Pushing every employee toward maximum AI use does not necessarily make a company more productive, according to ActivTrak CEO Heidi Farris writing in Fortune. Farris's Productivity Lab tracked 120,620 employees across 1,009 organizations over three straight quarters, from the final quarter of 2025 through the second quarter of 2026, to see how AI adoption actually changes the way people work day to day.
The data split workers into stages: 27 percent used AI mainly for research assistance, 14 percent used it to execute specific tasks, and just 2 percent had reached full workflow integration, where AI is woven into how they operate rather than used as a discrete tool. In total, 43 percent of workers were using AI in some form, while 57 percent rarely or never touched it.
The counterintuitive finding was that productivity does not simply climb alongside adoption. Healthy utilization peaked around 75 percent adoption maturity, then dropped roughly 5 percentage points once AI became fully embedded in a worker's process. Farris argues the sweet spot for most employees sits somewhere in the middle rather than at full automation, and that companies pushing everyone toward the highest stage may be working against their own goals.
She points to two specific risks in over pushing adoption: runaway infrastructure costs from deploying AI more broadly than jobs actually require, and what she calls operational disconnect, where increasingly sophisticated AI workflows make individual tasks faster without actually improving the broader business process around them. On a more encouraging note, the data showed AI usage sticks once it starts, with 82 percent of employees who adopted AI continuing to use it quarter after quarter.
Farris's takeaway for leadership teams is to match AI adoption levels to what a given role actually needs rather than treating maximum use as the universal goal, framing AI maturity as something to calibrate per job rather than maximize company wide.