"The Agentic Coding Productivity Illusion": Why Teams Overestimate AI Coding Gains

Summary

A widely-shared critique published jointly in InfoQ and The Pragmatic Engineer argues that current productivity metrics for agentic coding tools systematically overstate actual gains. The authors, based on data from 23 engineering teams across 8 companies, found that raw code output metrics (lines per day, PRs merged) improve 40-60% with agentic coding tools, but that review burden, integration debugging, and context-switching overhead nearly erase those gains when measured end-to-end across the