Human skills in the age of AI

The human skills that matter most as AI improves are not the ones AI cannot do. They are the ones that carry accountability. AI can now assist almost any task and imitate the surface of many human skills, tone, apparent expertise, even the words of empathy, but it cannot own the consequences of a decision, and it cannot tell you when it is confidently wrong. That is where durable human value now sits: in judgement, contextual wisdom, ethical reasoning, and the relationships and decisions a person has to stand behind. The right question is therefore not which jobs are AI-proof, a framing that treats this as a fixed contest between humans and machines. It is which capabilities hold or gain value as AI absorbs the routine work around them, and how you deliberately build and keep them. Those capabilities can be developed, augmented, or lost, depending entirely on how the work is designed.

The wrong question, and the better one#

Most advice starts from "which jobs are safe from AI?" It is the wrong place to start, because it assumes capability is a fixed property of a role rather than something a person builds and can lose. It leads to list-making, and lists of "skills AI can't replace" tend to be both generic and static, as if the line between human and machine were drawn once and for all.

A more useful distinction has three levels. Task resistance is whether AI can do a specific task; it is falling fast and will keep falling, so it is a poor thing to bet a career on. Role resistance is whether a whole job survives; it depends on the mix of tasks and changes slowly. Capability durability is different from both: it is whether a human capability holds or grows in value as AI spreads, regardless of which tasks or roles come and go. Durability is the thing worth building, because it survives the churn underneath it. The seven SuperSkills are chosen on exactly this test: not what AI cannot yet do, but what stays valuable as it can do more.

What AI can do, and what it cannot#

To see where durable value sits, separate three things AI does that are usually blurred together. It can assist: gather information, draft, summarise, generate options, find patterns. It can imitate: reproduce the tone, fluency and apparent confidence of expert human work, including the language of empathy and care. And there is a third category it cannot enter: accountability, the ownership of a decision and its consequences, the ethical judgement behind it, the contextual wisdom to know when a fluent answer is wrong. Assistance is real and valuable. Imitation is the trap, because imitation of a skill is easily mistaken for the skill. Accountability is the thing that does not transfer. It is where human capability keeps its worth.

The practical test for any skill is which column it falls into.

• AI can assist: research, drafting, summarising, option generation, first-pass analysis, pattern-finding. Use it here freely; the leverage is real.
• AI can imitate: tone and style, the appearance of seniority, fluent reasoning, empathetic wording, confident recommendations. Treat output here with care; the surface can outrun the substance.
• Requires human accountability: the final decision, ethical and values judgement, contextual and cultural wisdom, ownership of consequences, and knowing when to override the machine. This does not move to AI, and this is where the durable skills live.

How the labour market prices these skills#

The labour market is already repricing exactly these capabilities. The World Economic Forum's 2025 Future of Jobs report, drawing on employers of more than 14 million workers, names analytical thinking the single most sought-after core skill, followed by resilience, flexibility, leadership and social influence, and it finds skills gaps to be the biggest barrier to business transformation over the next five years. PwC's Global AI Jobs Barometer, built from close to a billion job postings, is sharper still: new tasks appearing in AI-exposed roles are two and a half times more likely to demand human skills such as empathy, judgement and creativity, and roles where AI raises the premium on human judgement are growing faster and paying salaries that rise markedly quicker than roles AI simply makes easier.

Two findings show the mechanism directly. PwC reports that entry-level roles most exposed to AI are now seven times more likely to require traditionally senior human skills such as leadership and face-to-face judgement, and that these "seniorised" entry roles have grown by more than a third since 2019: AI strips out the routine and leaves the human-intensive part exposed at every level. And Brynjolfsson, Li and Raymond, studying 5,172 support agents, found AI lifts the output of novices most while barely moving experts, because it transfers expert patterns downward. AI raises the floor of what people can produce. It does not raise the ceiling of what they can judge. The gap between those two is where human skill now earns its premium.

The seven human skills that hold their value#

The seven SuperSkills are not a list of nice-to-haves. Each answers a specific failure mode that AI introduces, so each grows more valuable, not less, as adoption rises.

