AI Lifecycle Governance
In one sentence
Oversight of an AI system at every stage, from first idea through to retirement.
What it means
An AI system is not a one-time purchase. Its data, users, and performance change over time. Lifecycle governance means setting clear review points across its whole life, rather than approving it once and moving on.
The stages of the lifecycle
Propose
Is AI the right answer to this problem?
Assess
What are the risks, and who could be affected?
Approve
Under what conditions can it proceed?
Monitor
Is it still working fairly and safely?
Escalate
What happens when something goes wrong?
Retire
When and how is it shut down?
Why it matters
Many AI harms appear after launch, as conditions change. Retirement is often forgotten entirely, leaving systems running long after anyone is actively responsible for them.
A simple example
Illustrative scenarioA chatbot is approved for general information. A year later, staff are using it for advice it was never tested on. Monitoring and escalation criteria catch that shift and trigger a new review.
A common misunderstanding
What people often thinkOnce a system is approved, the governance work is finished.
What is closer to the truthApproval is one step of six. Most of the system's life comes after it.