AI Teams Track 3 Metrics to Measure Hidden Software Costs
Updated
Updated · O'Reilly Media · Nov 11
AI Teams Track 3 Metrics to Measure Hidden Software Costs
3 articles · Updated · O'Reilly Media · Nov 11
Summary
Cost of change is presented as the main way to judge AI’s unintended effects in software development, tying maintainability, burnout and vendor lock-in to economic impact rather than coding style debates.
3 indicators anchor that assessment: value-add percentage, DORA metrics and token usage, with value-add percentage singled out as the most useful gauge of whether bug fixes, incidents and other non-value work are rising.
Monthly tracking can be lightweight, but the article says the signal is noisy—trends may take 3 to 6 months to emerge and about 1 year to confirm lasting improvement.
Scenario models should supplement those lagging metrics, including maintainability, excess code and lock-in cases; one example shows a 50% AI speed gain only breaks even if headcount falls 25% and maintenance costs double.
The piece warns value-add percentage is easy to game by reclassifying or deferring maintenance, arguing teams should use the metric to test AI adoption strategies, not judge individual managers.