10 to the 23 AI logo

Practice Library

Governance patternstructural

Improve the model

The default instinct — buy or build a better model — is a real lever with an honest, limited reach.

What it changes

decreasedAI modelIllustrative PAN-run result: ≈6.3% harm removed by a model upgrade alone in the same published PAN run that credited a verifier with ≈45.8%.

Who can pull it

DeveloperVendorDeploying organization

What it looks like institutionally

Reducing the model's base error rate helps every downstream pathway a little. It is also the lever institutions reach for first, because it requires no organizational change: procurement instead of governance.

Its honest limits: base error has an empirical floor (no current system reaches zero in demanding domains), and in propagation terms a better model shrinks the source while leaving every loop — adoption, records, retrieval, peer spread — untouched. Illustrative PAN runs across hundreds of deployment structures found system-side levers outperforming equal-effort model improvements in the overwhelming majority of cases; the ledgered scenario results carry the specifics and their caveats.

Use it, but use it last-alone: pair model improvements with the structural levers that govern what happens to the errors that remain.

Ledgered PAN-run results used above

In a published PAN model run over a supervised-plus-agent scenario, adding a verifier to the autonomous agent removed roughly 45.8% of steady-state harm and a coordinated governance package roughly 43.0%, while upgrading the model alone removed roughly 6.3%.[]

Across a published PAN sweep of 216 deployment topologies (432 sampled deployments), system-layer governance outperformed equal-effort model-layer improvement in 99.1% of cases at roughly 3.4x mean leverage - a figure the project corrected downward from an earlier 9.4x estimate.[]

Addresses: High base error. 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.

Sources & Evidence

Claims made on this page and what supports them. The full registry lives in Evidence.

ScenarioIn a published PAN model run over a supervised-plus-agent scenario, adding a verifier to the autonomous agent …

In a published PAN model run over a supervised-plus-agent scenario, adding a verifier to the autonomous agent removed roughly 45.8% of steady-state harm and a coordinated governance package roughly 43.0%, while upgrading the model alone removed roughly 6.3%.

Illustrative PAN-run result: PAN Social Work User Guide (v6.2.11.6), lever-ranking walkthrough (fig03). Direction-and-shape only — not a calibrated prediction for any specific deployment.

ScenarioAcross a published PAN sweep of 216 deployment topologies (432 sampled deployments), system-layer governance o…

Across a published PAN sweep of 216 deployment topologies (432 sampled deployments), system-layer governance outperformed equal-effort model-layer improvement in 99.1% of cases at roughly 3.4x mean leverage - a figure the project corrected downward from an earlier 9.4x estimate.

Illustrative PAN-run result: PAN Normalized Baseline Analysis; README direction-and-shape summary. Direction-and-shape only — not a calibrated prediction for any specific deployment.