# AI Adoption Increases Workload and Interruptions > Source: > Published: 2026-06-28 10:54:07+00:00 The operational takeaway from ActivTrak's largest-ever workplace study is counterintuitive: AI adoption correlates with more work activity, not less, and specifically with more fragmented, interrupted, multi-stream work. For practitioners designing AI systems, evaluating AI product ROI, or building telemetry for ML pipelines in enterprise settings, the data reframes what "productivity" means in practice. ### What happened ActivTrak's 2026 State of the Workplace report -- its fifth annual edition -- analyzed 443 million hours of digital workplace activity across 163,638 employees at 1,111 companies spanning finance, healthcare, insurance, and technology. A comparison cohort of roughly 10,584 employees showed what happened 180 days before versus after AI adoption: email activity rose **104 percent**, chat and messaging rose **145 percent**, and business management tool usage rose **94 percent**. Average focused, uninterrupted work session duration fell **9 percent** (to about 13 minutes). The Atlantic synthesized these findings and cited parallel UC Berkeley Haas research in which workers reported taking on tasks they had previously outsourced, doing work in evenings and waiting rooms, and concurrently supervising multiple automation flows. The Atlantic used the phrase "AI brain fry" to describe the resulting mental state. ### Technical context The ActivTrak dataset is behavioral, not self-reported -- it captures observed application activity, which makes it more reliable than survey-based estimates. The implication for AI tooling is twofold. First, session and state management become critical: users who fragment work into many micro-sessions incur repeated context setup costs, and systems that drop state or require re-prompting amplify cognitive overhead. Second, multi-agent orchestration surfaces a new consistency risk -- users supervising several automation flows concurrently create larger surfaces for prompt drift, data contamination, and cascading errors that may not appear in single-session evaluations. ### Measurement implications Teams measuring AI ROI with throughput metrics alone will miss the offsetting cognitive costs. Supplementing throughput with session continuity, context-reuse rates, and interruption frequency gives a more complete picture of whether AI deployment is net-positive for knowledge workers. ### What to watch Adoption curves for multi-agent orchestration toolkits, changes to session length and context-window reuse patterns as long-context models proliferate, and whether enterprise AI vendors begin shipping focus-mode or batching features that address the interruption problem identified here. ## Key Points - 1AI adoption raises task density and multitasking rather than freeing time: email rose 104%, chat 145%, business tools 94% among workers studied by ActivTrak. - 2Average focused-work session fell 9% after AI adoption, complicating claims of net productivity gain and pointing to design choices that reduce context-switching overhead. - 3AI product and tooling teams should instrument session continuity and interruption frequency -- not just throughput -- as primary UX and evaluation signals. ## Scoring Rationale Solid, data-backed workplace study covering 443M hours and 1,100+ companies, with a clear practitioner implication around session design and evaluation. Corroborated by multiple outlets. Downgraded from 6.8: it is a behavioral trend study, not a frontier model release or major industry decision, and the findings -- while counterintuitive -- are not entirely surprising given the trajectory of AI workload research. Practice with real Ad Tech data 90 SQL & Python problems · 15 industry datasets [Active Search Campaigns by BudgetEasy](/problems/sql/active-search-campaigns-by-budget) [High CPC Clicks & Poor Landing PagesMedium](/problems/sql/high-cpc-clicks-poor-landing-page) [Campaign ROAS by Attribution ModelHard](/problems/sql/campaign-roas-by-attribution-model) 250 free problems · No credit card [See all Ad Tech problems](/problems/datasets/adtech)