Skip to main content

6. People, Capability & Governance

What this flow tells you​

How fast the organisation can learn, adapt, and stay safe. Often under-instrumented because the events are slow and the metrics squishy — but the long-term agency of the business depends on this flow more than any other. AI agents are now part of the "people" inventory: they have roles, capabilities, access scopes, and a manager.

Data sources​

  • People flows — hiring pipeline, onboarding, role changes, skill development, performance, retention, exits
  • Agent inventory — which AI agents are live, their capabilities, their scopes, their cost-per-action, their consumed-rate
  • Policy and compliance — access logs, approvals, exceptions, audit findings, incidents, training completion
  • Knowledge — documentation usage, decisions made, playbooks followed or ignored, lessons captured
  • Governance — board decisions, voting records, policy changes, delegated authorities

Decision gates​

  • Hire / let-go — capacity, fit, culture, budget. Owner: hiring manager + HR + finance.
  • Agent activation / deactivation — capability proven, policy compliant, cost justified. Owner: agent-ops lead.
  • Access provisioning — least privilege, time-bound, auditable. Owner: identity system + manager approval.
  • Policy exception — documented, time-bound, escalated to right level.
  • Governance escalation — board-reserved matters, exceptions to delegation.

AI capabilities required​

  • Memory — agent capability registry; per-person performance and growth; institutional knowledge with provenance.
  • Escalation — approval routing: right approver, right SLA, audit trail.
  • Perception — policy drift, sentiment signals, attrition risk indicators, access anomalies.
  • Action — onboard, offboard, provision access, route approval, surface lesson.

Crypto-rail instruments​

  • Stablecoin payroll — global contractors paid in USDC, minutes not days, no FX gouging.
  • KYC + identity credentials — verifiable digital credentials for employees, contractors, agents; one identity primitive across systems.
  • On-chain governance votes — board votes + delegated-authority votes recorded on chain; immutable, traceable, queryable.
  • Role tokens — non-transferable tokens that gate access by role; revocation is on-chain and instant.
  • Lesson receipts — institutional knowledge captured as receipts (per agent receipt schema) so a future agent reads the history without re-asking.

Gauge​

Voluntary attrition rate, annualised, trailing 90 days. Single best read on whether the organisation is somewhere people want to be. For an agent-heavy organisation, pair with AI-agent consumed-rate (how often agent output is actually used downstream — if it trends down, agents are being deprecated by neglect, not decision).

Kill signal​

Voluntary attrition > 15% annualised OR audit exception unresolved > 30 days OR a board-reserved matter executed without board approval. Each signals a governance breach that compounds if uncorrected.

Skill coverage​

Shipped: kyc-screener — identity, sanctions, AML screening. 5p-players — role inventory, human/AI split.

Green-field: stablecoin payroll runner with KYC gate; identity-credential issuer/verifier composite; on-chain governance recorder + queryer; agent-capability registry as a live instrument (not a markdown page); attrition-signal aggregator with intervention recommendations.

Full matrix: skills-matrix.

Upstream / downstream​

  • Upstream: navigation system (Value anchor sets culture; Belief drives recruiting; Control enforces policy).
  • Downstream: every other flow (people + agents operate them); Financial Performance (payroll is the largest single recurring expense for most businesses); Meta-Flow (people + governance telemetry).

5P slice​

5P inspection: Principles (every person and agent earns their position by what they cause to happen), Performance (attrition + consumed-rate), Platform (HRIS + identity + governance system + agent CLI), Process (recruit → onboard → enable → review → exit / promote / let go), Players (HR + managers + board + agent-ops + AI agent on memory and escalation).