How Can AI Reduce Cognitive Friction?

If AI can increase cognitive strain, it can also help reduce it. The difference is not necessarily the technology itself; it is how intentionally organizations choose to use it.

A strong temptation is to measure AI’s success in simple productivity terms: How much faster did the employee complete the task? How many more outputs can they produce? How much time did automation save?

But saving time is only useful if we decide what that time is for.

This matters because of Parkinson’s Law, which argues that work tends to expand to fill the time available for its completion. Give someone an hour to complete a task and it may take an hour; make that task significantly faster with AI, and there is a danger that the saved time simply becomes capacity for more work. While that may sound good on the surface, how do we ensure it is high-quality or the work that truly needs to be done?  How do we ensure human workers are thriving and not being overwhelmed by unrealistic output demands?

That is not an argument against AI. It is an argument for being more thoughtful about productivity.

Recent evidence suggests there is real potential here. The International Labour Organization’s 2026 review of empirical evidence on generative AI found that productivity gains are real, although uneven, while worker-reported time savings have not consistently translated into higher measured output. That gap creates an opportunity for employers, as perhaps the time AI saves does not always need to be converted into more output. Sometimes, it can become recovery time, thinking time, learning time or simply breathing room.

Achieving the most effective and beneficial balance between human work and AI work requires employees to be taught how to use AI effectively. AI literacy should mean more than knowing how to write a good prompt. People need to understand when AI is useful, when it introduces additional work, when its output needs challenge and review, and when tasks should be completed by human workers.

Consider a simple example: asking AI to write an email. It may save time and reduce the immediate effort involved in drafting, but if the underlying problem is that someone struggles with communication, AI may only address the symptom. And if the employee cannot confidently judge whether the AI-generated email is appropriate, accurate or persuasive, the cognitive burden hasn’t disappeared. It has simply moved from writing to evaluating, which does not guarantee the removal of miscommunication.

This is why organizations should deliberately preserve opportunities for independent thinking. Research from Microsoft examining 319 knowledge workers and 936 examples of generative AI use found that greater confidence in AI was associated with less critical-thinking effort. The researchers also found that AI shifted critical thinking toward activities such as verifying information, integrating AI responses and overseeing the task.

The implication for the workplace is significant, as AI should be supporting expertise, not quietly replace the development of it. Cognitive friction creates tension between human reasoning and machine-generated information; it can make thinking harder and result in cognitive offloading or erosion over time.

That could mean creating “AI-free” stages of certain processes, asking employees to develop an initial view before consulting AI, or deliberately assigning tasks where independent analysis and divergent perspectives are required.

The goal isn’t to make work harder for the sake of it. It is to protect the human capabilities that make AI useful in the first place.

Ultimately, reducing cognitive strain through AI isn’t about asking technology to do everything. It is about using technology to remove the right things: unnecessary administration, repetitive work and friction that drains attention without adding meaningful value.

The best AI strategy therefore isn’t the one that maximizes the amount of work humans can complete. It is the one that creates more space for humans to do their best work, and enough space to recover from it.

If you would like to discuss how we can help ensure that AI is being optimized to its best capabilities in your workplace, please get in touch with me at amanda@orgshakers.com