Governance & Risk
Governance that runs through the whole evaluation.
AI systems change fast and create cascading effects. FRAME asks teams to govern AI use continuously — not just at set-up. Six mechanisms run across every phase.
Why continuous
Governance is not a one-time set-up
Set once, forgotten
Approve a tool at the start, then stop watching. Bias creeps in, performance drifts, and no one notices until the findings are already out.
Set once, forgotten
Approve a tool at the start, then stop watching. Bias creeps in, performance drifts, and no one notices until the findings are already out.
Cross-cutting governance
Six mechanisms, running throughout
Each one operates across every phase of the evaluation, keeping AI use accountable and responsive as conditions change.
01
Ongoing monitoring
Track AI performance against agreed criteria throughout implementation, and document any degradation or unexpected behaviour.
02
Bias surveillance
Monitor AI outputs for systematic bias, especially differential performance across demographic groups or contexts.
03
Stakeholder feedback
Create channels for participants, commissioners, and users to report concerns about AI-assisted processes.
04
Organisational learning
Run post-evaluation reviews with AI-specific questions, and share lessons across teams and the wider community.
05
Risk-escalation protocols
Define triggers, response procedures, and authority levels so the team can act quickly when problems emerge.
06
A living document
Treat FRAME itself as a living document, updating protocols as AI capabilities evolve and experience accumulates.
A closer look · Mechanism 05
Risk-escalation, aligned with NIST
When something goes wrong, everyone should already know what counts as a problem,
what to do, and who decides. FRAME’s escalation protocols set that out in advance.
Triggers
Know what counts
Validation failures, complaints, unexpected behaviour, evidence of bias, or legal and ethical concerns.
Response
Know what to do
A defined response procedure for each trigger, so the team acts consistently rather than improvising under pressure.
Authority
Know who decides
Clear authority levels so decisions can be made quickly, without bottlenecks or diffused responsibility.
Never finished
FRAME is a living document
AI capabilities move quickly, and so should the protocols that govern them. FRAME is meant to be revisited and revised as tools change and as teams learn what responsible use looks like in their own context.
Evaluators keep the deciding role — AI supports, people judge.
Ready to apply FRAME to your next evaluation?
Download the FRAME checklist, or enrol in the AI for MEL course to put the framework into practice.