AI is hollowing out the informal apprenticeships that used to form early‑career knowledge workers, and simple curriculum fixes won’t restore them. The author proposes a four‑part 'Embedded Formation Degree'—combining accelerated domain entry, a four‑year AI fluency track, embedded practice firms, and tied employer partners—to rebuild pathways from graduate to credible professional.

Apprenticeship after AI: Bridging Gaps in Early-Career Knowledge-Work Roles

April De Crescentis, Biff Baker · June 04, 2026 · Journal of Higher Education Theory and Practice

openalex review_meta medium evidence 7/10 relevance DOI Source PDF

AI is compressing early‑career entry tasks and eroding informal post‑degree apprenticeship pathways, meaning curricular tweaks alone are insufficient; the paper proposes an 'Embedded Formation Degree' combining accelerated domain entry, AI fluency, practice firms, and integrated employer partnerships to rebuild professional formation.

Artificial intelligence and automation are restructuring early-career knowledge-work roles by compressing the entry-level functions through which graduates historically built portfolios, developed professional judgment, and earned professional credibility. This paper argues that higher education has misdiagnosed the resulting challenge as curriculum misalignment—a content problem assumed to be solvable through revised syllabi, AI electives, and marginal expansions of experiential learning. Evidence from a systematic review of eighteen peer-reviewed studies and current labor-market analyses suggests a different conclusion: the informal post-degree apprenticeship system that historically completed graduate formation no longer reliably exists, and the architecture of the undergraduate degree is structurally incapable of replacing it through curricular revision alone. In response, this paper proposes the Embedded Formation Degree (EFD), a four-component framework consisting of accelerated domain entry, a four-year AI fluency track, an embedded practice firm, and structurally integrated employer partners. The framework reframes the education–employer gap as a structural failure in the pathway and outlines implications for universities, employers, accreditors, and policymakers.

Summary

Main Finding

AI and automation are structurally compressing traditional entry‑level knowledge‑work tasks, disabling the informal post‑degree apprenticeship pathway that universities historically relied on to complete professional formation. Curricular tweaks alone are insufficient. The authors propose an architectural reform—the Embedded Formation Degree (EFD)—that embeds apprenticeship‑style formation inside the four‑year degree via four integrated components so graduates arrive with judgment, portfolio evidence, and verified practice-ready skills.

Key Points

Compression of entry‑level work: Generative AI is automating routine drafting, research, preliminary analysis and similar tasks that historically formed on‑ramps for graduates (Hui et al., 2024; Brynjolfsson et al., 2025). The experience curve is compressed; employers increasingly expect judgment and oversight typically associated with mid‑career workers.
Epistemic shift: Value is shifting from execution to judgment, oversight, and accountability for AI outputs. AI literacy must therefore be about governance and critical use, not just tool proficiency (Ouyang & Jiao, 2021; Sun & Deng, 2025; Diaz et al., 2025).
Failure of post‑degree apprenticeship: Informal, employer‑provided formation no longer reliably exists because firms rationally reduce investment in novice formation when AI substitutes junior tasks (Beane & Anthony, 2024; Alimohammadi, 2025).
Labor‑market evidence of mismatch: Degree signals are weakening in AI‑exposed occupations; employers increasingly demand portfolios and verified competencies. Examples: bachelor’s‑level postings requiring any AI skill rose to 4.7% in 2024 (vs 1.4% for associate); demand for AI skills at bachelor’s level grew 240% from 2010–2024 with a sharp rebound 2023–24 (Galeano et al., 2025). Several growing knowledge‑work occupations now carry O*NET Job Zone 4 expectations (ETA, 2026).
Institutional responses so far are inadequate: Updating syllabi, adding AI electives, and marginal experiential learning treat the problem as content misalignment rather than a structural pathway failure.
Proposed solution (EFD): four integrated components designed to restore formation within the degree:
Accelerated domain entry (structured early immersion so students begin practical formation sooner).
A four‑year AI fluency track (progressive, accountability‑focused AI governance and judgment training).
An embedded practice firm (a university‑hosted, realistic practice environment where students generate portfolio artifacts under supervision).
Structurally integrated employer partners (long‑term, contractual engagements that create predictable channels for mentoring, real work, and signaling).

Data & Methods

Research design: Conceptual framework built from a structured synthesis of empirical literature and practitioner knowledge (Jabareen, 2009; Rocco & Plakhotnik, 2009).
Systematic literature review:
- Time window: peer‑reviewed studies published 2020–2025, with emphasis on 2023–2025 to capture rapid AI adoption effects.
- Databases searched: EBSCO, Google Scholar, and discipline‑specific repositories.
- Keywords: combinations around “artificial intelligence,” “early‑career employment,” “generative AI,” “knowledge work,” “experiential learning,” “apprenticeship,” and “entry‑level restructuring.”
- Screening and inclusion: dual‑author review; included studies had to be peer‑reviewed, relevant to the research questions, methodologically transparent, and reasonably applicable to North American higher education. Excluded studies lacking labor‑market connection or transferable methodology.
- Final corpus: 18 peer‑reviewed sources spanning causal field experiments, econometric analyses of labor platforms, qualitative ethnographies, simulation studies, job‑posting and establishment‑panel analyses, and institutional case studies (cited throughout the paper: e.g., Hui et al., 2024; Brynjolfsson et al., 2025; Mayer et al., 2025; Muehlemann, 2025; Galeano et al., 2025).
Supplementary evidence: labor‑market datasets and secondary analyses (O*NET/ETA projections, job‑posting analyses) used to illustrate changing employer expectations and degree signal attenuation.
Methodological framing: The paper intentionally synthesizes diverse methods (quantitative causal work + qualitative ethnography + institution‑level evaluations) to identify a structural phenomenon that individual empirical approaches alone might understate.

