Abstract
This literature synthesis examines whether reskilling programmes effectively reduce displacement risk for workers threatened by artificial intelligence and automation, or whether such initiatives primarily benefit already-advantaged workers. Drawing upon contemporary peer-reviewed research spanning multiple industries and national contexts, this study analyses the distributional outcomes of various reskilling approaches and identifies critical design features that determine programme equity. The findings reveal a nuanced picture: whilst proactive, well-targeted reskilling programmes demonstrably reduce displacement risk—with some organisations achieving 64% higher retention of displaced workers—generic market-led approaches tend to reinforce existing inequalities. High-skilled workers disproportionately access and complete training opportunities, whilst low-wage workers in automatable roles often lack training provisions despite facing the greatest displacement vulnerability. The analysis identifies three critical design choices that shape distributional outcomes: targeting vulnerable workers in automatable roles, embedding reskilling within social protection frameworks, and avoiding reliance on voluntary self-funded models. These findings carry significant implications for policymakers and organisations seeking to ensure that workforce transitions driven by technological change do not exacerbate existing socioeconomic inequalities.
Introduction
The rapid advancement of artificial intelligence and automation technologies presents one of the most significant labour market transformations since the Industrial Revolution. Across diverse sectors, from manufacturing to professional services, intelligent systems increasingly perform tasks previously requiring human cognitive and manual labour. This technological shift generates profound questions regarding workforce displacement, skills obsolescence, and the distribution of economic opportunities in digitally transformed economies.
Reskilling programmes have emerged as the predominant policy response to automation-induced displacement risk. Governments, employers, and international organisations alike champion workforce training as the mechanism through which displaced workers can transition to new roles created by technological change. The World Economic Forum (2020) estimated that 50% of all employees would require reskilling by 2025, whilst national governments have invested billions in workforce development initiatives. The implicit assumption underlying these investments holds that reskilling provides a pathway through which workers vulnerable to displacement can acquire competencies valued in emerging labour markets.
However, this optimistic narrative warrants critical scrutiny. Evidence increasingly suggests that reskilling programmes may not deliver equitable outcomes across the workforce. Workers already possessing educational advantages, financial resources, and employer support appear better positioned to access and complete training programmes. Conversely, those most vulnerable to displacement—typically low-wage workers in highly automatable roles—frequently lack access to meaningful reskilling opportunities. This pattern raises fundamental questions about whether reskilling functions as a genuine mechanism for mitigating displacement risk or primarily as an instrument that amplifies pre-existing workforce advantages.
This matter carries substantial academic and practical significance. Academically, understanding the distributional dynamics of reskilling programmes contributes to broader debates regarding technological change and inequality, skill-biased technological change theory, and the effectiveness of active labour market policies. Practically, as organisations and governments allocate substantial resources to workforce development, understanding which programme designs actually reduce displacement risk—rather than merely benefiting advantaged workers—becomes essential for evidence-based policymaking.
The stakes are considerable. Automation disproportionately threatens employment in routine cognitive and manual tasks concentrated among middle and lower-wage workers. If reskilling programmes fail to reach these populations, technological transitions may deepen existing inequalities, generating social and economic consequences extending far beyond individual displacement. Understanding the conditions under which reskilling programmes genuinely reduce displacement risk, rather than reinforcing advantage, represents a critical challenge for contemporary labour market policy.
Aim and objectives
The overarching aim of this study is to critically evaluate whether reskilling programmes reduce displacement risk for workers threatened by automation and artificial intelligence, or whether such programmes primarily benefit already-advantaged workers.
To achieve this aim, the following specific objectives guide the analysis:
1. To synthesise existing empirical evidence regarding the relationship between reskilling programmes and displacement risk reduction across diverse industries and national contexts.
2. To examine the distributional patterns of reskilling programme access and completion, identifying which workforce segments disproportionately benefit from current approaches.
3. To identify and analyse the specific programme design features that determine whether reskilling initiatives achieve equitable outcomes or reinforce existing advantages.
4. To develop evidence-based recommendations for policymakers and organisations seeking to design reskilling programmes that genuinely reduce displacement risk across the workforce rather than amplifying inequality.
Methodology
This study employs a systematic literature synthesis methodology to address the research aim and objectives. Literature synthesis represents an appropriate methodological approach when seeking to consolidate and critically analyse existing research findings to address specific research questions, particularly in rapidly evolving fields where primary data collection may lag behind emerging phenomena.
