Hiring
Teams are experimenting and building. Agents have local secrets, vague owners, and no standard personnel record.
symptom: “who owns this agent?”
Maturity model
Where is your enterprise on the AI workforce curve?
The gap is not model quality. It is operating readiness: credentials, managers, records, evaluations, and exits for the digital employees already entering the business.
Stuck
The agents work, but review blocks them. Evidence is assembled manually and approvals happen out of band.
symptom: “security has had it for months”
Operating
Agents have identity, supervision, evaluation, records, and termination. Work can move with evidence.
symptom: “show me the record”
Scaling
Digital and human workforce metrics roll into one operating cadence, with cohort-level performance and risk.
symptom: “which cohort do we hire next?”
Stage 02 realityThe agent works. The operating model does not.
Useful agents get stuck when the enterprise cannot prove who supervises them or what record they leave behind.
Ownermissing
No single accountable human for the worker.
Credentialsshared
Access is embedded, broad, or copied from a human.
Evidencemanual
Review depends on screenshots and one-off docs.
Exitunclear
No atomic revoke and attestation path.
The move
Stage 02 is not a technology problem. It is a governance problem.
ARX moves teams into Stage 03 by making the five primitives real for one cohort first: manager binding, scoped credentials, approval gates, personnel record, and termination attestation.
Map the roster
Declare agents, owners, managers, connectors, write actions, and records.
Run in shadow
Dry-run approvals and audit events against production-like activity.
Operate by cohort
Turn on governed work for the lowest-risk cohort, then expand by function.
Find your stage.
Eight questions. No login. A stage, the missing primitives, and the next operating move.