The paradox of expertise: demand rises as learning pathways narrow

Why reading a situation is becoming more valuable

A client describes a problem, and an experienced adviser hears the assumption underneath it. A doctor notices that the numbers fit while the patient's condition tells a different story. A tradesperson sees a previous repair and anticipates what they'll find behind the wall.

The pattern appears in relation to a particular situation. That is what makes expertise relational. It becomes visible through the question that reframes the problem, the warning raised before a decision or the option chosen because a similar case ended badly.

I've been writing about this subject for years. My position comes from research I've read and conversations with guests on the Wisepreneurs Podcast, including people who study business performance, workplace learning and AI adoption.

In 2023, I wrote about institutional failure and groupthink. Those concerns still hold. The research I've worked through since then gives me a clearer account of why an experienced professional's reading of a situation matters.

Facts, instructions and standard methods can be written down and copied cheaply. AI handles that kind of knowing well. Knowing what tends to happen when a client delays a decision, how a team reacts under pressure or which early warning deserves attention comes through having lived the situation.

Cedric Chin writes the business-expertise site Commoncog and joined me on the Wisepreneurs Podcast. He has spent years studying how people become capable in difficult fields.

When you ask an expert how they knew what to do, he says, they will say, "it just felt right. And that's the most you can do."

That felt sense draws on many encounters with similar situations. The person recognises a pattern before they can fully explain each step.

The cognitive scientist Keith Stanovich offers another way to understand what is happening. His work distinguishes between three parts of thinking: the quick response that registers something unusual, the processing used to work through information and the disposition to question an easy answer.

AI can process vast amounts of information. The experienced professional contributes something different: a memory of how this kind of client responded last time, a sense that one option carries an overlooked risk and the discipline to examine the answer again.

When generating an answer costs almost nothing, the constraint shifts to deciding which answers deserve action.

Organisations therefore need people who can compare an AI-generated recommendation with the circumstances in front of them, test its assumptions and take responsibility for the choice. The judgement involved in choosing that option is becoming more valuable.

Why the pipeline that produces expertise is narrowing

A junior lawyer learns by preparing drafts, receiving corrections and seeing which argument a partner keeps. An apprentice learns by attempting the task, making mistakes and working beside someone who recognises trouble early. A new consultant learns by sitting in meetings, preparing analysis and watching how the client responds.

Those activities create the patterns that later appear as expertise.

When organisations automate entry-level work, they remove part of this learning process. A beginner may produce acceptable work sooner, yet they receive fewer chances to struggle with the problem, make a decision and compare the result with what happened next.

Connie Malamed, a learning designer I interviewed for the podcast, has spent her career trying to capture knowledge learned through practice.

She explained that the hard part is drawing this knowledge out of the person who has developed it. A doctor may recognise that a patient is receiving too little oxygen without having previously described every cue involved. Someone has to keep asking what the doctor noticed, which detail mattered and what led to the next decision.

This transfer depends on close contact between experienced practitioners and people still learning.

I've seen that process in my own family. Two of my sons are tradies, an electrician and a carpenter. Their ability has grown through years on the job. They've attempted tasks, dealt with unexpected problems and learned when to ring a mate who has encountered something similar.

That pathway takes time, repeated practice and other people.

Cedric Chin makes the same point from the research. The Chinese business operators he admired "took a lifetime to get good," he says.

Deliberate practice is often presented as the standard route to faster skill development. In business, Cedric says, it is "impossible to apply to business" because the necessary coaching, repeated drills and clear feedback have rarely been built.

Developing someone who can read a complicated business situation has always required considerable time and attention.

AI adds pressure to that slow pathway. Filip Drimalka, who advises companies on digital change and joined me on the podcast, identified the risk directly. Junior employees, he said, "have much less opportunities to learn and gain this experience because managers are turning to AI instead of junior people."

In our conversation, I put the practical consequence back to him: organisations may replace interns and junior staff with a keyboard and AI. Fewer people then complete the early work that teaches them how to assess a result for themselves.

Michael Caosun and Sinan Aral at MIT model how this can play out. Their research suggests that heavy AI use before a professional reaches a sufficient level of skill can close off later development, even when the same technology helps an experienced colleague.

Medical residencies, law firm associate programs and engineering apprenticeships all rely on sustained practice during the formative years. Each task gives the learner another result to compare with their original reading of the situation.

Automating those tasks reduces the number of comparisons they experience.

This creates structural scarcity. Experienced professionals eventually leave their fields while fewer beginners receive the repeated, hands-on learning needed to replace them.

The augmentation trap

A cancer specialist reviews an AI recommendation, accepts it and moves to the next case. Repeated over months, the specialist gets less practice examining the evidence and forming an independent view.

Caosun and Aral describe a year-long study in which cancer specialists experienced what the researchers call intuition rust. Sustained reliance on AI decision support gradually dulled their ability to assess cases independently.

The early productivity gain carried a longer-term cost. As Caosun and Aral put it, using AI to bypass the reasoning "does not augment the worker's intelligence, it automates its decline."

This is the augmentation trap.

The safer approach is to form your own view before asking AI. Write down what you think the problem is, which option you favour and what evidence could change your mind. Then use AI to challenge, extend or test that view.

