Could the widespread adoption of generative AI (Gen AI) tools in the workplace erode our cognitive abilities? New research suggests that worrying reality is a possibility.

A paper from Microsoft and Carnegie Mellon University researchers investigated the use of Gen AI tools by knowledge workers and explored how automation contributes to the deterioration of cognitive faculties.

Also: The work tasks people use Claude AI for most, according to Anthropic

The researchers sampled 319 knowledge workers across professions in computer science, education, entertainment/media, administration, and financial and business services. 

The participants shared three real-world examples of using Gen AI in their work. Participants were also asked to provide examples for each task type (creation, information, and advice) to increase the variety of examples collected. About 936 first-hand examples of using Gen AI at work were collected.  

Also: Crawl, then walk, before you run with AI agents, experts recommend

Participants also completed a survey assessing their confidence in Gen AI-assisted work tasks compared to their perceived confidence in finishing the same responsibilities without Gen AI. 

According to the researchers, "higher confidence in Gen AI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking." 

Furthermore, the researchers argued that mechanizing routine tasks via AI and "leaving exception-handling to the human user" deprives workers of practicing their judgment. Ultimately, this process weakens their cognitive musculature, "leaving them atrophied and unprepared when the exceptions do arise."

Unsurprisingly, OpenAI's ChatGPT and Microsoft's Copilot were the most-used tools and some of the tasks cited in the paper include a trader using ChatGPT to "generate recommendations for new resources and strategies" to hone their skills, a lawyer using ChaGPT to find relevant laws for a particular case, a teacher using DALL-E to create images for a presentation about washing hands at school, and a nurse who "verified a ChatGPT-generated educational pamphlet for newly diagnosed diabetic patients."

Also: The work tasks people use Claude AI for most, according to Anthropic

The researchers found that "knowledge workers view critical thinking as ensuring the objectives and quality of their work."When workers partake in critical thinking, "their confidence in themselves doing and evaluating the task, and their general tendency to reflect on work strongly correlated with their perceived enaction of critical thinking." 

Key takeaways 

Generally, workers said they had more confidence in Gen AI-assisted tasks when they believed they enacted critical thinking.

When users had less confidence in AI-assisted responses, they used critical thinking to mitigate the quality of AI's output. About a third (36%) of surveyees used critical-thinking skills to mitigate potential adverse outcomes from using AI. 

However, while "it is possible that fostering worker's domain expertise and associated self-confidence may result in improved critical thinking when using Gen AI," the researchers wrote, users with access to Gen AI tools were also more likely to produce "a less diverse set of outcomes for the same task, compared to those without." 

Also: Cerebras CEO on DeepSeek: Every time computing gets cheaper, the market gets bigger

The research highlights a shift in knowledge workers' cognitive effort. Increasingly, professionals only intervene when AI outputs are not up to par. The researchers suggested that an overreliance on Gen AI in knowledge work is shifting AI use cases "from task execution to oversight." 

The researchers said: "Surprisingly, while AI can improve efficiency, it may also reduce critical engagement, particularly in routine or lower-stakes tasks in which users simply rely on AI, raising concerns about long-term reliance and diminished independent problem-solving."

Artificial Intelligence

The best AI for coding in 2025 (and what not to use - including DeepSeek R1)

I tested DeepSeek's R1 and V3 coding skills - and we're not all doomed (yet)

How to remove Copilot from your Microsoft 365 plan

How to install an LLM on MacOS (and why you should)

• The best AI for coding in 2025 (and what not to use - including DeepSeek R1)
• I tested DeepSeek's R1 and V3 coding skills - and we're not all doomed (yet)
• How to remove Copilot from your Microsoft 365 plan
• How to install an LLM on MacOS (and why you should)