Harness Report Finds AI Productivity Gains Outpacing Developer Measurement Systems

Harness announced in a press release the release of The State of Engineering Excellence 2026, a study showing that AI coding tools have improved developer productivity faster than engineering organizations can measure it. The report, based on responses from 700 practitioners and managers across five countries, highlights a growing gap between reported productivity gains and the ability to track their real impact.

Eighty nine percent of engineering leaders said productivity improved after adopting AI tools, and 88 percent said developer satisfaction rose. However, 81 percent reported that developers spend more time in code review, with nearly a third of that work going untracked. The study found that about 31 percent of developer time is now spent on unmeasured activities such as reviewing AI generated code and fixing related bugs.

Despite confidence in existing frameworks, 94 percent of leaders admitted that current metrics overlook critical factors such as technical debt, validation time, and burnout. Only 6 percent believe their current systems can adapt to these challenges. The report identifies measurement itself as the biggest issue, with leaders citing difficulty in assessing true productivity impact and maintaining code quality.

Developers expressed concern about how AI productivity data might be used. More than half fear it could influence individual performance evaluations, and nearly half reported pressure to work faster or concerns about privacy. The report recommends that organizations update their measurement frameworks to include indicators for code quality, cognitive load, and burnout, and to clearly separate improvement tracking from performance evaluation.