Upwork Research Institute Survey
The Upwork Research Institute's 2024 survey of 2,500 knowledge workers and business leaders documented the gap between AI's promised productivity gains and its lived reality. Seventy-one percent of full-time employees reported burnout. Ninety-six percent of C-suite leaders expected AI tools to increase productivity. Forty-seven percent of AI users said they had no idea how to achieve the productivity gains their employers expected. The data captured a structural condition: the expectation had outrun the capacity, the promise had outrun the delivery, and the workforce was absorbing the gap as personal inadequacy rather than as organizational mismeasurement. The findings provided the first large-scale empirical counterpoint to the executive narrative of AI-enabled productivity acceleration.
In the AI Story
The survey's findings appear to contradict the extraordinary individual-productivity gains that AI users consistently report — the engineer who ships in two days what previously took a week, the analyst who generates in hours analyses that previously required days. The contradiction resolves once fragmentation and cumulative cognitive cost are accounted for. AI tools increase output per unit of engagement while also increasing the total volume of engagement, and the additional engagement carries costs that offset — and in many cases exceed — the per-unit gains. Workers produce more and understand less, complete more tasks and master fewer domains, generate more artifacts and develop less judgment.
The survey also illuminated the role of executive expectation-setting in producing the gap. Ninety-six percent of C-suite leaders expected productivity increases; actual productivity increases, measured through worker self-report and objective output metrics, fell substantially short of these expectations. The gap was absorbed not by executives revising their expectations but by workers absorbing additional hours to attempt to meet them. The Upwork data documented this absorption: workers were spending more time with AI tools precisely to produce the gains their organizations expected, accumulating exactly the cognitive debt the simulation's framework predicts.
The survey became a reference point in the 2024–2026 debate about AI's actual organizational impact, cited in subsequent research including the Berkeley study that documented similar patterns in greater ethnographic detail. Together, the two studies established the empirical foundation for arguments that AI deployment without designed recovery produces intensification rather than efficiency gains — the exact finding Perlow's framework predicts and the exact condition the AI Practice framework exists to address.
Origin
The Upwork Research Institute conducted the survey in partnership with external researchers in spring 2024, sampling 2,500 workers and leaders across industries and geographies. Results were published in summer 2024 and circulated widely in business press and organizational-research literature through 2025.
Key Ideas
Burnout at scale. Seventy-one percent of full-time employees reporting burnout indicates systemic rather than individual pathology.
The expectation gap. Ninety-six percent of executives expected AI-enabled productivity gains that a plurality of workers had no idea how to produce.
Absorbing the gap as hours. The gap between expectation and delivery was closed by workers extending their engagement with AI tools, accumulating cognitive depletion.
Corroborating ethnographic findings. The survey data aligned with embedded-observation research showing AI tools intensifying rather than reducing total cognitive load.
Appears in the Orange Pill Cycle
Further reading
- Upwork Research Institute, AI-Enhanced Work Models Survey (2024)
- Xingqi Maggie Ye and Aruna Ranganathan, "AI Doesn't Reduce Work — It Intensifies It," Harvard Business Review (February 2026)
- Christina Maslach and Michael Leiter, The Burnout Challenge (Harvard University Press, 2022)
- Leslie Perlow, Sleeping with Your Smartphone (Harvard Business Review Press, 2012)