Augmentative or autonomous? Reframing artificial intelligence in talent management through a systematic review Open Access

Artificial intelligence (AI) is rapidly transforming how organizations attract, develop and retain talent, raising profound questions for management scholarship. While management research has long acknowledged the role of digital technologies in shaping organizational practices (Brynjolfsson & McAfee, 2017; George, Osinga, Lavie, & Scott, 2016), the specific integration of AI into talent management (TM) remains conceptually fragmented. Studies have examined algorithmic recruitment (Chamorro-Premuzic, Winsborough, Sherman, & Hogan, 2016), predictive retention models (Davenport, Guha, Grewal, & Bressgott, 2020) and AI-enabled performance evaluation (Tursunbayeva, Franco, & Pagliari, 2019), yet these lines of inquiry often proceed in isolation, grounded in disparate theoretical traditions.

This fragmentation reflects a broader issue in management scholarship: technology-focused studies tend to prioritize either micro-level behavioral implications or macro-level strategic outcomes, rarely integrating the two (Aguinis, Ramani, & Alabduljader, 2018; Raisch & Krakowski, 2021). In the case of AI in TM, the absence of an integrative framework leaves scholars and practitioners without a clear conceptual map for understanding how different forms of AI reconfigure the foundations of talent systems. Without such integration, the literature risks becoming descriptive rather than theory generative.

The present study addresses this gap by conducting a systematic review of peer-reviewed research at the intersection of AI and TM published between 2000 and early 2025. Beyond mapping the field, we propose a conceptual distinction between augmentative AI (technologies that enhance human judgment) and autonomous AI (systems that supplant human decision-making). This distinction provides a lens for synthesizing diverse strands of research, clarifying how AI applications in TM range from decision-support analytics to fully automated selection and development tools. Building on calls for theory-driven engagement with technological phenomena in management research (Haenlein & Kaplan, 2019; von Krogh, 2018; Theodorsson, 2025), our framework positions AI not only as an operational tool but as a transformative actor with implications for organizational strategy, employee experience and the future of work.

By systematically organizing two decades of scholarship and introducing an integrative conceptual framework, this paper makes three contributions. First, it identifies dominant thematic domains in the literature on AI and TM, including recruitment, development, retention and performance management, while revealing critical blind spots such as ethics, bias and cross-cultural variation. Second, it synthesizes these domains through the augmentative–autonomous distinction, offering a theoretical structure to guide future empirical and conceptual work. Third, it outlines a research agenda that highlights underexplored issues of fairness, accountability and the long-term dynamics of human–AI collaboration in talent systems. Taken together, these contributions aim to advance management theory by clarifying the conceptual terrain of AI-enabled TM and by offering a foundation for cumulative knowledge development.

Although research on AI in organizations has accelerated in recent years, it has not yet coalesced into a unified body of knowledge, particularly in the domain of TM. Existing studies tend to address isolated practices or narrow functional outcomes, producing a literature that is fragmented across fields and uneven in its theoretical grounding. To situate our contribution, this section reviews the foundations of AI in management research, the evolution of TM as a distinct scholarly domain and the points of intersection where AI has begun to reshape TM. Together, these strands highlight both the promise of AI-enabled talent systems and the persistent lack of theoretical integration that motivates our framework.

AI has become an increasingly prominent focus in management research, both as a technological phenomenon and as an analytical tool. Yet, much of this scholarship remains scattered across strategy, information systems and organizational behavior, with limited synthesis.

AI has increasingly attracted attention in management and organization studies, reflecting its transformative impact on strategy, operations and human resource (HR) practices. Early work on digitalization emphasized efficiency gains and data-driven decision-making (Brynjolfsson & McAfee, 2017), while more recent contributions highlight AI’s potential to alter managerial cognition, organizational routines and institutional logics (Haenlein & Kaplan, 2019; Raisch & Krakowski, 2021). Within organizational behavior, scholars have examined AI as a phenomenon that challenges established paradigms, calling for integration of data-driven insights with theory-driven research (von Krogh, 2018; George et al., 2016). Despite this momentum, AI-related research remains fragmented across subfields, with studies variously anchored in information systems, operations research or strategic management, but rarely synthesized into a coherent conceptual framework for TM.

