AI boosts coding productivity but creates human bottlenecks, Harvard study finds
A study by Harvard University researchers Fiona Chen and James Stratton reveals that AI tools boost coding productivity by 20-30%, but human bottlenecks in the review process are slowing down overall progress. The findings suggest that AI adoption shifts workloads rather than reduces them, with code reviewers facing increased demands.
Editor, Lazyfounder
30 SEC SUMMARY
- Harvard University research reveals AI boosts coding productivity by 20-30% in metrics like lines of code, commits, and pull requests.
- Human code review bottlenecks slow down progress, increasing review times from 7 to 10 days on average.
- AI-generated code requires more corrections, raising questions about its overall efficiency in software development.
- Workloads shift rather than reduce, with a 14% increase in employees involved in code reviews.
- No significant job losses reported, but AI adoption reshapes roles in software teams.
TABLE OF CONTENTS
KEY HIGHLIGHTS
- AI assistance increases lines of code, commits, and pull requests by 20-30%, according to Harvard University research.
- Human code review times rise from 7 to 10 days on average due to bottlenecks, delaying software completion.
- Comments per pull request increase by 49%, and reviewer comments rise by 35%, signaling more corrections for AI-generated code.
- The proportion of employees involved in code reviews grows by 14%, indicating a shift in workload rather than a reduction.
AI’s impact on coding productivity
Researchers at Harvard University have found that AI tools significantly boost coding productivity, with metrics like lines of code, code commits, and pull requests increasing by 20-30%. According to TechRadar, these gains are attributed to AI’s ability to accelerate initial code generation, allowing developers to produce more in less time.
Human bottlenecks in code reviews
Despite the productivity gains, the study reveals a critical bottleneck in the code review process. The average time to review and merge pull requests has increased from 7 to 10 days. TechRadar reports that this delay is due to a surge in pull requests—up by 23%—and a 49% increase in comments per pull request. Reviewer comments have also risen by 35%, suggesting that AI-generated code requires more corrections and scrutiny.
Workload shifts, not reductions
The research highlights that AI adoption does not reduce workloads but instead shifts them. The proportion of employees involved in code reviews has increased by 14%, indicating that more team members are now engaged in reviewing and refining AI-generated code. According to the findings, this shift has not led to significant job losses but has altered the dynamics of software development teams.
Questions about AI code quality
The study raises concerns about the quality of AI-generated code. TechRadar notes that AI-assisted code often requires substantially more corrections, which could undermine the efficiency gains. Despite the increase in code output, there is no statistically significant improvement in completed software, suggesting that human oversight remains essential to ensure quality and functionality.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
For founders and engineering leaders, this research underscores that AI-assisted coding tools are not a silver bullet. While they accelerate code production, they also expose and amplify existing bottlenecks—particularly in human-dependent processes like code reviews. The findings suggest that simply adopting AI without rethinking workflows or scaling review teams could lead to diminishing returns. Teams may need to invest in automating parts of the review process or refining AI-generated code quality to realize the full benefits of these tools.
Key takeaways
- AI-driven productivity gains in coding are real but may not translate into faster software completion if human bottlenecks persist.
- Founders should anticipate shifts in team workloads rather than reductions when adopting AI coding tools.
- The quality of AI-generated code remains a concern, requiring additional review and correction efforts.
- Scaling code review capacity or automating review processes may be necessary to fully capitalize on AI productivity gains.
- Job roles in software teams are evolving, not disappearing—planning for upskilling and role adjustments is critical.
FAQ
Does AI reduce the workload for software teams?
No. The Harvard study found that AI adoption shifts workloads, particularly increasing the burden on code reviewers, rather than reducing overall team effort.
Why isn’t completed software improving despite AI productivity gains?
Human bottlenecks in the code review process delay progress, and AI-generated code often requires more corrections, offsetting the productivity gains.
Has AI adoption led to job losses in software development?
According to the research, AI adoption has not resulted in significant job losses but has reshaped roles, particularly increasing the proportion of team members involved in code reviews.
Related on Lazyfounder
Sources
- TechRadar · 2026-10-12
AI agents are great at writing new code — but then the human bottleneck slows everything down
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.