The literature search strategy prioritised peer-reviewed academic sources published within the past five years, reflecting the rapidly changing landscape of artificial intelligence capabilities and their labour market implications. Sources were identified through academic databases including Scopus, Web of Science, and Google Scholar, using search terms combining reskilling, upskilling, workforce training, automation, artificial intelligence, displacement, and inequality. Additional sources were identified through citation tracking and reference list examination of key articles.
Inclusion criteria specified English-language, peer-reviewed empirical studies or systematic reviews examining the relationship between reskilling programmes and workforce outcomes in the context of automation or artificial intelligence. Studies were required to provide evidence regarding either displacement risk reduction or distributional outcomes of reskilling initiatives. Grey literature from reputable international organisations and government bodies supplemented academic sources where appropriate.
The synthesis process involved systematic extraction of key findings, methodological approaches, and conclusions from included sources. Findings were organised thematically according to the research objectives, enabling identification of patterns regarding both displacement risk reduction and distributional equity. Critical analysis examined the consistency of findings across contexts, methodological limitations, and implications for policy and practice.
This methodological approach carries acknowledged limitations. Literature synthesis cannot generate new primary data and remains constrained by the quality and scope of existing research. Publication bias may affect the available evidence base, potentially over-representing studies with significant positive findings. Additionally, the rapidly evolving nature of artificial intelligence capabilities means that studies examining earlier automation waves may not fully predict outcomes from emerging technologies. These limitations are acknowledged whilst recognising that synthesis of current best available evidence provides essential guidance for policy and practice.
Literature review
### The complementary and substitutive effects of automation
Contemporary research consistently demonstrates that artificial intelligence and automation technologies produce heterogeneous effects across the workforce, generally complementing high-skill workers whilst substituting for low-skill workers. This pattern carries significant implications for understanding who faces displacement risk and who benefits from technological change.
Oladele, Orelaja and Hameed (2024) examined workforce resilience in the United States context, finding that automation tends to enhance the productivity and earning potential of workers with advanced education and technical skills whilst reducing demand for routine task-intensive occupations typically held by workers with lower formal qualifications. This complementarity-substitution dynamic means that technological change generates both winners and losers within the workforce, with the distribution of outcomes closely tracking pre-existing skill hierarchies.
Kong, Li and Song (2025) analysed artificial intelligence-driven job replacement specifically within the service industry, documenting how low-wage service workers face particularly acute displacement risk whilst often lacking access to training provisions that might facilitate transition to alternative roles. Their analysis emphasised that without targeted intervention, market-led responses to automation tend to widen rather than narrow workforce inequality.
Maria et al. (2025) contributed cross-national evidence examining both job displacement and job creation dynamics, confirming that whilst automation creates new roles, these positions typically require skill sets that displaced workers do not possess without substantial retraining. The gap between skills destroyed and skills demanded represents the theoretical justification for reskilling programmes, yet their analysis highlighted that bridging this gap proves far more challenging for some workforce segments than others.
Costa, Machado and Conceição (2024) examined these dynamics within the automotive industry, traditionally a bellwether sector for automation trends. Their case study research demonstrated that whilst automation eliminated certain roles, simultaneous investment in reskilling enabled workers to transition to newly created positions, avoiding net employment loss within the studied plants. However, they noted that this positive outcome required substantial organisational commitment and investment that may not be replicable across all contexts.
### Patterns of reskilling access and completion
A consistent finding across the literature concerns the unequal distribution of reskilling opportunities. High-skilled workers demonstrate systematically greater likelihood of accessing and successfully completing training programmes, suggesting that generic reskilling initiatives may reinforce rather than address workforce inequalities.
Kong, Li and Song (2025) documented how low-wage workers in the service industry often do not have training provisions available to them, creating a paradox whereby those facing greatest displacement risk receive least support. Employer-provided training tends to concentrate among workers already possessing valued skills, as organisations rationally invest training resources where expected returns are highest. This market logic produces systematically inequitable outcomes.
Morandini et al. (2023) examined organisational approaches to upskilling and reskilling, finding that worker skill development opportunities cluster among already-advantaged employees. Their analysis attributed this pattern to multiple factors including employer selection effects, worker self-selection based on confidence and prior educational experience, and structural barriers including time poverty and financial constraints that disproportionately affect lower-wage workers.
Oladele, Orelaja and Hameed (2024) extended this analysis to examine how reliance on voluntary, self-funded training courses particularly disadvantages workers lacking financial resources, time flexibility, or awareness of available opportunities. Workers in precarious employment situations face particular barriers, as training participation may conflict with immediate income-generating requirements.
