A survey conducted by METR between February and April 2026 of 349 technical workers—including software engineers, researchers, and academics—indicates that AI tools have increased the value of their work by a median factor of 1.4–2x. While participants reported a median 3x increase in task speed, the researchers noted that speed often overstates actual value, as AI may encourage users to spend time on lower-value tasks that have become easier to execute.

Respondents retrospectively estimated a 1.3x value increase for March 2025 and forecast a 2.5x increase by March 2027. The survey, which compensated 70% of participants with an average of $200, utilized multiple questions to ensure internal consistency, finding that most participants' estimates held up under cross-validation.

METR staff cautioned that these self-reported figures may be inflated. METR employees themselves reported lower gains than the broader sample, suggesting that awareness of previous research on the gap between perceived and actual AI productivity might lead to more conservative estimates. The report recommends that organizations prioritize high-quality surveys to better track the automation of AI research and development.

A survey conducted by METR between February and April 2026 of 349 technical workers—including software engineers, researchers, and academics—indicates that AI tools have increased the value of their work by a median factor of 1.4–2x. While participants reported a median 3x increase in task speed, the researchers noted that speed often overstates actual value, as AI may encourage users to spend time on lower-value tasks that have become easier to execute.

Respondents retrospectively estimated a 1.3x value increase for March 2025 and forecast a 2.5x increase by March 2027. The survey, which compensated 70% of participants with an average of $200, utilized multiple questions to ensure internal consistency, finding that most participants' estimates held up under cross-validation.

METR staff cautioned that these self-reported figures may be inflated. METR employees themselves reported lower gains than the broader sample, suggesting that awareness of previous research on the gap between perceived and actual AI productivity might lead to more conservative estimates. The report recommends that organizations prioritize high-quality surveys to better track the automation of AI research and development.