The Governed AI Lifecycle
Eight stages. Seven gates. Every go/no-go is also a governance decision.
PMI-CPMAI's data-first phases form the spine; IAPP AIGP controls sit on every gate; Prosci ADKAR runs from week one.
- Stage 0
Mobilize and govern
Policy, inventory, risk tiers, literacy
◆ G0 · Policy approved - Stage 1
Frame the use case
Problem, AI pattern, value, owner
◆ G1 · Intake and tier - Stage 2
Prove the data
Lawful basis, quality, bias exposure
◆ G2 · Data rights - Stage 3
Prepare the data
Lineage, representativeness, proxies
◆ G3 · Lineage - Stage 4
Build
Models, vendors, guardrails
◆ G4 · Model card - Stage 5
Validate
Independent tests, impact assessment
◆ G5 · Validation - Stage 6
Deploy and operate
Human oversight, monitoring, rollback
◆ G6 · Go-live readiness - Stage 7
Scale and sustain
Value tracking, revalidation, retirement
◆ Ongoing review