# AI Coding Agents Lift Code Output, but Jira Gains Lack Significance

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/29266
Published: 2026-10-09T21:03:05.000Z
Updated: 2026-10-09T21:03:05.000Z
Section: Technology

> After adoption, code production rose sharply, while resolved Jira issues and epics showed no statistically significant increase as review demands grew.

Adopting AI coding agents increased software-development activity but did not produce a statistically significant rise in resolved Jira issues or epics, highlighting a gap between code production and completed work.

After firms adopted AI coding agents, lines of code increased 30%, while commits rose 20% and pull requests climbed 23%. The analysis covered about 300 million work events involving 725,938 workers at 718 firms from January 2021 through March 2026.

The data included GitHub activity, Jira issues, Google Calendar events and usage of AI coding tools. It compared intermediate measures such as code, commits and pull requests with completed work represented by resolved Jira issues and larger Jira epics.

Resolved issues and epics showed small positive effects after adoption, but neither result was statistically significant. The analysis also found no evidence that the average size or complexity of issues changed.

Code review demands increased alongside code output. Average pull-request review time rose 49%, the share of pull requests requesting changes nearly doubled and comments per pull request increased 35%. The share of workers performing code reviews increased 14%.

The results identify code review as a potential bottleneck, but they do not establish that longer reviews caused the lack of statistically significant growth in completed software work. Engineers spend about 30% to 40% of their time writing code, with the remainder devoted to planning, review and testing.

AI coding assistants help developers write code, while AI coding agents can handle longer and more complex tasks with greater autonomy. Fiona Chen and James Stratton wrote that “both technologies increase coding productivity” but that “productivity gains do not fully pass through to changes in software output or employment.”

The analysis found no statistically significant employment change attributable to AI agents. Its estimates ruled out an overall employment decline greater than 2.9% and an engineering-employment decline greater than 13.8%.

The paper is dated Aug. 4, 2026, and is a Harvard job-market paper rather than a peer-reviewed journal publication. Its conclusions are based on aggregated and anonymized data.