Companion assessment

Rate your program, honestly

A companion to The Non-Human Majority. Maturity is uneven, you can be advanced in one dimension and primitive in another, so rate each dimension on its own evidence, then track the metrics that a skeptic cannot wave away.

The maturity matrix
DiscoveryLifecycleAuthenticationAuthorizationRuntimeDelegation
Six dimensions, five levels each. Plot where you actually are on every row, the line will rarely be straight.
The scorecard

Metrics map to the four questions, plus the agent frontier. They are written to resist the easy answer, coverage against a denominator you can defend, privilege against observed use, runtime against enforceable response, and ownership against a tested path to stop the actor.

What are you?

Coverage and resolution
  • Percent of authoritative systems connected
  • Percent of authentication events resolved to a principal
  • Number of unresolved privileged principals
  • Entity-resolution confidence, and data freshness

What authority are you exercising?

Privilege and effective reach
  • Granted versus observed permissions
  • Unused privileged entitlements
  • Identities with cross-environment authority
  • Crown-jewel resources reachable, and effective high-risk paths

What are you doing now?

Runtime and containment
  • Percent of critical actions attributable to a resolved principal
  • Mean time to detect misuse, mean time to contain
  • Percent of high-confidence events safely contained, and the false-containment rate
  • Percent of critical systems with enforceable, not merely visible, controls

Who owns and can stop you?

Accountability
  • Verified technical-owner and business-sponsor coverage
  • Orphaned privileged identities
  • Percent with a tested revocation or containment path
  • Percent of on-behalf-of actions that retain the delegating user

The agent frontier

Agent and delegation governance
  • Agents using unique versus shared principals
  • Approved agent-to-tool relationships, and unsanctioned ones
  • Agents with destructive capabilities, sponsors, and tested kill switches
  • Actions that retain task, user, and purpose context

If you track only one line, track the gap between agents discovered and agents governed, against a discovery count that is itself still climbing. When those diverge, the accountability gap is widening.