The Finding

PwC's 2026 AI Jobs Barometer, which analyzed over one billion jobs worldwide and 2.4 million entry-level roles in the United States, landed this week with a finding that deserves more careful theoretical attention than it has received in the business press. Employers are now demanding senior-level competencies from entry-level candidates. The framing in most coverage treats this as a supply-demand problem: employers want more than the market currently provides, and workers need to upskill. That framing is not wrong, but it is incomplete. What PwC has actually documented is a structural coordination failure, and the mechanism behind it is worth examining carefully.

Why This Is Not Simply a Skills Gap

The conventional "skills gap" narrative implies that workers lack specific, identifiable competencies that training programs can remedy. But the PwC data points to something more disorienting. The competencies employers are demanding are not just technical. They are adaptive: the ability to work alongside AI systems whose outputs, constraints, and failure modes are not stable or fully legible to the workers using them. This is not a gap that procedural training closes easily. Hatano and Inagaki (1986) drew a foundational distinction between routine expertise, which is the ability to execute known procedures reliably, and adaptive expertise, which is the ability to modify and extend one's competence when context changes. What employers are describing when they say they want "senior-level skills" from entry-level workers is, in most cases, the second category.

Routine expertise scales through training pipelines. Adaptive expertise does not. This is the coordination failure the PwC data is actually surfacing, and it is one that classical labor market theory is not well-equipped to explain.

The Awareness-Capability Gap at the Hiring Stage

There is a specific structural problem embedded in the PwC finding that connects directly to what algorithmic literacy research calls the awareness-capability gap. Research by Gagnarin, Naab, and Grub (2024) and by Kellogg, Valentine, and Christin (2020) consistently shows that workers who are aware of algorithmic systems do not automatically develop the competence to perform better within them. Awareness is a necessary but not sufficient condition. The same asymmetry is now showing up in hiring. Entry-level candidates in 2026 are, almost universally, aware that AI is transforming their fields. Many have completed AI literacy modules, used large language models extensively, and can speak fluently about automation. But employers are signaling, through the PwC data, that this awareness is not converting into the adaptive performance they need. The candidates know the terrain exists; they do not yet know how to navigate it under novel conditions.

Why Consulting Firms Are Still Hiring Entry-Level Workers

A separate but related data point from this week is worth connecting here. Reporting on consulting firm hiring confirms that McKinsey, BCG, and similar firms are still recruiting new graduates, even as AI displaces significant portions of the analytical work those graduates would traditionally have performed. This seems paradoxical until you apply the coordination lens. These firms are not hiring for the procedures. They are hiring for the scaffolding: the structural understanding of how to frame problems, how to assess the reliability of algorithmically-generated outputs, and how to communicate recommendations to clients who are themselves navigating the awareness-capability gap. What consulting firms have recognized, perhaps ahead of most employers, is that the value of entry-level talent is no longer located in execution capacity. It is located in schema quality.

Gentner's (1983) structure-mapping theory is useful here. Workers who possess accurate structural schemas - representations of how systems behave across contexts, not just how to operate them in one context - are capable of analogical transfer when environments change. Workers trained only in platform-specific or tool-specific procedures are not. The consulting firms still hiring new graduates appear to be selecting, even if implicitly, for schema quality over procedural fluency.

What the PwC Data Should Change About How We Train Workers

The practical implication is that employer demands will not be met by adding more tool-specific training to university curricula or corporate onboarding programs. If the demand is genuinely for adaptive expertise - the ability to work competently with AI systems that will change - then the training model needs to target structural understanding rather than procedural knowledge. Rahman (2021) argues that algorithmically-mediated work environments create asymmetric power precisely because workers lack visibility into the structural constraints governing their performance. The PwC barometer is documenting the labor market consequence of that asymmetry at scale. Employers can see the gap. They are pricing it into hiring requirements. The field has not yet converged on a training model capable of closing it.

That is the coordination problem the 2026 AI Jobs Barometer is actually describing, and it is a more durable structural issue than the skills gap framing suggests.

References

Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7(2), 155-170.

Hatano, G., & Inagaki, K. (1986). Two courses of expertise. Research and Clinical Center for Child Development, 262, 27-36.

Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366-410.

Gagnarin, A., Naab, T. K., & Grub, J. (2024). Algorithmic media use and algorithm literacy. New Media & Society.

Rahman, H. A. (2021). The invisible cage: Workers' reactivity to opaque algorithmic evaluations. Administrative Science Quarterly, 66(4), 945-988.