AI Intensity is a per-day composite score that summarizes how much AI activity took place that day. Leanmote uses it as the X-axis when plotting AI usage against delivery metrics like Cycle Time, Throughput, and Bug Rate.
What it measures
AI Intensity is not a percentage of PRs. It is a dimensionless score built from four separate signals on each calendar day, summed after a logarithmic transform so that no single signal dominates.
How Leanmote calculates it
ai_intensity(day) =
log1p(org_metrics_total)
+ log1p(active_users)
+ log1p(ai_commit_count)
+ log1p(ai_loc / 100)
org_metrics_total — total AI tool events recorded across the organization that day (Copilot interactions and generations, Claude Code session activity, Cursor activity).
active_users — distinct users with at least one AI event that day.
ai_commit_count — count of commits tagged as AI-assisted that day (see AI-assisted PRs for tagging).
ai_loc — AI-attributed lines of code changed that day. Divided by 100 before the log transform so it doesn't outweigh the other terms.
The output is a positive number on a roughly logarithmic scale. Higher values mean more AI activity that day; the absolute number doesn't have an intuitive unit, which is why it's most useful as a relative comparison across days.
The scatter plots
AI Intensity is plotted against three Y-axes, one chart each:
vs Cycle Time — does higher AI Intensity correlate with shorter task cycle time?
vs Throughput — does higher AI Intensity correlate with more PRs merged per day?
vs Bug % — does higher AI Intensity correlate with more or fewer bug-tagged PRs?
Each point is a calendar day. These are correlation views, not causal — read the scatter for patterns over many days, not single-day verdicts.
How to interpret it
Negative slope on Cycle Time — high-AI days tend to have shorter cycle times. Encouraging, but a correlation, not proof.
Positive slope on Throughput — high-AI days ship more PRs.
Cloud of points / no clear slope — AI Intensity isn't a strong predictor of delivery on this team. Investigate per-team or per-repo subsets.
Reverse slopes — high-AI days correlate with worse delivery on this team. Worth a serious look at how AI is being used.
What to do about it
Treat AI Intensity as a relative ordering of days, not an absolute percentage. Compare days within the same period; don't compare absolute scores across very different periods.
Don't draw conclusions from short windows. Quarterly is the minimum reasonable view.
Pair with team-by-team breakdowns; org-wide patterns sometimes hide opposite team-level patterns.
Related metrics
AI-assisted PRs
AI-assisted Commits
AI Lead Time Comparison
Cycle Time
Throughput