In 2025, something unsettling happened to a group of veteran doctors in Poland: They started missing things. These weren’t rookies. They were highly experienced endoscopists, each with thousands of procedures behind them. Yet after just a few months of heavy reliance on AI-assisted tools, their ability to spot precancerous polyps without AI fell by a relative 20%.
The corporate world may be suffering from the same affliction. Employees are increasingly handing the heavy lifting to chatbots, drawn by their speed and polished output. The hidden cost? The underlying human capability can quietly begin to erode. MIT Media Lab has given this phenomenon a chilling name: cognitive debt.
The hidden cost of outsourcing your brain
Sara Connell has been tracking this decay firsthand. As an executive coach who works with leaders at GE and Johnson & Johnson, she noticed that relying on external processing is fundamentally changing how professionals operate. Her upcoming book, The Download, tackles what happens when we outsource so much of our thinking that our actual brains start to atrophy.
Connell isn’t anti-tech by any stretch. However, she points out a biological reality often ignored: You can’t send a machine to the gym and expect your own muscles to get stronger.
I asked Connell what this cognitive debt looks like on the ground for a manager. It usually starts with a sudden, bland sameness across the board.
Everyone’s memos begin to share that distinctive, slightly soulless AI cadence, with staccato sentences, heavy reliance on em-dashes, and weirdly poetic corporate jargon. But beyond the writing style, there is a deeply troubling loss of ownership. An employee might hand you a beautifully formatted strategy doc. Ask them to defend a specific assumption, however, or pivot the idea on the fly, and they often freeze. They can’t reconstruct the argument because they didn’t actually build it.
Microsoft researchers surveyed more than 300 knowledge workers and found evidence of the same pattern: The more people rely on generative AI, the less likely they are to engage in critical thinking themselves.
Sometimes the best AI strategy is a walk
So how do we fix it? Connell’s approach doesn’t require an expensive enterprise software licence or a weekend corporate retreat. It takes 10 minutes. She calls it the download walk.
The mechanics are almost absurdly simple. You decide on a problem you need to solve – a strategy snag, a hiring decision, a creative roadblock. Then, you step away from your desk and walk. The only hard rule is zero inputs. No podcasts, no taking a quick phone call, and absolutely no ChatGPT. Just you and ambient reality.
This isn’t just wellness woo-woo. Stanford research found that walking increases creative output by roughly 60%. The brain desperately needs white space to shift out of high-anxiety beta waves and into the alpha and theta states. That loosening of attention is where actual original insights tend to arrive.
Your brain should write the first draft
Before you even open a browser tab to start your next project, Connell suggested adopting a rule she borrowed from her son’s algebra teacher. It’s four words: Why. Try. Then, AI.
Figure out the actual purpose of the task. If it requires strategy or generating a novel idea, you must write the first draft yourself. Form your own argument. If you ask a bot for 10 ideas before you’ve come up with one of your own, you are instantly reacting to the machine’s boundaries instead of exploring your own. You can use the AI afterwards to format, systemise, or pressure test the logic.
Large language models are trained entirely on what humans have already created. If we stop doing the messy, frustrating work of thinking for ourselves, we risk a future of entirely derivative ideas.
Your best insights probably aren’t hiding in a better prompt. They might just be waiting for you to leave your phone on your desk and take a walk. – Inc./Tribune News Service
