AI ROI
Measuring the real return of agentic practice through honest proxies (session duration, commit ratio, tokens saved) instead of vanity metrics (number of prompts).
Definition
AI ROI is the attempt to concretely answer "does agentic practice pay back more than it costs?" with session data rather than impressions. Honest proxies are the ones that correlate with shipped work: session duration distribution, the share of sessions that end in a commit, tokens saved by filtering context upstream. Vanity metrics — number of prompts sent, hours "spent with AI" — measure usage, not value produced.
postcursors perspective
Without measurement, AI tooling debates are anecdote contests. "It changed how I code" isn't actionable data. AI ROI only matters if the numbers come from actually tracked sessions, over a period long enough to smooth out edge cases — a month of occasional use doesn't compare to a quarter of logged sessions.
In practice
Across 267 tracked sessions, 62% last under 10 minutes — most agentic work isn't the long, immersive session people picture. Only 14% of sessions actually produce a commit, which reframes what "using AI" means day to day. On the cost side: a local proxy (RTK) saved 4.4M tokens (54%) across 7,500 commands over a quarter — a far more actionable figure than a rough estimate.
Common misconceptions
- ✗ Using the number of prompts or hours 'spent with AI' as a success metric — that measures usage, not value produced
- ✗ Comparing tools based on personal anecdotes without equivalent session data on both sides