# Engine performance

> How well the automation is actually doing on your books, with every figure defined.

Section: Continuous bookkeeping · Canonical: https://intubu.intuitivecapital-dai.com/docs/engine-performance

Settings → Engine Performance.

Measured from what the system actually did on *your* books — not a benchmark, not a marketing figure.

**Every figure carries its definition.** "98% accurate" means nothing without knowing what was counted, over what period, and what an error was taken to be. Each metric here states that beside the number.

What it reports: how much was automated versus reviewed, how often an automated decision was later corrected by a person, where confidence sat when it got things wrong, and which categories it is weakest on.

**A falling number is useful information, not a fault to hide.** Automation degrades when a business changes — a new revenue line, a new vendor pattern — and the point of measuring is to notice before the close does.

Thresholds are tuned in [Confidence thresholds and evidence](https://intubu.intuitivecapital-dai.com/docs/ai-confidence).