We've cleared the highway, but the toll gate remains unchanged.

After we started using AI heavily, the time spent waiting for code reviews increased fivefold.

These numbers come from Faros AI, a US engineering data analytics firm, in its Q2 report this year. Over two years, they examined real work records from 22,000 developers across roughly 4,000 teams, comparing low AI usage versus high AI usage within the same organizations. Processed work grew by almost 34%. Yet review wait times grew fivefold, and weekly deployment frequency actually fell by nearly 12%. In other words, they produced more but shipped less.