The AI Skills Boom Has Two Job Markets, Not One

Nearly two-thirds of newly created roles in India’s multinational capability centres now call for AI skills, according to a hiring analysis that returned to the headlines last week. It is a striking number. It is also an easy one to misuse. Treat it as evidence that two-thirds of all workers need to become AI specialists, and a company could spend its training budget solving the wrong problem.

The figure comes from ANSR, a company that helps multinational businesses establish and run global capability centres, or GCCs. Its report, released on 24 September and discussed again by the Financial Express on 21 October, examines more than 5,000 live vacancies and hiring patterns covering more than 20,000 roles. ANSR says about 65% of new GCC roles created in 2026 require AI skills, while demand for AI and data talent rose around 45% year on year.¹²³ Those are the findings of a particular, technology-intensive hiring market, not a census of employment in India or a forecast for every occupation.

The more interesting story is what happens when this specialized market is placed beside evidence about ordinary workplaces. Research from ten countries finds that most explicitly AI-related vacancies remain concentrated in technical occupations. Meanwhile, a European Commission study released this month says nearly one in three EU workers has used AI at work. The apparent contradiction disappears when we distinguish people hired to build AI systems from people whose existing jobs increasingly involve using them.

The denominator changes the story

A global capability centre is not a miniature version of the whole economy. These units are often created by large multinationals to run software, analytics, product engineering, finance and other centralized functions. They are exactly where demand for AI integration specialists might rise quickly. ANSR reports that around 62% of GCC hiring in its sample is for people with three to eight years of experience, with a median of six years. Its strongest growth figure, roughly 1,000% for forward-deployed engineering and AI-engineering postings, concerns a narrow set of job categories.¹² Without the starting number of vacancies, a percentage increase of that magnitude cannot tell us how many people were hired.

The company describes two emerging profiles: AI-native engineers who integrate AI into software development, and forward-deployed engineers who combine engineering with business implementation. Those roles are credible examples of new or newly prominent work created around AI. They also demand a depth of technical knowledge that a charity administrator, shop manager or local-government caseworker is unlikely to need simply to use an AI assistant responsibly.

There is a further reporting limit. ANSR’s public report page and announcement describe the sample and headline figures, but do not publish enough detail on the classification of “AI skills” to let an outside reader reproduce the 65% measure from the public summary alone. It is unclear, from those accessible materials, how consistently the category distinguishes building an AI system from using one in a conventional role. That does not make the figure worthless. It means it should be used as a signal about GCC recruitment, with its scope stated, rather than as a universal benchmark for training budgets.

A technical core is hiring for a different toolkit

A July 2026 working paper by Rafiazka Hilman and Julia Koltai offers a wider comparison. Using online vacancies from ten countries collected between December 2025 and May 2026, the researchers find that STEM and technical occupations account for roughly 75% to 85% of advertisements containing explicit AI-related skills. Business and management occupations form a smaller second group; care, public service, education, logistics and other fields make up much smaller shares. The paper identifies a shared technical vocabulary around Python, SQL, machine learning and data analysis.⁴

This is evidence of where employers explicitly advertise AI expertise, not evidence that AI is irrelevant outside technology. The authors themselves warn that online postings can overrepresent formal urban white-collar employment, that keyword and model-based classification may miss AI use embedded in ordinary tasks, and that the data cover a short period. Their findings are descriptive. They do not establish that AI caused changes in hiring, pay or career mobility.

Still, the contrast with the GCC figure matters. A specialist hiring centre may ask most new recruits for AI skills, while the wider advertised market for those skills remains dominated by technical work. Both can be true. A leader who collapses them into a single slogan about “every job becoming an AI job” loses the most useful distinction: different occupations need different levels of capability.

Using AI at work is not the same as engineering it

The European Commission’s Joint Research Centre drew a different map in its AI and Skills report dated 15 October. Drawing on 2024-2025 evidence, it says about 30% of EU workers reported using AI at work. Reported use was close to 40% among white-collar workers but about 6% in elementary occupations. By 2025, around 20% of EU enterprises had adopted AI technologies.⁵ Those percentages have different denominators and measure different things: worker use, occupation, and enterprise adoption.

They also measure something different from a vacancy that explicitly requires Python or model deployment. A social worker may use AI to structure notes while remaining accountable for sensitive judgments. A small manufacturer may use a ready-made tool to classify incoming inquiries. A communications officer may check a translation drafted by a model. These are illustrative scenarios, not documented cases from the JRC report. None automatically calls for a machine-learning engineer on staff. Each does call for domain knowledge, verification, privacy awareness and a clear decision about what must remain human.

