What are core skills?

Core skills is a unit of measurement before it is an idea. It belongs to the World Economic Forum, it is defined by the Forum's own taxonomy, and it exists so that one number can be compared across editions of one survey. That number is 39 per cent, it is quoted in board papers more often than any other figure about skills, and the two things most often said about it are both slightly wrong.

A taxonomy, not a list somebody wrote down#

The 2025 report was built from responses by over 1,000 employers representing more than 14 million workers, across 22 industry clusters and 55 economies. Respondents do not write in whatever they consider a skill. Their answers are placed against the Forum's Global Skills Taxonomy, which is what makes 39 per cent comparable with the 44 per cent published two years earlier. Holding the categories fixed is the whole reason the series exists. It also explains why they read so abstractly: analytical thinking and resilience, flexibility and agility are taxonomy entries rather than phrases a hiring manager would use.

The five skills employers most often call essential today, in the Forum's order, are analytical thinking, which seven companies in ten name; resilience, flexibility and agility; leadership and social influence; creative thinking; and motivation and self-awareness. The next five are technological literacy, empathy and active listening, curiosity and lifelong learning, talent management, and service orientation and customer service. Against the 2023 edition the largest movements are leadership and social influence, up 22 percentage points, then resilience, flexibility and agility and AI and big data, up 17 each. Manual dexterity, endurance and precision records a net decline for the first time in the series, with 24 per cent of respondents expecting it to matter less.

The number has fallen for two editions running#

The 39 per cent almost always arrives in a slide as evidence that skill disruption is accelerating. The series says the opposite. The first edition of the report, in 2016, put expected skill disruption at 35 per cent. It reached 57 per cent in 2020, in the Forum's own words a high point in the wake of the pandemic. It came down to 44 per cent in 2023 and to 39 per cent in 2025.

So the number peaked six years ago and has fallen through two consecutive editions, including the two that cover the entire period since ChatGPT was released. The Forum offers a candidate explanation and marks it as one: the share of the workforce completing training rose from 41 per cent to 50 per cent over the same interval, which would mean employers feel better prepared rather than that the world became calmer. Whichever it is, anybody quoting 39 per cent as a measure of a rising storm is quoting a falling series.

That is the most useful thing on this page and it costs nothing to check. The figures sit in the Forum's own chapter three and in its own Key Findings digest, both public.

Transformed or outdated is not the same as obsolete#

The report's words are that workers can expect two-fifths of their existing skill sets to be transformed or become outdated. Transformed covers a skill that survives in changed form: the same underlying judgement exercised through different tools. Outdated covers one that stops being wanted. The report does not split the 39 per cent between them, and a great deal of the coverage restates the whole of it as obsolescence, which drops half of what the Forum said and turns a statement about change into a statement about loss.

A second drift happens inside the Forum's own house. Chapter three reads 39% of workers' core skills, a claim about the set of skills a given worker holds. The Forum's popular write-up of the same finding reads 39% of key skills required in the job market, which is a claim about the market. Those are different denominators, and the second is the version that travels. Neither is a fabrication and the gap between them is the kind that widens with every retelling.

A correction to our own use of it#

This research has used the figure loosely too, and the record should say so. The Half-Life of Skills, published on 8 June 2025, states that by 2027, 44 per cent of workers' skills will be disrupted. The 44 per cent is the 2023 edition's figure, and by the time that essay appeared the Forum had already published 39 per cent for 2030 in January of the same year. The date attached to it, 2027, comes from the 2023 report's own five-year horizon and does not belong to the newer number.

The same essay attributes to the Future of Jobs report a claim that the half-life of technical skills has collapsed to 2.5 years from ten to fifteen years in the 1980s. That figure appears nowhere in the 2025 report's skills chapter, which was read in full for this page, and no Forum publication carrying it could be located. It circulates widely and is attributed to several different originators depending on who is quoting it. It is not used anywhere on this site, and the estate's page on the underlying question rests on measured retention studies instead: see how fast do skills decay.

The substitution table is the model marking its own paper#

The 2025 report contains one analysis that gets quoted as though it were a measurement of what AI can do. Read the method note before repeating it. Box 3.1, produced with Indeed, sorts more than 2,800 granular skills by their capacity to be substituted by generative AI, and reports that none falls in the highest band while 69 per cent fall in the two lowest. The assessment was made by GPT-4o, rating its own ability to perform each skill.

A system's estimate of its own competence carries no external validation, and the direction of the bias is unknown rather than merely optimistic. The finding is recorded here because it is in the report and because the number is in circulation; it is not treated as evidence of what AI can or cannot do, and no page on this site uses it that way. The estate's material on the gap between how capable a system seems and how capable it is sits at the jagged frontier.

What a survey of stated preference can carry#

Nobody in the Future of Jobs Survey is tested. No skill is measured, no worker is observed, and no outcome is tracked. What is collected is what a senior person in a large employer says about their own organisation's needs, which is a real signal about demand and a poor guide to behaviour. Stated skill preference and hiring behaviour come apart routinely, and the report itself records employers planning to prioritise upskilling for 85 per cent of their workforce while 40 per cent also plan to cut staff whose skills become less relevant.

The figure that deserves more attention than 39 per cent is the one the Forum puts last. Per 100 workers, 59 will need training by 2030: 29 upskilled in their current role, 19 reskilled and redeployed, and 11 who are unlikely to receive it at all. That last group is the report's own estimate of the people the system will not reach. It measures access, where everything else here measures disruption. Whether the training that does happen works is a separate question, taken up at why reskilling programmes mostly fail.

Key sources

• World Economic Forum (2025). The Future of Jobs Report 2025, chapter 3: Skills outlook. Published 7 January 2025. Source of the 39 per cent, the core skills order, the percentage-point movements, the Global Skills Taxonomy and Box 3.1. Graded entry.
• World Economic Forum (2025). The Future of Jobs Report 2025, Key Findings. Source of the 57 per cent in 2020, the 44 per cent in 2023, the phrase transformed or become outdated, and the 59 in 100 training figures.
• World Economic Forum (2025). Future of Jobs Report 2025: the jobs of the future and the skills you need to get them. The Forum's own popular write-up, and the source of the widened wording quoted above.
• Hirji, R. (2025). The Half-Life of Skills. Box of Amazing, 8 June 2025. The essay corrected on this page.

Related SuperSkills research#

On which human capabilities hold their value and why, human skills in the age of AI and what becomes more valuable as AI gets cheaper. On the measured version of the decay question, how fast do skills decay and can you regain a skill you have lost. On the organisational reading of the same problem, capability debt and what is organisational capability. On why the training side of the answer keeps failing, why reskilling programmes mostly fail. On this site's own seven, the SuperSkills thesis.