The Future of Work Will Reward the Skills Machines Cannot Own

Artificial intelligence is forcing universities to confront an uncomfortable question: are they preparing students to use new technologies, or preparing them to think, judge and work effectively when those technologies become unavoidable?

A new study "Artificial Intelligence and Transversal Skills for Future Workforce: A Bibliometric Analysis of Research in Higher Education (2020–2025)," published in Education for Information and authored by J. M. Shalani Dilinika, analyses 1,487 Scopus-indexed journal articles to map how higher education research is responding to the skills demands created by artificial intelligence.

According to the study, AI literacy matters, but it is only one part of the emerging workforce equation. Critical thinking, collaboration, ethical reasoning, self-regulation and metacognition, the ability to monitor one's own thinking, are becoming just as important because they determine whether people can question, verify and responsibly act on machine-generated outputs.

The research also reveals a widening gap between recognition and reform. Scholarship on AI-related skills is expanding at extraordinary speed, yet universities are struggling to convert that knowledge into curriculum design, assessment, faculty development and meaningful preparation for employment.

The Research Boom Is Outrunning Institutional Reform

The study tracks a field that barely existed at scale six years ago and has since exploded. Annual publication output remained below 100 articles between 2020 and 2022. It rose to 133 in 2023, climbed to roughly 310 in 2024 and exceeded 850 in 2025. The calculated annual growth rate was 94.95%, with the sharpest acceleration following the widespread emergence of generative AI. This surge shows how quickly ChatGPT and related tools changed the higher education agenda.

Before generative AI became widely accessible, much of the debate focused on digital competence, educational technology and future-of-work skills in relatively broad terms. After 2022, research shifted toward questions of AI literacy, academic integrity, assessment, human–AI collaboration and graduate readiness, but publication growth should not be mistaken for institutional transformation.

The study identifies an "awareness-to-practice gap": universities increasingly understand that AI is altering learning and employment, yet many lack the organisational mechanisms needed to translate research into programme-level change. Curriculum review cycles are often slow, faculty capacity is uneven and accreditation systems may not keep pace with rapidly evolving competency demands. This creates a strategic risk.

Universities may add isolated AI modules, tool demonstrations or policy statements while leaving the deeper structure of teaching unchanged. Students could graduate knowing how to prompt a chatbot without knowing how to challenge its assumptions, detect errors, manage uncertainty or decide when human judgment should override automated advice.

The pace of technological change makes that approach unsustainable. Specific tools will evolve, disappear or become embedded in everyday software. The more durable educational task is to develop capabilities that remain valuable across technologies, industries and professional roles. It means higher education reform cannot be limited to adding content. It must reconsider what counts as knowledge, how students demonstrate understanding and which forms of human reasoning deserve greater emphasis when machines can produce plausible answers instantly.

AI Literacy Matters But Judgment Will Matter More

The study finds that AI literacy has become the most prominent transversal skill in the literature. It appeared 72 times among author keywords, ahead of digital literacy, digital competence and critical thinking. "ChatGPT" appeared 132 times, reflecting the extent to which one technology has shaped the field's recent direction.

AI literacy is broader than knowing how to operate a tool. It includes understanding what AI systems can and cannot do, how their outputs are produced, where bias or fabrication may arise and how to evaluate results against reliable evidence. According to the study, technical familiarity alone is not enough. The research landscape is increasingly moving toward human-centred abilities that help people work with AI without surrendering responsibility to it.

Critical thinking enables students to interrogate algorithmic outputs rather than accept them at face value. Information literacy helps them distinguish credible evidence from synthetic or misleading material. Ethical reasoning becomes essential when AI systems affect hiring, diagnosis, finance, education or public services. Collaboration and communication matter because AI-enabled work will still require negotiation, interpretation and accountability across teams.

Metacognition may be even more important. Students must recognise whether they genuinely understand a problem or have merely received a convincing answer from a machine. They must decide when AI support improves performance and when it weakens independent reasoning.

