From individual gains to company results

The evidence on individual productivity is becoming difficult to dispute. McKinsey’s The State of AI in 2026, published in August, reports that 80 percent of employees believe AI has increased their productivity.¹ Yet only 37 percent of companies report a positive contribution to operating profit, a share unchanged from the previous year.

This is not a contradiction. Individual productivity accumulates within a person’s working day. Company performance emerges through processes, roles, decisions and hand-offs. A task completed faster does not improve the income statement unless the surrounding system uses the time or capacity released.

Productivity has no automatic destination

AI can reduce the time needed to analyse a document, prepare a proposal or write code. The immediate benefit is real, but its destination is often undefined. The employee may use the saved time to refine the output, absorb additional requests, finish earlier or compensate for delays elsewhere. Each outcome feels productive. They do not create the same economic result.

The distinction matters because revenue growth and cost improvement require different choices. If AI is intended to support growth, released capacity has to move towards activities such as customer development, faster experimentation or higher service volume. If the intended outcome is lower cost, work has to be redesigned so that the same output needs fewer resources. Without that decision, measurement becomes ambiguous and the gain disperses across the working week.

This helps explain why broad access has produced uneven results. The binding constraint is increasingly the organization’s capacity to absorb change, rather than the capability of the technology. Processes designed around old task durations remain in place. Approval layers persist. Hand-offs continue at the same cadence. AI accelerates parts of the work while the operating model continues to govern the whole.

The organizational experience is uneven

Senior leaders can also misread adoption, because the same tool is experienced differently across levels. McKinsey found that 47 percent of middle managers and individual contributors report at least one negative effect from AI, compared with 31 percent of senior executives. Reported effects include pressure to take on more work, difficulty managing a larger volume of output, and mental fatigue.

That difference is consequential. A senior executive may experience AI as faster access to analysis and clearer preparation for decisions. One level down, the same capability can raise expectations without removing existing responsibilities. Work becomes heavier in ways conventional productivity measures rarely capture. Resistance may therefore reflect workload design rather than scepticism about the technology.

Licence counts and login rates provide little evidence of economic value. More useful signals include the share of work completed with AI, the proportion of output that requires correction, the cost per completed task, and early drop-off after initial adoption. Together they reveal whether AI is changing how work is performed or merely adding another layer to it.

For executives, the central question is no longer whether employees can become more productive with AI. Many already have. The question is where the released time, attention and capacity should go. Until that allocation is explicit, individual gains will stay visible in daily work and difficult to find in company results.

Footnotes

1. McKinsey, The State of AI in 2026, August 2026. ↩