An engineering leader writing in LeadDev compared one AI-assisted initiative with a similar one done without AI. Lead time fell by only about 4%, and one explanation the author gives is that coding was around 10% of lead time without AI, so faster coding changed little. The author suggests leads break lead time into wait, active and rework time, record a reason for each wait, and look for the real constraint before buying more tooling. It rests on one initiative in one organisation, so results will vary.