The study identifies metacognition, self-regulation and ethical engagement as emerging but still relatively underdeveloped research areas. Workforce readiness also remains at the edge of the field rather than its centre. The imbalance should concern policymakers and university leaders. Institutions appear to be studying AI use faster than they are examining the cognitive dependency, ethical risks and professional consequences that may follow.

The key employability challenge is therefore changing. Graduates will not compete only on what they know or how quickly they can produce an answer. They will increasingly be judged on whether they can frame the right problem, evaluate evidence, explain decisions, manage risk and take responsibility for outcomes produced with AI assistance.

Curriculum Reform Will Fail Without Assessment and Faculty Change

The study identifies five closely connected research themes: curriculum transformation, pedagogical innovation, AI literacy and digital competence, educator capacity building, and professional readiness grounded in ethics, information literacy and self-regulation. Together, these themes point to a simple conclusion: universities cannot prepare students for AI-driven work through one course, one policy or one software platform.

AI-related competencies need to be embedded across disciplines. Engineering students may need to evaluate automated designs and explain safety trade-offs. Medical students may need to scrutinise diagnostic recommendations and recognise when human judgment is indispensable. Business students may need to assess algorithmic bias in recruitment or lending. Journalism students may need to verify synthetic media and disclose AI-assisted production.

This does not mean every student must become a computer scientist. It means every graduate should understand how AI affects knowledge, evidence, responsibility and professional decision-making within their field.

Assessment is the pressure point. Traditional take-home essays and standardised assignments are increasingly vulnerable to undisclosed AI assistance. Simply banning AI may prove difficult to enforce and educationally shortsighted. But permitting unrestricted use without redesigning assessment risks making it impossible to know what students can do independently.

The study points toward authentic, process-oriented and competency-based assessment. Universities may need to evaluate how students define problems, document their reasoning, verify sources, revise outputs and defend conclusions, not only the final product they submit.

Faculty development is equally critical. Lecturers cannot teach AI literacy, redesign assessment or model responsible practice if they lack confidence with the technology themselves. The study treats educator capacity building as an immediate institutional priority, not a secondary support function.

Investment needs to extend beyond technology procurement. It should include teacher training, curriculum support, local research capacity, accessible infrastructure and safeguards that prevent students from becoming dependent on commercial platforms they cannot afford.

The Missing Link Is the Labour Market Itself

According to the study, the research connecting transversal skills with actual workforce demands remains comparatively weak. Terms such as "employability," "labour market" and "readiness" appear at the periphery of the keyword network. The field is highly active in discussing AI, education and digital competence, but less developed in demonstrating whether the skills taught by universities align with changing professional roles. This leaves a major unanswered question: are universities preparing graduates for work as it is genuinely changing, or for an academic interpretation of what future work might require?

Stronger engagement with employers, professional bodies and workers is essential. Industry partnerships could help identify which entry-level tasks are being automated, which roles increasingly require human oversight and where new competency gaps are emerging. But employers should not be allowed to define the curriculum alone.

Universities have responsibilities that extend beyond immediate labour-market demand. Ethical judgment, civic responsibility and the ability to question harmful systems may be economically useful, but they are also essential to democratic and socially accountable institutions. The real goal should be a broader model of workforce readiness, one that combines employability with agency. Graduates should be able to use AI productively, but also recognise when its use is inappropriate, discriminatory or unsafe.

The authors also caution that the bibliometric analysis maps publication patterns, citation relationships and keyword clusters; it does not establish which teaching approaches produce better employment outcomes. It relies exclusively on Scopus, includes only English-language journal articles and excludes conference papers, books and research indexed elsewhere. The author also cautions that the field may already have evolved since the February 2026 search.

Still, the research captures a decisive shift in higher education. The debate is moving beyond whether students should use AI. It is increasingly about what capabilities they will need when AI is embedded in nearly every profession.