AI is reshaping, not simply erasing, low‑skilled jobs in the UK: workers report task restructuring and mixed efficiency/displacement effects, while age matters mainly through digital skills and training access rather than as a deterministic barrier.

Main Finding

Artificial intelligence primarily causes task restructuring among low‑skilled workers in the UK rather than wholesale job elimination. Employment outcomes depend on workers’ adaptive capacity — shaped jointly by individual, organisational and institutional factors — with age acting as an indirect moderator via differences in digital literacy and access to training rather than as a deterministic factor.

Key Points

• Definition: Low‑skilled labour is defined multidimensionally (ISCED 0–2 baseline plus task/occupational features emphasizing routine, codifiable tasks), and can include formally better‑qualified workers whose core tasks are highly routinised.
• Primary mechanism: AI changes work at the task level (workflow optimisation, reduction/simplification of repetitive tasks, partial task substitution), producing simultaneous efficiency gains and displacement pressures.
• Adaptive capacity: The divergence in re‑employment outcomes is mediated by adaptive capacity — a combination of individual skills (especially digital literacy), firm practices (training, job redesign, welfare), and institutional supports (policy, regulation, training access).
• Age heterogeneity: Age shapes transitions indirectly. Older low‑skilled workers face greater barriers primarily because of lower digital literacy, less access to training and resource constraints; age per se is not a deterministic predictor once these mediating factors are accounted for.
• Stakeholder perceptions:
  • Frontline workers (n=153) expressed mixed views: top concerns were need for skills training, policy support and labour‑market reshuffling (each ~14.7% of open responses). Both fear of unemployment and recognition of new opportunities were present.
  • Managers (n=162) prioritized social fairness (17%), industry regulation (15.1%), data privacy (13.2%), human–AI collaboration (12.6%) and skills training (11.9%).
• Policy orientation: The paper argues for differentiated labour‑transition policies that target diverse groups and strengthen inclusive training, firm incentives, and social protections.

Data & Methods

• Context: United Kingdom; study aims to provide individual‑level qualitative evidence on AI‑driven employment change among low‑skilled workers.
• Design: Exploratory mixed‑evidence survey with grounded‑theory‑informed qualitative analysis. Original semi‑structured interview plan was revised; data were collected via structured questionnaires with closed and open‑ended items.
• Samples:
  • Frontline‑worker questionnaire: 153 total responses; usable open‑ended responses: Q14 (observed workplace changes) n=150; Q19 (concerns/expectations) n=150.
  • Management questionnaire: 162 total responses; usable open‑ended responses: Q22 (specific organisational response) n=4 substantive answers; Q23 (recommended organisational/policy response) n=159.
• Analysis:
  • Quantitative descriptive analysis of closed items.
  • Inductive coding of open responses using Constructivist Grounded Theory logic (initial → focused → axial → selective coding; constant comparison).
• Key empirical codes/findings from frontline workers (Q14): workflow optimisation, reduction of repetitive tasks, task simplification, partial job substitution, need for new skills, reduced overtime, income decline/diversification.
• Limitations noted: purposive (non‑probability) sampling, questionnaires (open responses shorter than interviews) — so results are exploratory and interpretive rather than statistically representative or fully theory‑saturating.

Implications for AI Economics

• Micro‑foundations: Models of AI’s labour market impact should prioritize task‑level change and incorporate adaptive capacity as a central mediator rather than treating age or skill groups as homogenous categories.
• Heterogeneity and pathways: Age should be modeled as an indirect moderator (via digital literacy, training access, resource constraints) rather than a direct causal determinant; policy evaluation must consider these mediating channels.
• Policy design:
  • Targeted reskilling: Subsidies and programmes should focus on digital literacy and transferable skills for vulnerable low‑skilled workers, with targeted outreach for older workers.
  • Firm incentives: Encourage employer investment in on‑the‑job training, job redesign that leverages human strengths, and welfare adjustments to smooth transitions.
  • Social protections & equity: Address distributional consequences (social fairness) through regulation, portable benefits, and active labour‑market policies aimed at re‑employment pathways.
  • Governance: Integrate data‑privacy, industry regulation, and human–AI collaboration standards into employment policy to align productivity gains with equitable outcomes.
• Research agenda:
  • Use representative and longitudinal data to quantify causal effects of task restructuring on employment trajectories.
  • Incorporate measures of adaptive capacity and institutional support into empirical models.
  • Compare sectoral and firm‑level heterogeneity in AI adoption and re‑employment pathways.
  • Evaluate cost‑effectiveness of targeted training and employer‑led re‑skilling interventions, with attention to age‑specific barriers and outcomes.

Reference (paper summarized): Liu, H. (2026). The Impact of Artificial Intelligence on Low‑Skilled Employment — Age Heterogeneity and Re‑employment Pathways. Journal of Economics and Management Sciences, 9(4). DOI: 10.30560/jems.v9n4p82.