Operations

Governance & rollout

AI is not self-running. Without an owner, outcome metrics, documented automation and a phased rollout, momentum fades and broken workflows multiply quietly.

What breaks

No AI owner or review cadence
Measuring usage instead of outcomes
Scaling before the pilot is stable
Undocumented, broken workflows

What we check

Named AI owner with a weekly cadence
Do now
When AI has no owner it has no future — no one improves outputs and it becomes ignored.
First step: Name an AI champion, give them time, and book a 30-minute weekly agent review.
Outcome metrics defined
Do now
Usage counts look like progress while impact is flat; you cannot prove or improve value.
First step: Pick one outcome metric per agent and build a dashboard that shows it weekly.
Workflows are healthy
Do nowAuto-checked by scan
A field drives a list, the list feeds a workflow, the workflow changes the owner. Broken automation upstream corrupts what AI reads.
First step: Review workflows with errors or no enrollments in 90 days; fix or archive them.
Automation is documented
Do laterAuto-checked by scan
Small changes become expensive when nobody can explain them six months later.
First step: Add a one-line description (owner, purpose, test records) to every active workflow.
Phased rollout with guardrails
Do now
Scaling too quickly multiplies issues; automation should follow confidence, not precede it.
First step: Write a rollout plan: one pilot, success criteria, then expand in phases.