Generative AI opens up almost limitless possibilities for content creation, promising to take over our most tedious and repetitive tasks. Sounds like paradise for leaders and specialists alike. But a fundamental question emerges: what impact will removing our daily grind have on our creativity and our ability to make sound decisions?
The Anatomy of Atrophy
How many phone numbers do you actually remember? I know just one—my wife's, which I memorized in case I ever lose my phone. Since Google Maps became ubiquitous, many of us can't drive to the seaside without a smartphone telling us where to go. When the battery dies, we lose not just navigation, but our confidence. Our sense of direction simply atrophies because we stopped using it.
Today, we're facing a far more serious challenge. AI promises to take over coding, content creation, even diagnosing diseases. The grand promise goes: the machine handles the boring stuff while you focus on innovation. But here's the trap. Your professional intuition and expertise didn't materialize from thin air—they were forged in the fire of thousands of repetitive tasks, stress, and learning from mentors. That struggle is precisely what built your competence.
When those challenges disappear, your cognitive capabilities will start declining faster than muscles after two weeks without training. The very mind that was supposed to give you an edge over AI begins losing its clarity of judgment. Matt Beane from the University of California warns that AI is disrupting the master-apprentice relationship. When a junior developer uses Copilot to generate code they couldn't write themselves, they bypass the mental process essential for learning. It's like trying to build physical strength by watching a robot lift weights.
Warnings from the Cockpit and the Operating Room
The erosion of human skills through automation isn't new, but it's now happening on a massive scale. A tragic example is the Air France Flight 447 disaster. When the autopilot disengaged due to sensor failure, highly trained pilots succumbed to the "startle effect." Despite numerous alarms, they couldn't respond correctly—their manual flying skills, unused in daily operations, had simply atrophied.
A similar "switching off" of vigilance is appearing in medicine. A 2025 study conducted in Polish endoscopy centers (published in The Lancet Gastroenterology & Hepatology) revealed a troubling trend. Among physicians using AI-assisted detection during colonoscopies, the adenoma detection rate (ADR) dropped from 28.4% to 22.1% over just 3 months. Accustomed to AI highlighting problems with a "green box," they stopped actively scanning the image. Their pattern recognition capability underwent rapid atrophy.
A Dulled Creative Edge
The same mechanism is beginning to affect knowledge work. While generative AI helps weaker creators, it simultaneously leads to homogenization. An analysis of 400,000 academic papers (Science Advances, 2024) shows that content is becoming increasingly similar, correct, polished, but mediocre.
What's worse, after stopping AI use, creators' creativity levels often drop below their starting point. The razor has been dulled. Creativity requires effort, wandering, and grappling with frustration. AI eliminates this necessary resistance, offering a ready-made highway to the solution. As a result, according to Stack Overflow's 2025 survey, developers write code 55% faster, but only one in three trusts what they've generated. We're creating solutions faster than we can understand them.
A Plan for Sharpening the Razor: How to Maintain Mental Fitness
This isn't about rejecting technology—it's about consciously maintaining our cognitive capabilities. We need to treat mental fitness the way athletes treat training and pilots treat simulator sessions.
1. Individual Level
Keep a Decision Journal: Write down why you're choosing a particular solution and what alternatives you see. This protects against hindsight bias.
Verify AI outputs: Treat AI like a bright but inexperienced intern. Never copy content you couldn't write yourself or don't understand.
Schedule Deep Work sessions without AI: Set aside at least 90 minutes daily for creative work without assistants. Sketch on paper, analyze architecture on a whiteboard, and do mental math.
Teach others: Explaining concepts to a junior (instead of sending them to a bot) forces your brain to go deep into knowledge structuring.
2. Team Level
Pre-mortem Analysis: Before starting a project, assume it has failed and try to identify what led to that failure. AI is usually too programmed for optimism to spot these risks.
"Humans vs. AI" Calibration: Run competitions—how would the team solve a problem versus the machine? This builds metacognitive sensitivity and teaches when to trust technology.
Peer Review: Don't abandon human code and document reviews. Conversations about mistakes are crucial for transferring tacit knowledge that AI doesn't possess.
3. Organizational Level
Communities of Practice: Support the exchange of experiences and failure stories you won't find in official documentation. I've written about their role in my post on communities of practice in building organizational culture.
Role Rotation: Introduce changes in areas requiring experience to enforce continuous learning and prevent routine.
Measure Decision Quality: Instead of evaluating only speed and effort (where AI always wins), measure quality and implementation effectiveness over the long term.
Heart and Data
The emergence of AI forces us to redefine the human role in value creation. Leaders must actively enforce practices that allow teams to maintain their sharpness of judgment.
Human-machine collaboration is the foundation of my "Leading with Heart and Data" approach. AI may process data more efficiently, but the "heart" must remain on our side—the moral compass and the capacity for critical analysis. History shows that calculators and Excel didn't destroy engineering. The same will hold true for us, as long as we learn to use new tools without losing what makes us something more than an advanced LLM.
Wojciech Pozarzycki, January 2026