# Harvard study shows AI code assistants add work without speeding delivery

> Research of 718 firms finds AI-generated code raises code volume but also boosts code‑review effort, limiting overall productivity gains.

Oossa · 2026-10-22 · https://oossa.com/en/harvard-study-shows-ai-code-assistants-add-work-without-speeding-delivery

A Harvard study using Jellyfish’s analytics platform examined 718 companies and about 300 million work events from January 2021 to March 2026. AI coding assistants raised lines of code by 12% and commits by 9%, while AI agents lifted lines of code by 30% and commits by 20%. However, the number of Jira issues resolved per worker rose only 12% with agents and not at all with assistants. The extra code created longer pull‑request reviews – review time grew 49%, comments per PR rose 53%, and the share of PRs needing changes doubled.

## The facts

- Study covered 718 firms and ~300 million work events (Jan 2021–Mar 2026).
- AI agents increased code lines by 30% but only raised resolved issues per worker by 12%.

## Why it matters

Engineering managers see that adding AI code generators may require more review capacity, so they must plan for extra QA resources.

## Sources & references

1. [Study Finds AI-generated Code Fails to Boost Productivity Due to Downstream Bottlenecks](https://www.infoq.com/news/2026/10/jellyfish-ai-productivity/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering) – InfoQ, 2026-10-22

Last updated: 2026-10-22