Li (2022) examined reskilling and upskilling requirements for Industry 4.0 readiness, emphasising that whilst extensive training needs exist across the workforce, current provision structures favour workers already possessing foundational skills and learning capabilities. The self-directed nature of many digital learning platforms, whilst offering scalability, may disadvantage workers lacking prior experience with technology-mediated learning.
### Evidence for displacement risk reduction
Despite concerns regarding distributional equity, substantial evidence indicates that well-designed reskilling programmes can meaningfully reduce displacement risk for participating workers.
Kanagarla (2024) conducted multi-sector analysis of workforce disruption and adaptation, finding that organisations with proactive reskilling programmes achieved 64% higher retention of displaced workers compared to organisations with reactive or no programmes. This striking differential suggests that reskilling, when properly integrated into organisational transition planning, provides tangible protection against displacement. Critically, proactive approaches embedded reskilling within broader workforce planning rather than treating it as an afterthought following displacement.
Costa, Machado and Conceição (2024) provided case study evidence from the automotive industry demonstrating that heavy investment in reskilling allowed job destruction from automation to be offset by new role creation, resulting in no net employment loss within the studied plants. Workers who completed reskilling programmes transitioned successfully to roles requiring different competencies, avoiding displacement despite substantial technological change.
Comparative analyses across industries and countries consistently find lower AI-induced unemployment in contexts where governments and sectors invest substantially in reskilling alongside appropriate governance policies. Maria et al. (2025), Li (2022), Shengelia (2025), and Güngör (2025) all document this relationship, suggesting that societal-level investment in workforce development moderates the displacement effects of technological change.
Shengelia (2025) examined labour market dynamics and employment trends, emphasising that coordinated approaches combining reskilling investment with broader industrial and social policies generate superior workforce outcomes compared to isolated training initiatives. This finding suggests that reskilling effectiveness depends partly on the broader policy ecosystem within which programmes operate.
Güngör (2025) analysed productivity gains versus job displacement, concluding that whilst automation inevitably displaces some roles, the magnitude of net displacement reflects policy choices including reskilling investment levels. Societies may choose whether technological change primarily generates displacement or primarily generates productivity enhancement with maintained employment, with reskilling representing a key mechanism through which this choice is exercised.
### Design features determining distributional outcomes
The literature identifies specific programme design choices that determine whether reskilling initiatives achieve equitable outcomes or primarily benefit already-advantaged workers.
Targeting represents perhaps the most critical design feature. Kong, Li and Song (2025), Maria et al. (2025), Oladele, Orelaja and Hameed (2024), and Li (2022) all emphasise that programmes explicitly targeting workers in automatable, low-wage roles produce more equitable outcomes than universally available programmes. Universal availability, whilst superficially fair, tends to result in disproportionate take-up by workers already possessing advantages that facilitate programme access and completion.
Funding models significantly influence distributional outcomes. Reliance on voluntary, self-funded courses systematically skews benefits toward already-advantaged workers who possess financial resources to invest in their own development. Kong, Li and Song (2025), Oladele, Orelaja and Hameed (2024), and Morandini et al. (2023) document how financial barriers exclude precisely those workers facing greatest displacement risk from training opportunities.
Integration with social protection systems emerges as a critical mechanism for reaching vulnerable workers. Kong, Li and Song (2025), Oladele, Orelaja and Hameed, (2024), and Maria et al. (2025) find that embedding reskilling within regional development programmes and social protection frameworks substantially improves access among disadvantaged populations. Such integration addresses multiple barriers simultaneously, providing financial support, reducing opportunity costs, and connecting reskilling to broader transition assistance.
The Organisation for Economic Co-operation and Development (2019) has similarly emphasised that active labour market policies achieve superior outcomes when training components connect with income support, job search assistance, and regional economic development initiatives. Isolated training programmes, however well-designed, cannot address the multi-dimensional barriers facing workers in precarious circumstances.
Discussion
The synthesised evidence presents a nuanced picture regarding the capacity of reskilling programmes to reduce displacement risk. Reskilling demonstrably can reduce displacement risk, with compelling evidence that proactive, well-designed programmes achieve substantially superior workforce outcomes. Yet the evidence equally demonstrates that without careful attention to design features, reskilling primarily benefits already-advantaged workers, potentially widening rather than narrowing workforce inequality.
### Meeting the research objectives
Regarding the first objective—synthesising evidence on the relationship between reskilling and displacement risk reduction—the literature provides consistent support for the proposition that reskilling can materially reduce displacement risk when properly implemented. The finding that organisations with proactive reskilling programmes achieved 64% higher retention of displaced workers (Kanagarla, 2024) represents particularly compelling evidence. Similarly, case study evidence demonstrating that substantial reskilling investment enabled automotive plants to avoid net employment loss despite extensive automation (Costa, Machado and Conceição, 2024) confirms that reskilling provides genuine protection under favourable conditions.
