Agent as Human — complete RACI & control model

23 lifecycle stages Ward 1 · Human proxy
Core principle: An agentic AI that acts on behalf of a person must be governed with the same rigour as a human employee — across identity, authority, supervision, performance, conduct, and exit. Below is the complete lifecycle model with RACI assignments, control requirements, evidence expectations, and risk exposure per stage.
R — Responsible
A — Accountable
C — Consulted
I — Informed
BU = Business owner IT = Platform / engineering / IAM RISK = Op / model / tech risk COMP = Compliance / legal / privacy OPS = Operations / service mgmt IA = Internal audit
#Control areaHuman eq. BUITRISKCOMPOPSIA Evidence
23
Lifecycle stages
290+
Control points
6
RACI roles
5
Lifecycle phases
14
Control domains
The blunt version: If a human employee needs an identity, manager, role, access rights, supervision, performance reviews, conduct controls, and an exit process — so does every agentic AI that acts on behalf of your organisation. Any stage missing from this lifecycle is a governance gap a regulator will find before you do.