Last week, I came across new research on cognitive atrophy from overuse of automation. The findings were striking. When automation carries everything, core skills can and will fade. Like a muscle, if you don't use it, you lose it.
That research caught my attention, so I set a personal challenge for myself. A friend was flying in, and I decided to drive to the airport to pick her up without GPS. No blue dot. No voice prompts.
On the highway, my calm faded with every exit sign. A small surge of worry, then the next. Is my exit ahead? Did I pass it? Those feelings mattered. In that moment, I realized my navigation skills had gone soft. My little experiment validated what the research warned about. Overreliance can dull fundamentals.
After confirming the atrophy of my personal navigation muscle, I decided I had to be proactive to maintain a healthy balance. In Waze, my GPS go-to, I found a setting for minimal prompts: fewer instructions, just key cues at the right moments. This new setting helps me stay both alert and calm. I am not outsourcing the whole drive, and I am not white-knuckling it either.
As I continued to think about achieving a healthy balance with AI, it reminded me of lessons I learned growing up. When my mother thought I was working too much, she would say in Spanish,
Ni tanto que queme al santo, ni tanto que no lo alumbre. Not so much that it burns the saint, not so little that it does not light him.
I once asked her what that meant. She reminded me of all the times she asked me to light the candle at our home altar. And she would always instruct me to place it close enough to illuminate the saint, but not so close that it would burn or damage the statue. It was a delicate balance.
That picture of balance is how we should approach AI. Use automation just enough to enhance and amplify your capabilities, expanding the value you bring while keeping your skills sharp and your judgment in the loop.
Using AI is not an all-or-nothing proposition. Like placing a candle near a saint, you still use the flame, but at a safe distance so it illuminates without burning.
Use AI to illuminate your work, not scorch it.
THIS WEEK'S INSIGHTS
The risk is not the tool. The risk is what happens to your brain when you stop doing the work yourself.
1. Half of all companies will soon test whether their people can think without AI. Gartner is calling it, and the implications reach far beyond hiring.
2. Neuroscience confirms "use it or lose it" applies to your thinking, not just your body. The same principle that weakens unused muscles weakens unused thinking patterns. When AI does the cognitive work, your brain quietly lets go of the wiring it no longer needs.
3. The cognitive offloading trap is real, and it lingers after you stop. A controlled study found a 47% drop in neural engagement among AI users. The worst part: when they stopped using AI, the dip did not bounce back. It stuck.
TRENDS
The evidence is building fast. Here is what the latest research tells us about skill atrophy and AI dependence.
The Gartner "AI free" mandate is here. Gartner's Top Strategic Predictions for 2026 warn that erosion of critical thinking from GenAI will push 50% of global organizations to require skills evaluations performed without AI tools. In high-stakes industries like finance, healthcare, and law, the scarcity of independent thinkers will raise talent costs and force entirely new hiring strategies. Specialized testing methods designed to isolate human reasoning ability are already emerging as a secondary market.
Think of it like a company requiring employees to parallel park before they can use the self-parking feature. They want to know: Can you still drive without the assist?
Cognitive offloading is measurable, and the numbers are sobering. A study published in Cognitive Research: Principles and Implications found that AI assistants may accelerate skill decay among experts and hinder skill acquisition among learners. Even more concerning, the researchers found that AI may also prevent both groups from recognizing these effects. You lose capabilities you don't realize you're losing. A separate study of 666 participants across diverse age groups confirmed a significant negative correlation between frequent use of AI tools and critical thinking abilities. Younger participants showed the highest dependence and the lowest scores.
"Cognitive debt" is the new term leaders need to know. In a March 2026 article in Cognitive World, researcher Mohammad Hossein Jarrahi introduces the concept of cognitive debt: the hidden cost of frictionless AI. As autonomous systems shift work from thinking-by-doing to choosing among outputs, human expertise atrophies while outputs appear to improve in the short run. He also flags AI sycophancy, where AI systems are optimized to agree with you rather than challenge you, as an accelerator of confirmation bias and further skill erosion.
The medical field is sounding the alarm. A 2026 paper in Frontiers in Medicine coined the term "diagnostic deskilling" to describe the phenomenon of clinicians becoming overly dependent on AI models. They rely less on their own skills, assume the AI is always more accurate, and become less confident in making independent decisions. The researchers also identify "moral deskilling," the decline in ethical sensitivity and moral judgment that comes from placing too much trust in technology. This pattern is not limited to medicine. It applies anywhere humans are quietly stepping back from judgment calls.
