Sources

Every statistic and quotation on this site comes from the list below. For each one we say what it found, what it cannot tell you, and how we checked it.

We read each source before citing it. Where we could only read an abstract, we say so, and we describe only what the abstract reports. Where a study has important limits, such as being a preprint, a laboratory task or a single company, we say that too.

All sources

  1. Harris, D. J., Allen, K. L., Vine, S. J., & Wilson, M. R. (2021). A systematic review and meta-analysis of the relationship between flow states and performance. International Review of Sport and Exercise Psychology. Read the source

    What it found
    Across 22 studies of sport and gaming tasks, flow and performance were positively related (pooled r = 0.31, 95% CI 0.24 to 0.38), a medium-sized relationship.
    What it cannot tell you
    The studies were of sport and gaming tasks, not office or leadership work. It is a correlation, and the authors report that no proposed mechanism has convincing evidence.
    How we checked
    Full abstract read (author preprint record). Retrieved .

  2. Lomas, T., Medina, J. C., Ivtzan, I., Rupprecht, S., & Eiroa-Orosa, F. J. (2019). Mindfulness-based interventions in the workplace: An inclusive systematic review and meta-analysis of their impact upon wellbeing. The Journal of Positive Psychology, 14(5). Read the source

    What it found
    Across 35 randomized controlled trials, mindfulness programs at work reduced stress (SMD -0.57) and burnout (SMD -0.36) and improved job performance (SMD 0.43) on average.
    What it cannot tell you
    The authors note that study quality was inconsistent and call for more high-quality trials. The programs studied were structured mindfulness courses, not advisory work like ours.
    How we checked
    Full abstract read. Retrieved .

  3. Killingsworth, M. A., & Gilbert, D. T. (2010). A wandering mind is an unhappy mind. Science, 330(6006), 932. Read the source

    What it found
    People are thinking about what is not happening almost as often as they are thinking about what is, and doing so typically makes them unhappy.
    What it cannot tell you
    Volunteers answering prompts on a phone app, not a representative sample of workers.
    How we checked
    Full abstract read. Retrieved .

  4. Harvard Gazette (2010). Wandering mind not a happy mind. Read the source

    What it found
    Reports that people spend 46.9% of their waking hours thinking about something other than what they are doing, across 2,250 volunteers and 250,000 data points.
    What it cannot tell you
    A news report on the Killingsworth and Gilbert study, which is cited separately.
    How we checked
    Article text read; figures and quote copied from it. Retrieved .

  5. Mark, G., Gudith, D., & Klocke, U. (2008). The cost of interrupted work: more speed and stress. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 107-110. Read the source

    What it found
    In a controlled experiment, interrupted people finished their tasks in less time with no difference in quality, but reported more stress, frustration, time pressure and effort.
    What it cannot tell you
    A laboratory experiment with an email task. It does not measure flow or how long it takes to regain focus.
    How we checked
    Full paper text read, including the abstract. Retrieved .

  6. Toyota Motor Corporation. Toyota Production System. global.toyota. Read the source

    What it found
    Toyota says that "all employees implement daily incremental kaizen" as part of its production system.
    What it cannot tell you
    Describes Toyota's own manufacturing practice. It is not evidence that kaizen works in other settings.
    How we checked
    Page text read; quoted sentence copied from it. Retrieved .

  7. Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science. Read the source

    What it found
    In a preregistered experiment with 453 college-educated professionals doing writing tasks, those given ChatGPT took 40% less time on average and their output quality rose by 18%.
    What it cannot tell you
    Short, incentivized writing tasks done online. Not a study of a whole organization or of long-term effects.
    How we checked
    Full abstract read. Retrieved .

  8. Brynjolfsson, E., Li, D., & Raymond, L. (2023). Generative AI at work (NBER Working Paper 31161, revised November 2023). Read the source

    What it found
    Across 5,179 customer support agents, access to an AI assistant raised issues resolved per hour by 14% on average, by 34% for novice and low-skilled workers, with minimal impact on experienced, highly skilled workers.
    What it cannot tell you
    One company's customer-support teams, in a working-paper version. Gains varied a lot between workers.
    How we checked
    Full abstract read on the NBER page. Retrieved .

  9. Dell'Acqua, F., McFowland, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science. Read the source

    What it found
    In a preregistered experiment with 758 consultants, on 18 tasks inside the frontier of AI capabilities, those using AI completed 12.2% more tasks, 25.1% more quickly, with significantly better quality. On a complex managerial task outside the frontier, those using AI were 19% less likely to produce correct solutions.
    What it cannot tell you
    Consultants at one firm, using GPT-4, on tasks designed for the study. The frontier moves as tools change.
    How we checked
    Full abstract read in the published paper. Retrieved .

  10. Lee, H., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. C. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. CHI 2025. Read the source

    What it found
    In a survey of 319 knowledge workers, higher confidence in generative AI was associated with less critical thinking, while higher self-confidence was associated with more.
    What it cannot tell you
    A self-reported survey, so it shows association, not cause.
    How we checked
    Full abstract read. Retrieved .

  11. Fang, C. M., Liu, A. R., Danry, V., Lee, E., Chan, S. W. T., Pataranutaporn, P., Maes, P., Phang, J., Lampe, M., Ahmad, L., & Agarwal, S. (2025). How AI and human behaviors shape psychosocial effects of extended chatbot use: A longitudinal randomized controlled study. arXiv:2503.17473 (preprint). Read the source

    What it found
    In a four-week randomized experiment (n = 981, over 300,000 messages), the assigned conditions had no significant effects, but participants who voluntarily used the chatbot more showed consistently worse outcomes on loneliness, emotional dependence and problematic use.
    What it cannot tell you
    A preprint, not yet peer reviewed as far as we checked. It is about chatbot companionship, and heavier use may reflect existing differences between people.
    How we checked
    Full abstract read; author list and preprint status confirmed on the arXiv record. Retrieved .