BOTTLENECKS
More Code Does Not Mean More Adoption
Wharton research found that AI coding tools sharply increase developer output, with newer agentic systems lifting coding activity far more than earlier autocomplete tools. But the gains translated into smaller increases in software projects and releases, and did not produce a matching rise in user adoption.
The finding is a useful correction to AI productivity hype. Writing code is no longer always the binding constraint; review, integration, distribution and customer iteration can absorb the gain. What follows is a different yardstick for AI work: shipped usefulness, not how much more material teams can generate.
Source: Knowledge at Wharton · September 8, 2026