The second objective—examining distributional patterns of programme access and completion—reveals troubling inequities. High-skilled workers systematically demonstrate greater access to and completion of training programmes across diverse contexts. Low-wage workers in highly automatable roles, despite facing greatest displacement risk, often lack meaningful training provisions (Kong, Li and Song, 2025). This distributional pattern suggests that market-led reskilling initiatives may reinforce rather than address structural workforce inequalities.
The third objective—identifying design features determining equitable outcomes—yields actionable findings for policy and practice. Three design choices emerge as particularly consequential: targeting workers in automatable, low-wage roles; avoiding reliance on voluntary, self-funded participation models; and embedding reskilling within broader social protection and regional development frameworks. These design features distinguish programmes that genuinely reduce displacement risk across the workforce from those that primarily benefit already-advantaged workers.
### Theoretical implications
These findings carry implications for theoretical understandings of skill-biased technological change and labour market adjustment. Standard human capital theory suggests that workers rationally invest in skills yielding positive returns, implying that training markets should efficiently allocate reskilling opportunities to workers who would benefit most. The evidence reviewed here challenges this assumption by documenting systematic market failures in training provision.
Workers facing greatest displacement risk frequently cannot access training due to financial constraints, time poverty, information asymmetries, and employer selection effects. These market failures mean that absent targeted intervention, reskilling provision concentrates among workers least in need of displacement protection. This pattern suggests that complementary policies addressing access barriers represent essential components of effective reskilling systems.
The evidence also suggests limitations of firm-level responses. Whilst some organisations demonstrate exemplary practice through proactive reskilling integrated with workforce planning, such approaches remain far from universal. Employer incentives may not align with workforce-wide equity objectives, particularly when training investments yield uncertain returns and workers may subsequently depart for other organisations. This observation supports arguments for public investment and coordination in reskilling provision rather than sole reliance on employer-led initiatives.
### Policy implications
The evidence base supports several policy implications for governments and organisations seeking to design reskilling programmes that genuinely reduce displacement risk rather than primarily benefiting advantaged workers.
First, targeting matters fundamentally. Universal programmes, whilst politically attractive, tend to produce regressive distributional outcomes. Resources should concentrate on workers in demonstrably automatable occupations with limited current access to training. Occupational forecasting can identify roles facing elevated displacement risk, enabling pre-emptive rather than reactive intervention.
Second, funding models determine access. Self-funded approaches exclude precisely those workers facing greatest need. Public funding, employer mandates, or hybrid approaches providing subsidised access for targeted populations represent necessary conditions for equitable outcomes. The opportunity costs of training participation—including foregone wages during training periods—warrant explicit recognition through income support mechanisms.
Third, integration with broader support systems enhances effectiveness. Isolated training programmes cannot address the multi-dimensional barriers facing vulnerable workers. Embedding reskilling within regional development initiatives, social protection systems, and employment services enables holistic support addressing financial, informational, and motivational barriers simultaneously.
Fourth, proactive timing generates superior outcomes. Evidence that proactive reskilling substantially outperforms reactive approaches (Kanagarla, 2024) suggests that early intervention before displacement occurs yields greatest benefit. Policies encouraging organisations to invest in reskilling prior to workforce restructuring, rather than as severance support following displacement, warrant consideration.
### Limitations and future research directions
This analysis carries several limitations requiring acknowledgement. The evidence base, whilst growing, remains constrained by methodological challenges inherent in evaluating complex interventions within dynamic labour markets. Establishing causal relationships between programme participation and displacement outcomes proves difficult given selection effects and confounding factors. Longitudinal studies tracking workers through technological transitions remain relatively rare.
Additionally, the rapidly evolving capabilities of artificial intelligence systems mean that historical evidence may imperfectly predict future dynamics. Large language models and generative artificial intelligence systems demonstrate capabilities extending into cognitive task domains previously considered automation-resistant. Whether existing evidence regarding reskilling effectiveness generalises to these emerging technologies remains uncertain.
Future research should prioritise rigorous evaluation of targeted reskilling programmes serving disadvantaged populations, enabling assessment of design features most conducive to equitable outcomes. Longitudinal studies examining long-term career trajectories following reskilling participation would strengthen understanding of sustained displacement protection versus temporary effects. Comparative analysis across national policy contexts could illuminate how broader institutional frameworks shape reskilling effectiveness.
