Skip to content
The GOVENANT Standard

Motion vs. delivery

Motion is what an AI agent does: runs completed, drafts written, decisions logged. Delivery is what verifiably shipped: the email that truly sent, the outcome row that exists in the record. The diff between the two is the primary health signal of any agent system — and the axis on which performed autonomy is detected.

The rule

Under the GOVENANT Standard, a unit of work is not “done” when the agent says so; it is done when the substrate can verify the outcome exists (Law 2 — Assertion). Dashboards MUST show runs completed vs. outcomes verified per terminal edge — and the two numbers must be able to diverge, because their divergence is the finding.

Why activity metrics mislead

An agent platform can complete thousands of runs, all logged “success,” while shipping nothing: queues without consumers, gates that delivery routes around, completions keyed to the agent’s return value instead of a verified outcome. Every one of these renders motion as delivery. The standard’s audit instrument (Part 9) exists to catch exactly this class of failure — its rule of evidence: ground truth is the database and the code, never the logs’ self-report.

Performed autonomy · outcome-keyed completion (Part 5) · the money query (Part 9)