TM has matured as a distinct field within human resource management, yet conceptual ambiguity and theoretical fragmentation persist. This provides an important backdrop for understanding how AI might reshape core assumptions about talent.

TM has traditionally been conceptualized through the lenses of strategic human resource management (Collings & Mellahi, 2009), human capital theory (Ployhart & Moliterno, 2011) and resource-based perspectives of competitive advantage (Barney, 1991). Research emphasizes the identification, development and retention of key individuals whose skills and knowledge disproportionately shape organizational outcomes (Tymon, Stumpf, & Doh, 2010). While TM scholarship has matured into a distinct field, it has been criticized for conceptual ambiguity (Gallardo-Gallardo & Thunnissen, 2016; Theodorsson, 2024) and for privileging normative models over theory-driven explanations (King, 2015). The infusion of AI into TM offers opportunities to address these limitations by introducing new forms of data, analytics and algorithmic decision-making. However, without theoretical integration, the risk is that AI applications will be treated as isolated tools rather than as forces reshaping the foundations of TM.

Despite a surge of studies on AI-enabled HR practices, research remains largely practice-driven and domain-specific. As a result, the field lacks cumulative insights or shared conceptual frameworks.

Empirical studies of AI in TM have proliferated across specific domains. In recruitment and selection, AI-enabled tools for résumé screening, psychometric analysis and predictive hiring have been widely documented (Chamorro-Premuzic et al., 2016; Tambe, Cappelli, & Yakubovich, 2019). In learning and development, adaptive learning platforms and algorithmic coaching are increasingly studied (Bessen, 2019). Research on performance management and retention highlights predictive analytics for turnover and AI-driven appraisal systems (Davenport et al., 2020; Tursunbayeva et al., 2019). While these contributions illustrate the breadth of AI’s application, they remain dispersed across journals and disciplines, often grounded in practice-driven narratives rather than robust theory.

The central challenge is therefore not a lack of research activity, but the absence of an integrative framework that connects these disparate strands. Addressing this gap requires conceptual clarity that bridges micro- and macro-level perspectives.

The lack of integration between AI and TM research reflects broader tensions in management scholarship. Scholars have noted the persistent divide between micro-level analyses of individual behavior and macro-level studies of organizational strategy (Aguinis et al., 2018). In the context of AI and TM, this divide manifests as either a focus on employee attitudes toward AI tools (e.g. fairness, bias, acceptance) or on strategic outcomes such as efficiency and performance. Few studies explicitly connect these levels, leaving a gap in understanding how AI both enables and constrains talent systems in practice. Bridging this divide requires conceptual frameworks that can synthesize disparate findings, clarify mechanisms and provide direction for future research.

This study responds to that need by advancing an augmentative–autonomous AI framework, which distinguishes between systems designed to enhance human decision-making and those intended to operate independently of human judgment. By situating this framework within the established foundations of TM and AI research, the paper contributes to clarifying how different modes of AI adoption reshape talent processes, with implications for theory, practice and the future of work.

This review follows the Preferred Reporting Item for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines and established protocols for management and organizational research (Tranfield, Denyer, & Smart, 2003; Snyder, 2019). The objective was to synthesize the state of knowledge on AI in TM and to develop a conceptual framework that integrates fragmented insights across disciplines.

To ensure breadth and quality of coverage, searches were conducted in Scopus, Web of Science, EBSCO Business Source and ProQuest ABI/Inform, four of the most comprehensive databases for peer-reviewed management, information systems and organizational research (Aguinis et al., 2018). Boolean operators were used to capture combinations of terms relating to AI (e.g. “artificial intelligence,” “machine learning,” “algorithmic decision-making”) and TM (e.g. “talent management,” “recruitment,” “performance management,” “workforce analytics,” “employee retention”). The full database-specific search strings are reported in  Appendix 2.

