Every team that adopts coding assistants reports the same thing in week two: it feels faster. The question a CFO asks is different: faster at what, compared to when, and did quality hold?
We define velocity as accepted deliveries per comparable period. Not tickets closed, not lines written. A delivery counts when the business accepted it. That single definition removes most of the arguments.
Alongside it we track three guardrails: cycle time from definition to production, defects found after release, and rework hours. If cadence goes up while any guardrail degrades, the pilot has not succeeded yet.
The pattern that actually produces the multiple is not the tool. It is context: clear standards, documentation an agent can read, and tasks scoped small enough to review in one sitting. Assistants amplify whatever process they land in.
Only after the four numbers hold for a full month do we publish the multiple. Until then it is a hypothesis, and we say so.