The OECD made this point in a 2024 analysis of job advertisements: most workers exposed to AI will not need specialized skills such as machine learning or natural-language-processing engineering. Their tasks and skill mix may change nonetheless. The OECD also found that management and business capabilities are prominent in AI-exposed occupations, while the direction of demand for particular skills is not uniform.⁶ A training plan built around technical keywords alone would miss much of this transition.

The opportunity is in the bridge, not only the lab

The market between specialist engineering and everyday tool use may be especially consequential for smaller organizations. A capable operations employee who can identify a useful process, understand its data, test an AI-enabled workflow, recognize failures and work with an outside technical provider is performing a valuable bridging function. The job may never be advertised as “AI engineer”. Its responsibilities may nevertheless expand. That is an editorial interpretation of the evidence, not a measured count of newly created positions.

The risk runs in both directions. If employers demand advanced technical qualifications for jobs that mainly require careful tool use, they can raise entry barriers unnecessarily. If they assume a few hours of chatbot training qualifies someone to deploy automated decisions in sensitive settings, they can create unsafe work and expose staff to responsibilities beyond their competence. The July cross-country paper raises a related concern: AI-specific requirements are most visible in an already advantaged technical core and may reinforce barriers to entry. Its data do not prove that this is happening to every occupation.

The European Commission’s 14 October analysis of AI adoption across strategic sectors adds another constraint. It reports adoption rising from 7% of firms in those sectors in 2021 to 18% in 2025, but with substantial differences between industries. It identifies shortages of relevant expertise and regulatory uncertainty among obstacles to adoption.⁷ For a small organization, the binding constraint may be neither the number of employees who can write prompts nor the absence of a data-science team. It may be the lack of someone who understands the workflow well enough to choose a safe, useful application and judge its results.

A skills plan should start with the work, not the headline

A practical way to avoid confusing these markets is to separate three kinds of responsibility. Some people must build and integrate systems: they need software, data, security and evaluation skills. Others must run work with AI support: they need enough tool literacy to frame tasks, protect information and recognize errors. A third group must authorize and oversee uses that affect customers, employees or vulnerable people: they need process ownership, escalation rules and the authority to say no. In a small organization, one person may cover more than one group, but the capabilities are not interchangeable.

For a hypothetical 35-person service company, this distinction changes the investment decision. Before commissioning a company-wide course on advanced AI development, the manager could identify which tasks actually require integration work, which could benefit from approved tools, and which require independent human sign-off. An experienced employee might be trained to test and document a workflow; specialized implementation could be purchased when needed. That is a proposed decision framework, not a tested intervention or a promise of savings.

The right measure of progress would then be whether the organization can complete a defined task more reliably, safely or efficiently, and whether employees understand the limits of the tools they use. Counting AI certificates or inserting “AI” into job descriptions is much easier. It is also less informative.

The wrong question is whether everyone needs AI skills

India’s GCC hiring data show that advanced AI capability is becoming a serious recruitment priority in a specialized, expanding part of the economy. Cross-country vacancies show how concentrated that explicit demand remains. European worker surveys show AI reaching far beyond those technical job titles. None of the three sources cancels out the others. They describe different layers of the same labor market.

For leaders deciding whom to hire and whom to develop, the useful question is not “What percentage of jobs require AI?” It is “Which people must build it, which must use it well, and which must be able to challenge it?” Confusing those answers could create an expensive skills gap of the organization’s own making.

How hilighter can help

When AI skills mean different things in different roles, teams need a clear map of work, evidence and growth. hilighter helps map skills by role, compare people with requirements in Skills & Talent Insights, recognize learning in Skills Evidence & Recognition, set personalized growth paths, and explore an indicative view of AI’s potential impact on jobs. A free plan covers up to three roles.

Sources

1. ANSR, Talent Trends in GCCs in India 2026, 24 September 2026.
2. ANSR, New ANSR Report Finds AI Skills Driving GCC Hiring in India, press release, 24 September 2026.
3. Financial Express, 65% of new GCC roles need AI skills, 21 October 2026.
4. Hilman, R. and Koltai, J., Occupational Convergence or Divergence?, working paper, 30 July 2026.
5. European Commission Joint Research Centre, AI and Skills, 15 October 2026.
6. OECD, Artificial intelligence and the changing demand for skills in the labour market, 10 April 2024.
7. European Commission Joint Research Centre, AI adoption more than doubles in almost every EU strategic sector, 14 October 2026.

Sources last reviewed: 27 October 2026. ANSR figures describe GCC hiring samples; the 65% “AI skills” classification is not fully reproducible from the public summary alone. JRC and OECD figures measure different denominators (worker use, vacancies, enterprise adoption).