Your AI Is Getting Smarter. Your Marketing Instincts Are Getting Weaker.

A group of experienced doctors in Poland just proved something that should concern every business owner who has been handing their thinking to AI.

These were gastroenterologists. Each had performed over 2,000 colonoscopies. When their clinics introduced AI-assisted polyp detection, something unexpected happened. Not when they used the AI. When they stopped using it.

Their detection rate for precancerous growths dropped from 28.4% to 22.4%00133-5/abstract). A 20% relative decline. In experienced professionals. After just months of AI exposure. The study, published in The Lancet Gastroenterology & Hepatology in 2025, represents the first clinical evidence of what researchers call automation-induced deskilling with measurable patient outcomes.

The skill didn't vanish because they stopped practising. They were still performing colonoscopies every day. The skill eroded because the AI had been doing the noticing for them. Their pattern recognition, the subconscious ability to spot something wrong, had quietly atrophied.

If you're using AI to analyse your campaigns, write your ads, draft your strategy, and evaluate your results, the same process is happening to your marketing judgment right now.

How Expert Intuition Actually Works (And What AI Breaks)

What makes an expert an expert? This sounds like a philosophical question. It's not. It has a precise, research-backed answer.

Daniel Kahneman, the Nobel Prize-winning psychologist behind Thinking, Fast and Slow, and Gary Klein, the pioneer of naturalistic decision-making research, spent years arguing about whether expert intuition could be trusted. They eventually reached a joint conclusion, published in American Psychologist, that identified exactly two conditions required for reliable expert intuition.

First, the environment must provide high-validity cues. The patterns between what you observe and what happens next must be stable and learnable. Second, the expert must have prolonged opportunity to practise with rapid, unambiguous feedback.

Marketing qualifies on both counts. Search terms that convert follow patterns. Audiences that respond share characteristics. Landing pages that work have recognisable structure. These patterns are stable enough to learn. But they can only be learned by doing the analysis yourself: reading the search terms, studying the conversion data, testing the landing pages, watching what works and what doesn't over months and years.

This aligns with what's known as the Dreyfus model of skill acquisition, which describes five stages of expertise development: Novice, Advanced Beginner, Competent, Proficient, and Expert. At the Expert level, practitioners no longer follow rules. They operate on intuition built from thousands of hours of accumulated pattern recognition. As Google Ads specialist Mike Rhodes describes it, "that last 10% is all about intuition. It's the tacit knowledge. You cannot put it into a flowchart."

Here is the critical problem: AI lets you skip the middle stages of expertise development. And those middle stages are where intuition gets built. You jump from Novice straight to apparently competent output, without ever developing the pattern library that would let you evaluate whether that output is actually good.

The Jagged Frontier: Where Confidence Meets Incompetence

A Harvard Business School study with 758 BCG consultants, published in Organization Science in early 2026, put hard numbers on this phenomenon.

When consultants used GPT-4 for tasks within AI's capabilities, they completed 12.2% more tasks, 25.1% faster, with 40% higher quality results. Impressive.

But when they used AI for tasks beyond what the researchers called the "frontier" of AI's reliable capabilities, they were 19% less likely to produce correct solutions than consultants working without AI at all.

The researchers called this the jagged frontier because the boundary isn't smooth or predictable. You can't easily tell which tasks AI will handle brilliantly and which it will get wrong. And the consultants couldn't tell the difference either. They trusted AI's confident-sounding wrong answers just as readily as its right ones.

This is what the research calls automation bias: the tendency to over-trust automated recommendations, even when they contradict your own judgment. It's the same phenomenon that caused doctors to accept AI's polyp assessments without developing their own diagnostic instincts.

For marketing, the jagged frontier runs straight through the decisions that matter most:

We've written before about why AI amplifies strategy rather than replacing it. The deskilling research explains the mechanism: when you stop exercising a cognitive skill, the neural pathways that support it weaken. It's not a metaphor. It's neuroscience.

Why This Matters More for Small Businesses Than Anyone Else

Rory Sutherland, Vice Chairman of Ogilvy, argues that competitive advantage lives in the counterintuitive. His central thesis from Alchemy: "It is much easier to be fired for being illogical than it is for being unimaginative. The fatal issue is that logic always gets you to exactly the same place as your competitors."

His eighth rule of alchemy is direct: "Test counterintuitive things only because no one else will." That's where the breakthroughs hide.

