More AI code, same software output: what a 700-firm study found
Your AI coding agent is writing more code than ever. According to a new study from Harvard University, that extra code is not translating into more shipped software.
What the Study Measured
Researchers Fiona Chen and James Stratton analyzed aggregated engineering data from Jellyfish, a platform that tracks granular output across engineering teams. The dataset covers 300 million individual work events (commits, pull requests, and issue management activity) across more than 700,000 employees at over 700 software development firms, spanning 2021 through March 2026.
The researchers tracked exactly when each firm introduced AI coding assistants or AI coding agents, then ran a difference-of-differences regression to measure before-and-after changes on key output variables.
The Numbers
The raw code volume jump after AI agent adoption is hard to miss:
- +30 percent in total lines of code generated
- +20 percent in total commits
- +23 percent in pull requests
On the surface, that looks like a productivity win. The problem shows up one layer down. Resolution rates for Issues and Epics tracked in tools like Jira did not change in a statistically significant way after AI tools were introduced. The researchers also found no shift in the size or complexity of those tracked issues, ruling out the possibility that teams were simply tackling harder problems.
Where the Efficiency Goes
The study identifies human code review as the bottleneck that swallows the gains. More AI-generated code means more review volume, more pull request revisions, and more reviewer comments. The efficiency added during the coding phase gets absorbed downstream before it reaches actual software output.
The researchers conclude there is little evidence that firms increase software output or reduce employment by using AI coding tools, at least at current adoption levels and with current review practices.
For solo operators and small teams shipping their own products, the implication is direct: adding an AI coding agent to your workflow without adjusting how you review output may produce more code to manage rather than faster product delivery.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
