Enterprise AI Measurement Guide
Agent Effectiveness
Cost per Outcome
How do we calculate the true cost per unit of engineering output from AI coding agents, and how does that cost vary across the different tools our team uses?
What it shows
Cost per Outcome calculates the dollar cost of each merged pull request produced with AI coding agent assistance, dividing total agent spend by the number of sessions that resulted in a merged PR. This converts AI tool cost from an aggregate budget number into a unit economics metric: what does it actually cost, per shipped outcome, to operate AI coding agents in this engineering organization?
Why it matters
The AI coding agent conversation has been dominated by adoption rates and per-seat costs. Neither of those metrics answers the question CFOs are now asking: what are we getting per dollar spent? Cost per Outcome is the answer. An organization paying $29.62 per merged PR from one agent and $10.97 from another, at identical adoption rates, has data to make a tool selection decision. Without this metric, that decision gets made on vendor relationships or developer preference.
The Larridin angle
Larridin's Agent Effectiveness tab is the only view that connects per-tool spend to per-tool outcome rates in the same table, making Cost per Outcome calculable per app, per engineer, and over time, rather than only at the aggregate level.