How Tempo calculates AI commit rate
Tempo first removes merge commits because they are repository operations rather than a useful signal of code creation. It then counts commits with a positive AI detection and divides that count by the remaining commits.
Organization and team rates are calculated from the underlying commit counts, not by averaging repository percentages. This prevents a small repository from carrying the same weight as a high-volume repository.
How to interpret the number
A rising rate means detected AI assistance is appearing in a larger share of commits. It does not, by itself, show that delivery is faster, quality is higher, or cost has fallen.
Interpret adoption alongside PR cycle time, PR size, review behavior, change failure signals, and team context. The useful comparison is whether the delivery system changes as adoption changes.
Detection coverage matters
AI commit rate is only as complete as the evidence available. CLI file matching provides stronger coverage than commit-message forensics alone. An undetected commit should be read as “no supported signal found,” not “proven manual.”
Useful ways to segment AI commit rate
The organization-wide number is a starting point. Repository, contributor, tool, and time-window breakdowns explain where adoption is happening and whether it is sustained.
- Compare repositories with similar work profiles.
- Track adoption before and after enablement or policy changes.
- Separate CLI-attributed work from pattern-detected work.
- Compare AI-assisted and unclassified PR outcomes rather than ranking people.