system contracts and dependency-failure modes are demonstrated and operated under change control. Changes, bypass signals and failures drive targeted retests; recurring jobs retain current checks as capability and scale change.
Collaboration and state Handoffs, authorities, persistent stores and downstream obligations have accountable owners. Grant chains, shared limits, revocation and memory/output lifecycle are exercised across the declared operating boundary. Cross-boundary failures, state contamination and coordination limits are measured and used to improve controls and contracts.
Intervention and continuity Intervention, ownership and service dependencies are identified; gaps are visible. Pause, revoke, handoff and recovery are exercised with deadlines, measurable bounds and explicit authority. Exercises and incidents lead to demonstrated improvements in continuity, residual-effect reporting and restoration discipline.
Evidence and learning Evidence sources, access, retention and known omissions are documented. Sampled missions are reconstructable; claims, limits, failures and retests are recorded with consistent assessment methods. Evidence sufficiency, false denials, useful completion and human burden are tracked; disagreements and incidents improve methods and controls.

Not assessed is used when the evidence cannot support a rating. Below Identified means affirmative evidence shows the stated Identified practices are absent or materially incomplete. A missing document alone may instead mean unavailable evidence; the evaluator states which.

For each dimension, assign the highest stage whose practices and preceding practices are evidenced throughout the declared evaluated scope. Document exceptions and sampling limits. A policy document without operating evidence does not demonstrate the Governed or Continuous stage.

Required evaluation record

Record the dimension, rating, evidence and period, sample and exclusions, assessor and interests, unresolved gaps, owner and improvement commitment. State whether evidence is documentary, examined operation, exercise or independently reviewed.

Use the dimension profile as the primary result. If an overall stage is reported, use the lowest evidenced dimension, disclose the entire profile and do not turn not-assessed dimensions into a stage. Explain the organizational boundary so that a mature team does not imply an unevaluated enterprise is equally mature.

What the rating means

Identified means governance responsibilities and gaps are visible. Governed means the organization demonstrates and operates the controls required for its declared deployments. Continuous means it measures their performance and changes them through an accountable feedback process. These describe operating capability, not increasing tiers of permitted effects.

A newly formed organization can run a narrow deployment that satisfies all applicable baseline and profile requirements. A highly mature organization can operate a nonconforming deployment. Report both facts rather than forcing them into one score.

Review and improvement

The organization declares a review interval appropriate to its change rate and risk, with event-triggered reviews for material boundary, authority or operating changes. Each improvement commitment names the affected dimension, required evidence, owner and review date. An upward rating requires new evidence; a published plan to improve is not the improvement itself.

Nothing here prescribes staffing ratios, professional certification, a commercial product, or an enterprise-wide security replacement. Adopters integrate the evaluation with their existing organizational risk and assurance processes.