10 to the 23 AI logo

Practice Library

Governance patternprocedural

Understand the system

Fund continuous learning about what the AI deployment is actually doing across the sociotechnical system, so a limited budget can target hidden dynamics and real failure modes instead of guessing.

What it changes

increasedPeople or agents using it(correction, all three components)
dampenedContaminated records read and believed(via reliable checking sources)

Who can pull it

Deploying organizationOversight board

What it looks like institutionally

Governance under scarcity fails when it spends on assumptions. What a deployment is actually doing — which pathways carry the most unverified work, where operator deference is quietly rising, which checks run against stale or contaminated sources — is rarely visible from the org chart. Funding continuous learning about the live sociotechnical system turns that invisible behaviour into something you can see and budget against.

Treat understanding as a standing budget line, not a one-time audit: instrumentation, incident review, and protected time for the people closest to the work to notice and report what the dashboards miss. The payoff is targeting — the same limited budget aimed at the dynamics and failure modes that actually matter in this deployment, rather than spread evenly across everything.

The diagnostic for any governance claim becomes: what do we actually know about how this system behaves, how did we learn it, and how stale is that knowledge now?

Addresses: Nominal oversight without capacity · Checks against unreliable sources. Test a version of this lever in the PAN Lab.

Deciding whether this lever fits your deployment?

Which patterns matter — and in what order — depends on your system's actual shape. Ranking your options on evidence, with what can backfire stated, is engagement work.