Specialised skills in the age of AI: After engineering, and later when I studied rural development management, I heard the same advice repeatedly: be adaptable, learn a little of everything and do not limit yourself to one field.
Business schools reinforce much the same idea. A good manager, we are told, should know enough about finance, marketing, operations, human resources, strategy and technology to work across all of them.
After nearly 13 years working in development, rural livelihoods, agribusiness, journalism, higher education, programme management and institution building, I have started to question that advice.
READ | Generative AI is forcing universities to rethink assessment
Breadth has value. It helps people move between disciplines, understand different perspectives and work with specialists. But there is a point at which breadth without sufficient depth becomes a liability. In a labour market where many occupations are becoming more specialised, knowing something about many subjects may not be enough.
The World Economic Forum’s Future of Jobs Report 2025, based on a survey of more than 1,000 employers representing more than 14 million workers across 55 economies, estimates that 39% of workers’ core skills will change by 2030. The report also finds that 63% of employers regard skills gaps as a major barrier to business transformation. AI and big data, networks and cybersecurity, technological literacy, creative thinking and analytical thinking are among the skills expected to grow in importance.
The change can be seen across professions. Engineering now encompasses specialised fields such as semiconductor design, robotics, renewable energy and electric mobility. Technology companies need expertise in areas ranging from artificial intelligence and cybersecurity to cloud architecture and data engineering. Finance has its own increasingly specialised fields in risk, taxation, investment and financial analytics. Agriculture and development require expertise in areas such as climate adaptation, rural finance, impact evaluation, agricultural value chains, digital agriculture and farmer-producer organisations.
The OECD’s Skills Outlook 2025 points in a similar direction. It identifies continuing changes in the skills required by specialist occupations and stresses the need for workers to keep updating their capabilities as technology and labour markets change.
The lesson I draw from this is not that young people should specialise early and ignore everything outside their chosen field. It is that depth should come before breadth.
READ | Generative AI has changed the meaning of university degree
Specialised skills: Build an area of real expertise
A useful way of thinking about this is the idea of a T-shaped professional. The vertical line represents a field in which a person has developed substantial knowledge and experience. The horizontal line represents an ability to understand adjacent areas and work with people who have different expertise.
Consider an AI professional working on machine learning. Technical knowledge alone may not be enough. The person also needs to understand the problem the technology is meant to solve, the circumstances in which it will be used and the ethical or commercial consequences of its application.
An agribusiness professional may develop deep knowledge of agricultural value chains while also understanding production, climate risks, finance, consumer behaviour and digital technologies. That breadth makes the specialist more useful. It does not substitute for the specialist knowledge itself.
My own career is perhaps a useful example. I began with engineering, moved into rural development and later worked in journalism, constituency development, agribusiness, government programmes and higher education. On paper, the progression can look less like a career plan than a collection of unrelated experiences.
It took me some time to see the connection between them. Experience across fields becomes useful when there is enough knowledge in at least one area to give it a centre of gravity. Otherwise, a CV can simply become a record of places and roles. Expertise gives those experiences a way to connect.
This matters more now because artificial intelligence is changing the value of general knowledge.
READ | Civil services reform must move beyond training
When information becomes cheap, judgement matters more
AI can already retrieve, summarise and explain information across a remarkable range of subjects. Someone who knows a little about many things therefore has less of an informational advantage than before.
But information is not the same as understanding.
The recent debate over an AI-generated proof related to the Navier–Stokes problem illustrates the distinction. OpenAI announced in September that an internal AI system had produced a proof addressing the long-standing mathematical problem. The Clay Mathematics Institute subsequently described the problem as apparently settled while noting that its formal process for evaluating such a claim still has to take place.
Even if the mathematical result survives scrutiny, another question remains: what have human mathematicians learned from the result? A machine can arrive at a solution without necessarily providing the kind of understanding that allows researchers to explain why it works, identify the ideas behind it or use those ideas to tackle another problem.
That is not a problem confined to mathematics.
An AI system can produce a financial analysis, summarise legal material, suggest an engineering solution or generate an agricultural recommendation. A professional still has to decide whether the assumptions are reasonable, whether important information is missing and whether the answer makes sense in the circumstances at hand.
That requires knowledge of the field. It also requires experience.
A rural-finance specialist, for instance, will be better placed to assess an AI-generated recommendation if the person understands how farmers borrow, how agricultural incomes behave and how local institutions actually work. A marketing professional who understands consumer behaviour can question an AI-generated campaign in ways that someone relying only on the machine’s output may not.
The advantage is therefore shifting. Knowing where to find information is useful, but increasingly common. Knowing what information matters, how it fits together and when an apparently plausible answer is wrong is harder.
Breadth still matters, but it needs an anchor
This is why specialisation should not be confused with narrowness.
Someone specialising in rural finance should know enough about digital finance and farmer institutions to understand how the field is changing. A marketing professional should understand data and technology. An engineer should know something about business and sustainability. A development professional needs familiarity with quantitative methods, institutions and public policy.
The point is not to accumulate a long list of secondary skills. It is to develop enough knowledge outside one’s main field to ask better questions and work effectively with people who possess different expertise.
This has implications for education as well. Students should have room to explore before committing themselves to a field. Early experimentation can reveal interests that are difficult to identify from a classroom alone. But exploration cannot remain the permanent strategy. At some stage, a person has to spend enough time with one subject to move beyond familiarity and become genuinely competent.
That takes sustained effort. It also takes patience, something career advice often understates.
The pressure to remain adaptable can sometimes encourage people to keep moving: another qualification, another role, another sector, another skill. Movement can be useful, but it does not automatically produce expertise. Spending years around a subject is not the same as understanding it deeply, either. Depth comes from sustained engagement with difficult problems, repeated application and learning from mistakes.
AI will make this distinction more important. As machines become better at producing competent first answers, the ability to evaluate those answers will matter more. The person who understands the underlying field can use AI as a tool without becoming dependent on its apparent confidence.
For a young professional, then, the more useful question may not be how many skills to acquire or how many fields to sample.
It may be this: What problem do I want to become exceptionally good at solving?
Finding an answer to that question does not require giving up breadth. It gives breadth a purpose.
Krishna Uppuluri is a development professional, educator, and research scholar whose work explores the intersection of public policy, governance, rural development, and institution building.