A Jack Of All Trades? Why AI May Rescue Us From Early Specialisation

“A jack of all trades is a master of none.” Most of us remember this familiar saying. But fewer remember the words often added to it: “But oftentimes better than a master of one.”

Whether or not those words belonged to the original proverb, they capture a fundamental weakness in how we educate our children. From middle school onwards, we encourage students to specialise in mathematics, science, economics, or the arts, often before they have discovered their interests or understood how different disciplines connect. By university, these choices have hardened into professional identities.

Specialisation has produced extraordinary advances in human knowledge. But it has also encouraged intellectual tunnel vision. Doctors approach problems through medicine, economists through incentives and data, and engineers through technical solutions. Yet most important problems require knowledge from several disciplines. The danger is not simply that specialists lack knowledge outside their fields, but that they may fail to recognise what they are missing.

Artificial intelligence may fundamentally change this equation. For the first time, we can combine deep professional expertise with broad access to complementary knowledge at remarkably low cost. AI offers us the possibility of becoming masters of one discipline without remaining strangers to all the others.

The problem with teaching children to think in silos

Consider a doctor treating patients with a complex disease. She understands the biology of the disease, the available treatments, and the clinical characteristics of her patients. Suppose she observes that patients receiving a particular treatment recover more quickly than others. Does this mean the treatment works? Or were healthier patients more likely to receive it?

An economist trained in causal inference would ask whether treated patients were systematically different from untreated patients. Perhaps doctors prescribed it to patients with better prognoses. Without addressing these questions, the observed relationship may not establish that treatment caused improvement.

Medical researchers and biostatisticians are familiar with these problems. Randomised clinical trials exist precisely because medicine recognises causal identification. But not every excellent clinician is familiar with the full range of statistical and econometric methods for observational data.

Now reverse the example. An economist studying the effect of extreme heat on worker productivity may establish a convincing causal relationship. But why does productivity fall? Dehydration? Cognitive impairment? Sleep disruption? Cardiovascular stress? Changes in effort?

The economist may have the statistical tools to establish the effect but lack the biological knowledge to understand its mechanisms. Different mechanisms imply different solutions: water access, night-time cooling, or adjustments to work schedules.

The danger of specialisation is not simply that we know too little about other fields. It is that our training influences which questions we think to ask in the first place.

Why don’t we simply collaborate?

Specialists can and do consult other specialists. Doctors collaborate with statisticians. Economists work with climate scientists. Engineers consult psychologists. Universities increasingly establish interdisciplinary institutes, offer collaborative grants, and encourage researchers to work across departments.

Yet genuine collaboration remains difficult. Researchers’ networks are concentrated in their own fields. Finding the right expert takes time. Understanding unfamiliar terminology requires effort. Explaining a problem to someone trained in another discipline can be surprisingly hard.

The barriers are also human. A distinguished professor may hesitate to ask a junior researcher in another field to explain basic concepts. An accomplished doctor may be reluctant to admit unfamiliarity with an econometric method. Professional ego, status, and fear of appearing uninformed can discourage useful exchanges.

Institutional incentives also matter. Academic careers are often evaluated within disciplines. Departments control hiring and promotion. Researchers worry about authorship, credit, and whether interdisciplinary work will be valued by those assessing their careers.

Similar frictions shape broader social relationships. Shared language, culture, education, and professional experience can make interaction easier, although genuine preferences and values matter too. We tend to work with people we understand, limiting the ideas we encounter.

AI changes the economics of accessing knowledge

For the first time, a specialist can obtain substantial help with unfamiliar disciplines at an extremely low marginal cost. A doctor can ask AI to explain identification strategies, evaluate assumptions, and suggest statistical tests. An economist can explore physiological mechanisms behind heat exposure. An engineer can investigate behavioural economics without first earning another degree.

Unlike approaching a senior colleague, asking AI carries no embarrassment. Researchers can ask elementary questions, seek clarification, challenge responses, and repeat until a concept becomes understandable.

