Does AI adoption actually reduce the work that employees do?

ActivTrak's Productivity Lab reports that across the organizations it monitors, AI adoption has risen sharply: "80% of employees now use AI tools at work," average time spent in AI tools is "up eightfold," and the typical organization now runs "seven or more AI tools on average, up from two in 2023." Over the same window, the behavioral trace data ActivTrak collects from employee computer activity show work intensifying, not easing: "time spent across work applications increased between 27% and 346%," including a 104% rise in email, a 145% rise in chat and messaging, and a 94% rise in business-management-tool use; collaboration "surged 34%," multitasking rose 12%, weekend work rose "more than 40%," and "focus time fell to a three-year low" — AI users' average daily focused time down 23 minutes.

The report treats this as a direct rebuttal of expectation: "Despite expectations that AI would reduce workloads, the findings show otherwise: AI is amplifying work activity across nearly every category measured." Its account is that adoption expands the surface area of work — more tools, more channels, more hours — faster than it removes any single one. The same data show the workday's edges eroding: weekday starts moved earlier (7:48 a.m. versus 8:02 a.m.), Saturday productive hours rose 46% to 4h37m with starts pulled to 7:11 a.m., and Sunday hours rose 58% with starts pulled to 10:58 a.m. ActivTrak's Chief Customer Officer, Gabriela Mauch, states the conclusion directly — "AI isn't reducing work, it's increasing the speed and density of how work happens" — and names an "AI measurement gap": most organizations "lack reliable data on how AI is actually changing productivity, focus and workforce capacity."

This sits in tension with Does generative AI shift knowledge workers away from communication?, which reads Microsoft 365 trace data as a shift away from communication toward solo documentation. ActivTrak's trace data, from a different and unspecified customer base, show communication and coordination activity (email, chat, collaboration) among the fastest-rising categories, not falling relative to others — the two are reconcilable only if total work is expanding enough for both documentation and communication to grow at once, which is what ActivTrak's "amplifying" framing implies but doesn't test directly against the Microsoft data. The focus-time decline also runs alongside rather than through the mechanism in Does AI really save time, or just change how we spend it?: that note locates reallocated time inside a single task, while ActivTrak describes reallocation across the whole workday and week, with no task-level account of where the lost minutes go.

The excerpt gives no sample size, no definition of "AI tools" or "focus time," and no description of which industries or organizations ActivTrak's customer base represents, so none of the percentages generalize beyond whatever population the Productivity Lab monitors. The findings are correlational — adoption and activity both rise over the same multi-year window, with no comparison group — so "AI is amplifying work activity" describes an observed pattern, not a demonstrated cause. ActivTrak sells workplace-analytics and productivity-monitoring software, and its call to close "the AI measurement gap" doubles as a pitch for the visibility product it sells, a reason to read the framing, if not the trace numbers themselves, as interested. What the data support, at the strength available, is that within this monitored population, AI adoption has coincided with more work, not less.