Enterprise AI Measurement Guide
Team Performance
Output by Category
What proportion of our engineering team's AI-assisted output is going toward new features versus maintenance work, and is AI investment changing that mix over time?
What it shows
The Output by Category view breaks engineering output into four intent-classified buckets: Features (net new product work), Bug Fixes, KTLO (Keeping the Lights On, operational maintenance and support), and Unclassified. The percentage share each category represents shows what AI is actually helping the team do, not just whether output is going up, but whether the output mix is moving in a strategically useful direction.
Why it matters
A team with rising Output per Engineer is encouraging. A team with rising Output per Engineer but KTLO growing from 27% to 50% of that output is a different story. AI is making maintenance faster, not freeing capacity for product work. This is the innovation rate question: are AI tools buying the engineering org time to build, or just making it cheaper to maintain? The category breakdown is the only way to answer that question from system data rather than manager estimates.
The Larridin angle
Larridin's WorkGraph classifies engineering work by intent derived from actual commits, PRs, and ticket data, not from what engineers self-report or what project labels say. That makes the category breakdown reliable enough to act on, rather than a rough estimate built on survey responses.