Treat its output as material requiring examination. Check the assumptions, trace the sources and compare the answer with what you know about this client and situation.

Protect the tasks that keep your abilities active. Working through a difficult brief, structuring an argument or reviewing an uncertain recommendation may take time. These tasks also give you the repetitions that keep your pattern recognition sharp.

The professionals I've interviewed who use AI well keep themselves inside the process.

Filip Drimalka describes it this way: "You use AI within this process, but AI doesn't do the whole work," he said. "You're the one who is orchestrating."

His summary is: "it's about using AI for your own expertise."

Here, that expertise appears in the prompts he chooses, the output he rejects and the final decision he remains responsible for.

Connie Malamed reaches a similar conclusion from workplace learning. She uses AI tools "as a starting point and not as the end point," then reshapes the material for the learners, purpose and voice involved.

Independent professionals have an advantage because they can decide where AI enters their work. They can use it for transcription, initial research or alternative options while keeping diagnosis, interpretation and final recommendations under their own control.

Caosun and Aral's model suggests that people who control their own AI use are better able to protect future capability. A manager focused on this quarter's output may set a higher level of automation than the worker would choose for their longer career.

When you decide which tasks to keep, self-direction becomes a form of protection.

Auditing what you know

A method may have worked with five previous clients because they shared similar conditions. An industry rule may have survived through repetition without anyone checking the evidence. A memorable success may hide the many similar attempts that failed.

These are the assumptions that can become attached to genuine expertise.

The responsible response is to audit what you think you know. Ask where the belief came from, when it last held true and what evidence would lead you to revise it.

Cedric Chin found this quality among the capable operators he studied. He calls it cognitive agility and describes it as being "able to change their minds when new data, new evidence presents itself."

In business, he says, "you cannot lie to yourself."

That willingness to update separates a capable practitioner from someone relying on seniority. It requires the person to revisit a familiar answer, consider how the last comparable decision played out and accept evidence that points elsewhere.

What this means for your practice

The paradox of expertise is a market condition. AI makes answers abundant while increasing the value of people who can decide which answer fits the situation.

Three practical consequences follow.

First, your ability to interpret is becoming more valuable. An organisation can generate a strategy, report or set of recommendations quickly. It still needs someone to find the weak assumption, compare the options and explain which course deserves action.

This is where expertise appears in your practice: through the client's problem, the questions you ask and the choice you help them make.

Second, independence gives you control over how you think. You can decide which tasks AI handles and which ones remain part of your own analysis. That choice helps preserve the repeated practice behind reliable decisions.

Employees can seek similar control, although task allocation often sits with a manager whose priorities may focus on immediate output.

Third, credibility depends on regular review. Return to your assumptions, test old methods against current conditions and ask what recent results have taught you. The professional who updates a view when the evidence changes is more useful than the person who repeats a familiar answer.

This work also benefits from another person's perspective.

Almost every successful independent professional I've interviewed has received help with their thinking. A coach, mentor or partner can ask the question that's hard to see when you're close to the problem.

Michael Zipursky says his firm has invested heavily in coaches and programs because there is always more to learn. He also says many of his clients already know how to solve the technical or professional problems in their field.

As he puts it, "they already have that expertise."

In practice, that means they can diagnose the operational problem, advise the executive team or complete the specialist work clients bring them.

Their challenge is to "communicate" their value, then "package it and position it and place value on it."

After a few successful engagements, the realisation becomes clearer: "people want my expertise."

They want the questions that clarify the brief, the patterns recognised from similar cases and the recommendation shaped for their circumstances. The next task is building a practice that delivers those outcomes and supports the life you want.

The wider argument sits inside the Relational Vitality writing on cognitive vitality and what protects it.

Frequently asked questions

References

Cedric Chin on accelerating business expertise.

Wisepreneurs Project podcast, on expertise acceleration, tacit knowledge and cognitive agility.

Connie Malamed on mastering the art of instructional design.

Wisepreneurs Project podcast, on tacit knowledge and using AI as a starting point.

Filip Drimalka on thriving in the age of AI.

Wisepreneurs Project podcast, on orchestrating AI within the process and the narrowing pathway for junior staff.

Michael Zipursky Consulting Success: Why Conversations Beat Marketing for Independent Consultants.

Wisepreneurs Project podcast, on AI amplifying rather than replacing expert advisers.

Keith Stanovich, What Intelligence Tests Miss: The Psychology of Rational Thought (2009). The basis for the distinction between tested working knowledge and untested assumptions, and for the three levels of thinking.

Michael Caosun and Sinan Aral, The Augmentation Trap: AI Productivity and the Cost of Cognitive Offloading (2026). Research showing how rational AI adoption can erode the expertise it depends on, the intuition rust finding, and the structural protection self-directed professionals hold

Every Tuesday I send The Wisepreneur, a short letter for experienced professionals building independent practice: one usable idea a week. Sign up at wisepreneurs.com.au/newsletter

When you want a second pair of eyes on your own practice, the Positioning and Practice Audit is how I work with people: wisepreneurs.com.au/marketing-partnership