3 MYTHS TO REFRAME
These are the beliefs I hear most often from smart, capable professionals who are using AI but not thinking about what it is quietly doing to their sharpest skills.
Myth #1: "The more I use AI, the more advanced I am."
Why we believe it: Volume feels like progress. If I am using AI for everything, I must be ahead.
Reframe: More usage appears to be progress, but quality declines when critical thinking fades. A study of 666 participants across age groups and educational backgrounds found a significant negative correlation between frequent AI use and critical thinking ability. The driver was cognitive offloading. The more people handed their thinking to AI, the less sharp their independent reasoning became.
What to do: Design prompts that force engagement. Ask the model to list its assumptions, show each step, and cite sources. Add one quick human check for logic and context fit before anything ships.
Myth #2: "Manual practice is a step backward."
Why we believe it: Polished AI outputs make manual work feel unnecessary. Why would I go slow when the tool goes fast?
Reframe: Short manual reps keep core skills and confidence ready for the moments when AI cannot help you. Researcher Jarrahi calls the accumulation of unchecked AI dependence "cognitive debt," where frictionless AI quietly hollows out the very skills that make your judgment valuable. As people stop engaging with the execution of tasks, their cognitive muscles decline, and that decline directly undermines their ability to choose well.
What to do: Schedule brief manual reps each week. Rebuild one deliverable without AI. Compare to the AI version. Update your checklist of steps only you will verify.
Myth #3: "AI should just work. If it doesn't, it's not ready."
Why we believe it: Every tool promises to be plug-and-play. We expect AI to be the same.
Reframe: The best collaborators in your career did not show up on day one already knowing your communication style, your priorities, and your standards. You built that relationship over time. AI is the same. The difference between a bad AI experience and a great one is not the tool. It is whether you set it up to know who you are before it starts working. Context files, rules, and model selection. When the AI already knows how you think, the prompt barely matters.
What to do: Build a simple context file with three things: who you are, how you think, and what you need. Teach the tool before you test the tool.
TOOLS
This week, try the following tools and Prompts to Steal to help you take action. These will help you stay sharp while still moving fast.
NotebookLM as a research partner
Prompts to steal:
"Upload these three documents. Summarize the key findings across all of them. Flag any contradictions between sources. Do not use information outside of these documents."
Power tip: Use NotebookLM when you want to stay grounded in your own sources rather than relying on the AI's general training data. It only responds based on what you upload, which keeps you in the driver's seat and forces you to curate the inputs.
Claude as a thinking partner
Prompts to steal:
"I want to [TASK]. Read all files first. Ask me questions before you execute. Do not guess. After you finish, list the assumptions you made and tell me where your reasoning could be wrong."
Power tip: Turn on Extended Thinking so Claude reasons through the problem before responding. The difference between a quick answer and a thought-through answer is often a matter of setting.
Perplexity as a fact checker
Prompts to steal:
"Show me the top studies on [topic]. Extract findings, methods, sample sizes, and limitations in a table. Flag weak designs. Tell me what new evidence would change the conclusion."
Power tip: Use Perplexity when you need cited, verifiable sources fast. Every sentence includes a citation, making it easy to check the AI's work rather than just trusting it.
TRY IT THIS WEEK (Micro-Actions)
These are designed to take you from knowing to doing. Share these with the leaders in your circle.
1. Go manual on one task this week. Pick something you normally hand to AI entirely: drafting an email, summarizing a report, or prepping for a meeting. Do it without AI first. Then do it with AI. Compare the two. Notice where your instincts were sharp and where they had gone quiet.
2. Write a visible guardrail. Before your next AI session, write one sentence at the top of your prompt: "AI can start these steps. I will check accuracy, logic, sources, and final tone." That one line changes your posture from passenger to driver.
3. Ask AI to show its work. On your next real task, add this instruction: "List each step you took. Note assumptions. Point to the source lines you relied on." Compare its process to your own manual flow. If they match, delegate more. If they drift, bring the work back to you.
POWER TIP
Start with thinking, not answers. Ask the tool to outline a plan, list assumptions, show each step, cite source lines, and rate confidence. Save that prompt as a template. Compare the AI's steps to your manual checklist. If they match, delegate more. If they drift, bring the work back to you.
👉 What's one boundary you will set so AI keeps you sharp, not rusty?
Closing Thought
Use AI for speed. You set the standard.
Leaders who stay in the loop keep their edge!
♻️ Share this with a leader who needs to hear this.
¡Hasta la próxima, un abrazo fuerte! (Until next week, a big hug!)
Monica Marquez
🔔 Follow me for more on LinkedIn
