The productivity paradox: why AI is increasing cognitive load, not free time

I had a conversation recently with an EGM of people and culture at a council. Like many organisations they were rolling out AI tools across the business. Tasks that used to take an hour were now taking people five minutes. But instead of enjoying more free time, employees were suddenly expected to do more.

She also noticed that people were experiencing even more mental fatigue, despite the removal of admin heaavy simple tasks from the work day. An exhaustion and brain fry from the constant stream of strategic work. 

Unexpectedly, instead of pushing to automate everything with AI, this organisation is encouraging people to use their agents less.

The tasks you thought were the problem might have been the solution

What if the low-effort admin tasks that used to fill your day were actually the recovery time your brain needed? Clearing your inbox or formatting that report didn’t require much cognitive load, and that was the point.

Your brain needs variation in cognitive load the way your body needs variation in physical effort. When AI removes the low-effort tasks it creates an unbroken stream of high-intensity cognitive work. 

Research from Microsoft's 2025 Work Trend Index found that knowledge workers are already operating in environments shaped by overload. A 2026 study highlighted by Harvard Business Review identified what researchers call 'AI brain fry' - mental fog, slower decision-making, and exhaustion linked to the cognitive load of managing multiple AI tools and validating their outputs (HRD Connect, 2026). The effect was especially pronounced in HR, operations, and marketing.

A UC Berkeley longitudinal study found that 67% of workers who adopted AI tools in 2025 reported working more hours, not fewer, by the end of the year (AI Magicx, 2026). The tasks got faster, so the day got fuller. 

"The way we learn, and the way we build critical thinking capability, is by doing the work.

When we outsource all the routine tasks to AI, we stop understanding how the pieces fit together."

- Lisa Hammond, The Next Step

So what can you do to avoid cognitive load from AI?

The answer is to be more intentional with the use of AI tools. 

1. Audit what you have automated. Get clear on which tasks are genuinely low value, and which were helping your team stay grounded in the detail. 
2. Protect low-cognitive time. Inbox management, note-taking, document formatting: these are recovery stops as much as admin. Keep some of them.
3. Set output expectations based on capacity, not speed. If AI makes a task take five minutes instead of an hour, consider what else that hour could be used for. Deep thinking rather than just more output.
4. Watch for the signs. If your team looks tired, is more reactive than creative, and struggles to explain why - It may be a cognitive load problem.

The most advanced organisations are the ones using AI thoughtfully, with a clear view of what human capacity is actually for.