Study Finds AI Productivity Gains Lead to Increased Workloads
A UC Berkeley ethnographic study of 200 tech employees found that time saved by AI tools often results in higher management expectations and increased workloads without corresponding headcount adjustments. The study suggests that efficiency gains from AI can be "weaponized" to justify more demands on teams. This can undermine morale and lead to burnout if not managed with a focus on reinvesting saved time into strategic initiatives.
- Research from Faros AI, analyzing over 10,000 developers, found that while individual developers with high AI adoption merge 98% more pull requests, the review time for those pull requests increases by 91%, creating a significant bottleneck in the development lifecycle. - A study by Uplevel Data Labs on 800 developers using GitHub Copilot found no significant improvement in pull request cycle time or overall throughput and noted a 41% increase in bugs within pull requests. - While a GitHub study showed developers completed tasks 55% faster with Copilot, other research indicates a "productivity plateau," with initial 10% gains in team productivity not increasing further over time, even as the volume of AI-authored code grows. - A 2024 survey revealed that employees who frequently use AI report a 45% higher rate of burnout than their colleagues who do not use AI tools. - The amount of AI-authored code in production environments is growing, with one study of 4.2 million developers showing it increased from 22% to 26.9% in a single quarter. - The marketing automation company Klaviyo implemented a company-wide initiative where 1,800 employees were required to develop and demo a value-driving AI use case, tying the effort to their performance reviews to ensure adoption. - AI tools are being positioned as a way to combat burnout by automating repetitive tasks, providing smart break recommendations, and monitoring workload distribution to reallocate tasks when an employee is overburdened. - Some studies show a disconnect between perceived and actual productivity; one randomized controlled trial found that experienced developers using AI tools took 19% longer to complete tasks, despite believing the tools made them 20% faster.
