|
A Harvard study of 718 companies found that engineers using AI coding agents produced 30 percent more code, yet overall work output did not increase and code review time lengthened by 49 percent, shifting the productivity bottleneck from coding to the review stage. The findings suggest that legal AI vendors should measure review-stage metrics like revision rates and reviewer time alongside output metrics to assess true productivity gains. The research indicates that for AI tools to deliver real value, organizations need to address the review and approval checkpoint, not just the drafting speed.
Read Full Article →
← Back to latest