• Curiosity answers the failure mode where a fluent first answer ends the inquiry. When the machine always has a response, the discipline of asking the better question, and of not stopping at the plausible, becomes the scarce input.
• Change Readiness answers the churn: the tools change monthly, and the capability to keep adapting without exhaustion is what lets a person stay effective through it rather than freeze.
• Big Picture Thinking answers the tunnel vision of AI micro-outputs. Models optimise the task in front of them; holding the system, the second-order effects and the long horizon is the human job that keeps the local answer from breaking the whole.
• Empathy answers the imitation trap most directly. AI can produce empathetic words; it cannot be accountable in a human relationship. As simulated care becomes cheap, real care becomes the differentiator.
• Global Adaptability answers the model's blind spots. Training data has a centre of gravity; the judgement to read local, cultural and contextual meaning that the model flattens is a distinctly human edge.
• Principled Innovation answers the risk of deploying capable systems without ethical grounding. The capacity to ask not only whether something can be built but whether it should be grows in value as building gets easier.
• The Augmented Mindset answers the crutch-versus-amplifier choice. It is the meta-capability of working with AI so it extends your thinking rather than replacing it, and it determines whether the other six are strengthened or eroded by the tools.

Where the framing is contested#

Two honest caveats. First, no list of human skills is permanent; the specific tasks under each will keep shifting, and anyone who claims a fixed catalogue of "things AI will never do" is likely to be wrong about the details. The value of the seven is that they are chosen on durability rather than on current machine limitations, but they are a lens, not a law. Second, the evidence that these skills are being repriced upward is strong and current, from the WEF and PwC data above, but it measures demand and wages, not a proven long-run causal claim about which capabilities are irreplaceable. The safer and more useful reading is the one this whole body of work rests on: capability is not fixed, it can be built or lost, and the job is to design work so the durable skills are practised rather than automated away.

How to build them, and what leaders should protect#

For individuals, the method is consistent across all seven: keep doing the thinking AI could do for you, deliberately, so the capability stays yours. Use AI to assist and to test your reasoning, not to replace the first move where judgement is formed. Build what might be called a judgement portfolio: a record of the decisions you made, what you overrode in the machine's output and why, which is the evidence of capability that a polished deliverable no longer provides.

For leaders, the task is to protect the capabilities that carry accountability while letting AI take the rest. That means deciding where human judgement must remain rather than letting automation settle it by default, keeping deliberate practice in the system so people still build the durable skills, treating verification as real skilled work, and measuring capability directly rather than trusting that good output means a capable person. The organisations that thrive will not be those that automate fastest. They will be those that are clearest about which human capabilities they cannot afford to lose, and that design the work to keep them. This is the practical content of drift versus design, and the reason unmanaged automation accrues capability debt.

Key research and primary sources

• Bacigalupo, M., Kampylis, P., Punie, Y. and Van den Brande, G. (European Commission JRC) (2016). EntreComp: The Entrepreneurship Competence Framework. Publications Office of the European Union, JRC101581.
• ILO, ETF, Cedefop, Eurofound, European Commission and UNESCO (2026). Changing landscape of skills in the age of AI. Joint publication, 13 August 2026.
• OECD (2019). OECD Learning Compass 2030. OECD Future of Education and Skills 2030 project.
• OECD (2026). AI and skills: What we know so far. OECD policy brief, 5 June 2026.
• OECD (2026). Skills in the AI age. OECD Artificial Intelligence Papers No. 60, July 2026.
• Pearson (2022). Pearson Skills Outlook: Power Skills. Pearson, November 2022.
• Vuorikari, R., Kluzer, S. and Punie, Y. (European Commission JRC) (2022). DigComp 2.2: The Digital Competence Framework for Citizens. Publications Office of the European Union, JRC128415.
• World Economic Forum with the McKinsey Health Institute (2026). The Human Advantage: Stronger Brains in the Age of AI. World Economic Forum, 15 January 2026.
• World Economic Forum (2025). The Future of Jobs Report 2025.
• PwC (2025). Global AI Jobs Barometer.
• Brynjolfsson, E., Li, D. and Raymond, L. (2025). Generative AI at Work. Quarterly Journal of Economics, 140(2), 889-942. Earlier version NBER Working Paper 31161.
• Lee, H.-P. et al. (2025). The Impact of Generative AI on Critical Thinking. Microsoft Research and Carnegie Mellon, CHI 2025.

Related SuperSkills research#

Each of the seven has its own page in the research, and this argument connects to AI and human judgement, AI and critical thinking, capability debt and drift versus design. Start with the introduction to the seven SuperSkills. On what remains distinctly human even where AI performs it better, see what stays human; on the career question, staying valuable in the age of AI. On what not to build, see why "learn to prompt" is weak career advice. For the wider map of the field, see the essential works.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. This work draws on research across more than 200 organisations in 30 countries over seven years. Findings are attributed to the studies that produced them and kept separate from the interpretation and the framework, which are the author's. This is a living reference, reviewed and updated as significant new evidence appears.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.