Implications for AI Economics

Human‑capital formation and returns: The shift implies that private returns to a traditional bachelor’s degree will diverge across institutions and students depending on whether formation (judgment, portfolios, supervised practice) is delivered. Standard estimates of degree returns that ignore embedded formation will overstate average readiness and understate heterogeneity in labor outcomes.
Signaling and market sorting: As degrees lose signal value in AI‑exposed occupations, alternative signals (verified competencies, work portfolios, apprenticeship credentials) will gain market value. This reshapes screening costs, matching frictions, and wage dispersion—potentially increasing returns to students with access to embedded formation or networks.
Employer investment incentives: Evidence (Muehlemann, 2025) shows firms already committed to training may increase apprenticeships despite AI, but most firms do not self‑generate formation. That creates a coordination failure: private firms underinvest in entry‑level formation when AI raises the near‑term productivity of experienced workers, producing under‑provision of human capital that universities could internalize.
Policy and public economics: There is a potential public‑good role for subsidizing embedded formation (e.g., seed funding for university practice firms, wage subsidies for employer‑partners, accreditation pathways that reward integrated formation). Policies should target the externalities of underprovided apprenticeship—inequality, skill mismatch, and underemployment of graduates.
Labor productivity and distributional effects: If universities successfully implement EFD‑style reforms, aggregate productivity could benefit through faster on‑ramp to high‑value judgment roles; but without broad adoption, inequality may widen as formation concentrates among well‑resourced institutions and students with existing network advantages.
Research and measurement needs: Economists should track (a) changes in degree wage premia by occupation and institution, (b) returns to verified competency and apprenticeship credentials, (c) employer hiring criteria dynamics (portfolios vs. degrees), and (d) long‑run career trajectories of cohorts graduating with embedded formation versus traditional degrees.
Practical levers: From an economic policy perspective, useful interventions include subsidizing university‑based practice firms, creating standardized competency verification instruments, incentivizing durable employer‑university contracts (to internalize long‑term returns to training), and aligning accreditation standards to reward formation outcomes rather than seat‑time/content lists.

Overall, the paper reframes the AI‑education problem from “update the syllabus” to “rebuild the institutional architecture of professional formation,” with direct consequences for human‑capital models, signaling theory, employer incentives, and policy interventions in the age of AI.

Assessment

Paper Typereview_meta Evidence Strengthmedium — Based on a systematic review of 18 peer‑reviewed studies and contemporary labor‑market analyses, the paper synthesizes observational and descriptive evidence that the post‑degree apprenticeship has weakened; however, the underlying studies appear heterogeneous, largely non‑experimental, and do not provide strong causal identification of AI as the singular driver. Methods Rigormedium — The work uses a systematic review plus labor‑market analysis and builds a coherent theoretical framework, but the description given does not report detailed inclusion/exclusion criteria, risk-of-bias assessment, meta‑analytic techniques, or primary causal estimation; empirical bases are therefore informative but not methodologically decisive. SampleA synthesis of eighteen peer‑reviewed studies (mix of qualitative, descriptive, and observational quantitative work) combined with current labor‑market analyses and policy/sector reports; no new randomized or quasi‑experimental data collection is reported. Themesskills_training labor_markets human_ai_collab org_design adoption GeneralizabilityLikely concentrated on early‑career knowledge roles in particular economies (e.g., US/UK/OECD) — geographic context may not generalize globally, Heterogeneity across occupations and industries — effects differ by sector, firm size, and unionization, limiting broad extrapolation, Rapidly evolving AI capabilities mean findings may age quickly as tools and employer responses change, Small number of reviewed studies and likely mixed methodologies reduce external validity across all knowledge‑work fields, Policy and institutional contexts (education systems, labor market regulations, employer training norms) vary, constraining transferability

Claims (6)

Claim	Direction	Confidence	Outcome	Details
Artificial intelligence and automation are restructuring early-career knowledge-work roles by compressing the entry-level functions through which graduates historically built portfolios, developed professional judgment, and earned professional credibility. Skill Obsolescence	negative	high	compression of entry-level functions used for portfolio-building, judgment formation, and credibility acquisition among early-career knowledge workers	n=18 0.24
Higher education has misdiagnosed the resulting challenge as curriculum misalignment—a content problem assumed to be solvable through revised syllabi, AI electives, and marginal expansions of experiential learning. Training Effectiveness	negative	high	adequacy of curricular fixes (revised syllabi, AI electives, marginal experiential learning) to address the education–employer gap	n=18 0.24
The informal post-degree apprenticeship system that historically completed graduate formation no longer reliably exists. Skill Acquisition	negative	high	presence/reliability of informal post-degree apprenticeship pathways for graduate formation	n=18 0.24
The architecture of the undergraduate degree is structurally incapable of replacing the informal post-degree apprenticeship system through curricular revision alone. Training Effectiveness	negative	high	capacity of undergraduate curricular revisions to substitute for post-degree apprenticeship in graduate formation	n=18 0.24
The paper proposes the Embedded Formation Degree (EFD), a four-component framework consisting of accelerated domain entry, a four-year AI fluency track, an embedded practice firm, and structurally integrated employer partners. Training Effectiveness	positive	high	proposed structural intervention (EFD) to address gaps in graduate formation and early-career pathways	0.04
The framework reframes the education–employer gap as a structural failure in the pathway and outlines implications for universities, employers, accreditors, and policymakers. Governance And Regulation	negative	high	characterization of the education–employer gap (structural pathway failure) and associated implications for stakeholders	0.04