Decision rights
Who can register an initiative, approve assessment, start delivery, accept risk, and close impact.
AI governance defines roles, decision rights, risk routing, and gates so strong initiatives move faster and weak or risky ones stop before large spend.
Good AI governance does not route every initiative through one heavy committee. It routes initiatives by value, risk, and readiness.
Who can register an initiative, approve assessment, start delivery, accept risk, and close impact.
Clear checks before resource spend: data, security, process owner, product route, impact metric.
Different routes for self-service, pilots, strategic initiatives, sensitive data, and prohibited use cases.
Every continuation, return, stop, or exception is recorded so the portfolio does not become a black box.
There is a problem, initiator, short description, expected impact type, and business context.
No obvious duplicate, owner exists, base fields are complete, and the next analysis step is clear.
Product route is selected, similar initiatives checked, risks are not blocking, and priority is clear.
Actual metrics, data source, and decision on scaling, support, revision, or closure are available.
Teams launch tools and agents without shared architecture, data rules, or access controls.
Platforms, licenses, and pilots are purchased without a clear mechanism for selection and stopping.
Criteria are not known in advance, so every initiative becomes a manual debate between functions.
Initiatives reach demos but lack process owners, adoption metrics, and dates for impact validation.
Start with an audit of portfolio and gates: risks, decision rights, blockers, and first management rules.