Conclusions
This literature synthesis addressed whether reskilling programmes reduce displacement risk or primarily benefit already-advantaged workers. The evidence supports a conditional answer: reskilling can materially reduce displacement risk, but programme design critically determines whether benefits distribute equitably across the workforce or concentrate among already-advantaged workers.
The research objectives have been substantially achieved. Evidence confirms that proactive, well-designed reskilling programmes significantly reduce displacement risk, with some studies documenting 64% higher retention of displaced workers in organisations with proactive approaches. Simultaneously, evidence documents troubling distributional inequities, with high-skilled workers systematically accessing training whilst low-wage workers in automatable roles frequently lack meaningful provisions. Critical design features determining distributional outcomes include targeting of vulnerable workers, avoidance of self-funded models, and integration with social protection systems.
These findings carry significant implications for policy and practice. Governments and organisations investing in reskilling must attend carefully to programme design if equity objectives are to be achieved. Universal, market-led approaches tend to reinforce existing advantages; targeted, publicly supported, and integrated approaches offer superior prospects for genuinely reducing displacement risk across the workforce.
The broader significance extends to fundamental questions regarding technological change and social equity. Automation and artificial intelligence will continue transforming labour markets, creating both opportunities and displacement pressures. Whether these transitions widen inequality or support inclusive prosperity depends substantially on policy choices including reskilling programme design. Evidence-informed approaches can ensure that workforce development genuinely protects vulnerable workers rather than merely amplifying advantages enjoyed by those already well-positioned.
Future research should prioritise rigorous evaluation of targeted programmes, longitudinal tracking of reskilling participants, and comparative analysis across policy contexts. As artificial intelligence capabilities continue advancing, ongoing evidence generation regarding effective workforce development approaches becomes increasingly urgent. The stakes—determining whether technological transformation creates broad-based prosperity or deepened inequality—warrant sustained scholarly attention and evidence-based policymaking.
References
Costa, I., Machado, C. and Conceição, O. (2024) ‘Automation and the importance of reskilling workers: A case study in the automotive industry’, *European Conference on Management Leadership and Governance*. https://doi.org/10.34190/ecmlg.20.1.2923
Güngör, D. (2025) ‘The role of AI in transforming labor markets: Productivity gains versus job displacement’, *Next Generation Journal for The Young Researchers*. https://doi.org/10.62802/vk8h9g79
Kanagarla, K. (2024) ‘Artificial intelligence and employment transformation: A multi-sector analysis of workforce disruption and adaptation’, *International Journal of Scientific Research in Computer Science, Engineering and Information Technology*. https://doi.org/10.32628/cseit24106170
Kong, Z., Li, Z. and Song, Y. (2025) ‘Analysis of artificial intelligence-driven job replacement in the service industry and unemployment response strategies’, *Advances in Economics, Management and Political Sciences*. https://doi.org/10.54254/2754-1169/2025.bl23302
Li, L. (2022) ‘Reskilling and upskilling the future-ready workforce for Industry 4.0 and beyond’, *Information Systems Frontiers*, pp. 1–16. https://doi.org/10.1007/s10796-022-10308-y
Maria, S., Purwinahyu, P., Fitriansyah, F., Rachmawaty, A. and Aini, R. (2025) ‘Artificial intelligence and labor markets: Analyzing job displacement and creation’, *International Journal of Engineering, Science and Information Technology*, 5(2). https://doi.org/10.52088/ijesty.v5i2.830
Morandini, S., Fraboni, F., Angelis, M., Puzzo, G., Giusino, D. and Pietrantoni, L. (2023) ‘The impact of artificial intelligence on workers’ skills: Upskilling and reskilling in organisations’, *Informing Science: The International Journal of an Emerging Transdiscipline*, 26, pp. 39–68. https://doi.org/10.28945/5078
Oladele, I., Orelaja, A. and Hameed, A. (2024) ‘Artificial intelligence and automation: Creating a more resilient United States workforce’, *Journal of Scientific Papers “Social Development and Security”*, 14(4). https://doi.org/10.33445/sds.2024.14.4.7
Organisation for Economic Co-operation and Development (2019) *OECD employment outlook 2019: The future of work*. Paris: OECD Publishing. Available at: https://www.oecd.org/employment/outlook/
Shengelia, R. (2025) ‘Artificial intelligence and labor market dynamics: Employment problems and development trends’, *Economics*. https://doi.org/10.36962/ecs107/3-5/2025-07
World Economic Forum (2020) *The future of jobs report 2020*. Geneva: World Economic Forum. Available at: https://www.weforum.org/reports/the-future-of-jobs-report-2020