But here's the catch nobody talks about. To identify what's counterintuitive, you first need strong intuition about what's conventional. If AI has been doing your marketing thinking, you don't have that baseline. You can't break rules you never internalised.

This connects to Byron Sharp's research on how brands grow. Sharp's data across 13 product categories and 130+ brands shows that growth overwhelmingly comes from reaching new buyers. Not deeper loyalty from existing ones. The insight that a particular trade show audience, a specific community group, or a seasonal shift in your local area could become your next growth channel isn't something that appears in a dashboard. It comes from years of accumulated understanding of your customers, your competitors, and your market.

For an SME spending $3,000 to $5,000 a month on Google Ads, your marketing intuition is your only real competitive advantage. A multinational competitor can outspend you. They can afford better tools. What they can't replicate is your 15 years of knowing that tradies in your region respond to different messaging than office workers, or that enquiries spike after school holidays for reasons no seasonality table would predict. This is the kind of knowledge that feeds what we've described as the judgment bottleneck in AI marketing: AI can generate a thousand options, but only human expertise can pick the right one.

When every business in your category is feeding the same data into the same AI tools and getting the same recommendations, the one who wins is the one whose human judgment spots what the algorithm missed.

The Compounding Problem: How Deskilling Hides Until It's Too Late

A case study from Aalto University in Finland reveals how deskilling compounds silently. Researchers studied an accounting firm whose automation system failed unexpectedly. Staff couldn't perform essential accounting tasks they had routinely done before the system was introduced.

As researcher Esko Penttinen observed: "The system was in a way 'too perfect.'" So perfect that nobody noticed skills were disappearing until the technology wasn't there. His colleague Joona Ruissalo added a line that should be framed on every marketer's wall: "Higher thoughts cannot be had without engaging in detail."

The same dynamic appears in the medical research. The Lancet study noted that junior doctors who trained alongside AI may never develop the diagnostic instincts their predecessors built through years of unassisted practice. The researchers described them as "never skilled": not deskilled, because the skill was never acquired in the first place.

In marketing, the equivalent is the account manager who has always had AI generate their search term analysis, write their ad copy, and recommend their bidding strategies. Their output looks polished. Their reports are formatted beautifully. But if you asked them to look at a raw data export and tell you what's wrong with a campaign, they would struggle. Not because they're unintelligent. Because they never built the pattern library that Kahneman's research shows is essential for expert judgment.

This is the real risk that sits underneath the conversations about whether AI makes marketing faster but not better. It does make it faster. But if "faster" means skipping the learning that would make you better, you're trading long-term capability for short-term efficiency. And the trade gets worse over time, because the skills you're not building today are the ones you'll need to evaluate AI's recommendations tomorrow.

What to Actually Do About This

The answer isn't to stop using AI. That would be like refusing to use a calculator because you want to practise long division. The answer is to be deliberate about which tasks you keep and which you automate.

Nita Farahany, professor of law and philosophy at Duke University, frames this as preserving your "constitutive competency": the core thinking capacity that makes you effective in your domain. Her advice to anyone using AI tools: figure out where your generative thinking happens, and protect that above all else.

For marketing, constitutive competency means:

What This Means for Your Business

AI tools are getting better every quarter. That is not the risk. The risk is that your judgment is getting weaker at the same rate, and you won't notice until a campaign fails and you can't diagnose why.

The businesses that will thrive aren't the ones using the most AI. They're the ones using AI for the right things: execution, scale, speed, data processing. While deliberately preserving the human judgment that no algorithm can replicate.

Your decade of knowing your customers, your market, and your industry is not a quaint relic of the pre-AI era. It is your most valuable competitive asset. The Lancet study proved that even highly trained professionals lose their edge when they stop practising their core skills. Your marketing instincts work the same way.

Use AI to do more. But never stop doing the thinking yourself.

Further Reading

• Endoscopist deskilling risk after exposure to AI in colonoscopy00133-5/abstract) - The Lancet study proving AI dependency erodes professional diagnostic skills
• Navigating the Jagged Technological Frontier - Harvard/BCG study on AI's uneven impact on knowledge worker performance
• The Vicious Circles of Skill Erosion - Aalto University research on how automation failure reveals hidden skill loss
• Conditions for Intuitive Expertise: A Failure to Disagree - Kahneman and Klein's joint paper on when expert intuition can be trusted
• Rory Sutherland, Alchemy - The case for testing counterintuitive ideas as a competitive strategy