The disruption unfolding across today's labour market is unlike anything that came before. Where past waves of automation swept through factory floors and manual work, AI is hitting white-collar jobs, especially entry-level ones. In the first quarter of 2026, tech companies laid off more than 78,000 workers, with 48% attributed to AI automation. Even Jerome Powell, the United States Federal Reserve chair, has warned that AI could “absolutely have implications for job creation”.

How can we gird ourselves for AI’s impact? In our “AI & Jobs” series, we ask INSEAD professors to analyse the situation from the perspectives of individuals – the focus of this article – as well as organisations and policymakers. 

The consensus: AI is redesigning and restructuring jobs far more than it’s making them obsolete. Whether and how individuals exploit AI while honing their own skills (especially for entry-level and junior workers) will decide the security of their future. 

Meta-skills will be a game changer 

Phanish Puranam, The Roland Berger Chaired Professor of Strategy and Organisation Design 

The jobs most vulnerable to AI displacements in the next few years will be those that are low-level, don’t require collaboration (e.g. modular work) or in-person presence (e.g. purely knowledge work), and are done the same way across companies. This isn’t necessarily because algorithms will become effective substitutes for all tasks in all such roles – the evidence suggests they aren’t (yet) – but because organisations are no longer hiring as they expect AI to eventually catch up. 

That said, completely new tasks and roles have been created since the advent of generative AI. I categorise them into four types: AI operations, AI compliance, jobs related to human-AI interaction (e.g. prompt librarian, AI personality coaches), and perhaps the biggest group is simply AI-augmented versions of old roles in software, medical services and other sectors where demand is elastic.

Although we can’t forecast what skills will be in demand in the future, I think “meta-skills”, which allow humans to acquire new skills quickly, will matter more than specific skills. Meta-skills are, as I explain in a separate article, unlike domain knowledge or technical expertise. Meta-skills such as analogical reasoning, metacognitive regulation, higher-order thinking and social coordination don't directly produce output. Instead, they accelerate learning, enable knowledge transfer across contexts and help people adapt when tasks evolve.

In new work with Alessandro Sforza and Matteo Devigili, I’m trying to pin down the signature of meta-skills by studying “super-jumpers” – individuals who make big leaps in the skills they seem to acquire when transitioning to new jobs. The danger of relying heavily on AI tools is that our meta-skills could atrophy. This means we might be more efficient in the short term but increasingly fragile and commoditised over time.

Build your networks, reputation and visibility

So Yeon Chun, Associate Professor of Technology and Operations Management

AI is often discussed in terms of jobs created or destroyed, but this framing misses a more fundamental shift. AI is altering how work is structured, how value is defined and how opportunity is distributed. In short, AI is rewriting the rules of work.

Rather than focusing on jobs gained or lost, it’s more useful to look at how tasks are redistributed between humans and machines. Even without large-scale unemployment, AI may lead to a more invisible form of disruption in terms of responsibilities, scope and career progression.

This shift changes the nature of value. When high-quality output becomes easy to generate, value moves away from production and towards judgment (i.e. the ability to interpret, evaluate and make decisions). To stay relevant, humans must become skilled at guiding AI systems, assessing their outputs, and applying context and causal reasoning. 

Judgment is also critical for evaluating others’ work. As AI makes it easier to produce polished output at scale, distinguishing truly valuable work becomes more difficult. Thus, those who are better at making their work – valuable or not – visible may be better positioned to capture opportunity. 

This reflects a broader dynamic I highlight in my research: When people have less time and trust to judge an ever-increasing amount of output, visibility increasingly decides whose work is recognised. Appearing busy or visible can replace doing valuable work, and credibility becomes rare.

If you can’t beat them, use them

Winnie Jiang, Assistant Professor of Organisational Behaviour

In an ongoing study of professional workers on the Upwork platform, I’ve observed that those who invest time in learning which tools are best suited for particular tasks, how to combine tools effectively, and how to use them skilfully succeed in turning AI from a threat into a resource.

AI tools also free up time and cognitive capacity, enabling workers to try out new tasks, create side projects and think of new ways to create value. For example, market researchers who use AI for data collection and initial analysis can devote more effort to interpretation and application. In this way, individuals become more career-resilient, while organisations also benefit. 

There’s a caveat: Early-career employees should prioritise hands-on learning, which can mean avoiding AI use when it’s readily available. Our study shows that the professionals who benefit most from AI are those who already know what “good” actually looks like in a given context. In contrast, when individuals rely on AI without first developing this foundational understanding, they often struggle to detect errors or meaningfully improve AI outputs.

For individuals, the prospect of being displaced by AI breeds uncertainty, anxiety and a sense of diminished status and agency. Socially, widespread job insecurity can deepen the divide between those who benefit from AI and those who are displaced. In “mass unemployment” scenarios, the legitimacy of the political and economic status quo could be destabilised. 

To mitigate or avoid these outcomes, workers need to be provided with not only support that helps them reskill but also support to help them reinterpret AI disruption as a temporary transition and an opportunity to find more meaningful work. Leaders, on their part, should treat workers as capable contributors who can identify and create new value, rather than as surplus labour. 

Cultivate deep domain knowledge and “taste”

Victoria Sevcenko, Assistant Professor of Strategy

Most of the labour market change happening now is the restructuring of existing roles to incorporate AI, not the appearance of new categories. The floor on acceptable output seems to have risen: People are expected to come in able to do things they might have been given more time to learn, and the average entry-level job might start to look more like a mid-level job.

Within those redesigned roles, there will likely be more demand for deep domain knowledge and what I will call “taste”, which I shall explain below. Here are some specific actions individuals can take to stay competitive:

• Use AI extensively, preferably the best available models. This will teach you about what AI can and cannot do, and what you, as a human, are uniquely better at. As the models evolve, you can also spot the direction of improvement faster and more accurately.
• Build “taste” in your specific field, which is that intuitive sense of what good work looks like, what counts as a contribution, and what is novel or sloppy. Taste is typically socially constructed: it’s co-created by a community of practitioners and shaped by what they think matters, and much of it isn’t documented well enough for AI to pick up on. Build taste by interacting directly with people in your field and getting regular feedback from peers. You can’t rely on AI alone for this, and it’s also harder to be original in your thinking if you do.
• Build critical thinking. By this I mean the ability to assess what you know, what you don’t and what you are uncertain of, and to step back, reflect and adjust. It’s debatable how “trainable” people are in these skills, but some people are clearly stronger, and this gives them a big advantage. Getting feedback on your work and learning to test your own knowledge are all part of the training.
• Learn how AI works. You don’t need to build the models, but you need enough knowledge to anticipate what AI will be good or bad at, and to make sense of improvements as they emerge.

Next week, we’ll look at what companies and organisations can do to mitigate AI’s impact on jobs and the talent pipeline.