AI does not eliminate the need for human experts. Its answers can be wrong, and consequential conclusions require verification. But it sharply reduces the initial cost of exploring unfamiliar knowledge.

Because AI is a general-purpose technology, the same interface can help a doctor understand statistics, an economist understand biology, and a lawyer understand financial incentives. The marginal cost of asking another question is often close to zero, even when the answer draws on another discipline.

From understanding another discipline to using it

The next development is even more consequential. AI is moving beyond information towards implementation. An assistant can explain causal inference; an agent with approved access to data and tools can help organise patient records, write code, run diagnostics, and identify weaknesses. Clinical judgement and methodological validation remain indispensable.

AlphaFold illustrates how advanced computational methods can become accessible to molecular biologists who did not build the underlying models. Its protein-structure predictions support questions that once demanded more specialised computational resources, although experimental validation remains important.

Coscientist, an AI system described in Nature, combined scientific information retrieval, programming, laboratory documentation, and robotic equipment to plan and execute selected chemical experiments. Its demonstrations were bounded and supervised, but they show how AI can coordinate capabilities that traditionally reside in separate disciplines.

The promise is not that every doctor becomes an econometrician or every economist becomes a biologist. It is that professionals can draw more effectively on knowledge beyond their original training.

Some disciplines are more portable than others

Not all specialised knowledge is equally transferable. Economics, mathematics, statistics, and computer science contain methods applicable across many fields. Economics studies incentives, constraints, information, strategic interaction, and causal relationships—concepts relevant to medicine, education, law, engineering, and environmental policy.

Yet applying a portable method correctly still requires domain knowledge. Economists cannot infer a drug’s biological effects from economic theory alone. Conversely, doctors and engineers can benefit from economic reasoning, statistical methods, and behavioural science. AI can make these complementary capabilities more accessible, increasing the productivity of expertise people already possess.

Professionals who fail to use AI to complement their expertise risk making themselves increasingly replaceable. Using AI merely to perform existing tasks faster misses its greatest potential. A doctor can draw on economics and statistics, an economist on biology and psychology, and a lawyer on the economic and technological implications of legal decisions. The real opportunity is not to use AI to do what we already know, but to do what our specialised training previously prevented us from doing. This is how AI can make professionals more valuable rather than more replaceable.

What should our schools do differently?

If AI makes complementary knowledge easier to access, we should reconsider how we educate children. Too often, students choose an intellectual direction before they understand the alternatives. Years of specialised training reinforce those early choices and can create intellectual rigidity.

The answer is not to eliminate specialisation. We still need excellent doctors, economists, engineers, scientists, and lawyers. But we should delay unnecessarily restrictive choices and provide broader exposure to different ways of thinking.

A future doctor should understand why incentives and causal inference matter. An economist should appreciate biology and psychology. An engineer should recognise that technical solutions succeed or fail partly because of human behaviour.

Students should also learn to use AI to explore unfamiliar disciplines, challenge assumptions, and identify missing knowledge. They must learn to evaluate AI critically: fluency is not accuracy, and access to information is not understanding. The goal is deep expertise without intellectual confinement.

The next productivity revolution

For centuries, economic progress has depended on specialisation. But specialisation also creates intellectual boundaries, reinforced by our education system and professional institutions.

The consequences are not simply gaps in knowledge. People may overlook alternative explanations, miss important mechanisms, or approach complex problems through unnecessarily narrow frameworks.

Collaboration has always offered a remedy, but it requires time, money, networks, and willingness to acknowledge what we do not know. AI changes this equation by making unfamiliar knowledge more accessible and increasingly helping specialists implement methods outside their training.

The gains could extend well beyond doing existing tasks faster. AI may enable better questions, fewer mistakes, new connections, and solutions to problems that previously required expertise people could not readily access.

For generations, we have educated people to become specialists and then expected them to solve problems requiring knowledge from many disciplines. AI may finally help us overcome this contradiction.

The next productivity revolution may not come from replacing human expertise. It may come from freeing human expertise from the intellectual boundaries our education system helped create.