Searches were executed in September 2025 and results were limited to English-language, peer-reviewed journal articles published between January 2000 and September 2025. To enhance coverage, forward and backward citation tracking was also conducted for key studies to identify additional relevant literature not captured in the initial database searches.

Studies were considered eligible if they were published in peer-reviewed academic journals, addressed the application or implications of AI within TM or human resource management (HRM) and offered either empirical, theoretical or conceptual insights relevant to organizational contexts. Conference proceedings, editorials, dissertations and studies that mentioned AI or TM only incidentally were excluded. A summary of exclusion reasons is provided in  Appendix 1.

The expanded search across all four databases initially yielded 238 records. After removing duplicates, 212 unique records remained. Titles and abstracts were screened for relevance, leading to the exclusion of 32 studies. The remaining 180 articles underwent full-text review, during which 56 were excluded for reasons such as limited conceptual depth, absence of a clear organizational context or minimal relevance to AI–TM intersections. This process produced a final data set of 124 articles, which formed the basis for analysis (see Figure 1, PRISMA 2020 flow diagram).

The addition of two databases generated a small number of new records but did not materially alter the final corpus, confirming the comprehensiveness of the original search. Screening and coding were conducted independently by both authors, with strong inter-rater reliability (Cohen’s κ = 0.82). Discrepancies were discussed until consensus was achieved.

Each article was coded for publication year, journal outlet, methodological design, theoretical framing and TM domain (recruitment, development, retention and performance management). A hybrid approach combining deductive and inductive coding (Miles, Huberman, & Saldaña, 2014) was applied. Deductive codes reflected established TM domains, while inductive codes captured emerging themes such as ethical AI, cross-cultural variation and human–AI collaboration.

The resulting data set supported both descriptive mapping (publication trends, methodological distributions and geographic focus) and conceptual synthesis. The analysis informed the development of the augmentative–autonomous AI framework, which clarifies distinct modes of AI adoption in TM and their implications for fairness, efficiency and trust.

The review identified 124 peer-reviewed studies published between 2000 and 2025. Together, these studies cover the main domains of TM: recruitment and selection, learning and development, performance management and retention. Across this body of work, two distinct patterns of AI adoption emerge: augmentative systems that support human judgment and autonomous systems that replace it. These patterns structure how organizations are integrating AI into talent practices and are captured in the augmentative–autonomous framework shown in Figure 2.

The analysis of the 124 studies points to four domains where AI is reshaping TM: recruitment and selection, learning and development, performance management and retention and ethical and cross-cultural considerations. Each domain illustrates how augmentative and autonomous systems take shape in practice and the different consequences they have for managerial judgment, fairness and trust.

Recruitment has been the most studied area of AI in TM. Augmentative approaches include algorithmic résumé screening and psychometric analytics that assist recruiters (Tambe et al., 2019; Chamorro-Premuzic et al., 2016). Autonomous applications include interview bots and predictive hiring models that operate with minimal human oversight (Langer, König, & Krause, 2021). As shown in Figure 2, augmentative systems are positioned as enhancing human judgment, while autonomous systems promise scalability but heighten concerns about bias and fairness (Bogen & Rieke, 2018).

AI has been deployed to tailor developmental experiences. Augmentative applications include adaptive learning platforms that provide individualized training recommendations while managers retain oversight (Bessen, 2019). Autonomous approaches include AI-driven coaching systems that independently generate developmental feedback and career advice. Figure 2 illustrates how augmentative systems tend to support managerial roles, while autonomous systems offer efficiency at the potential cost of diminishing human involvement in development (Meijerink, Bondarouk, & Lepak, 2021).

In performance management, augmentative AI is often used in dashboards and analytics tools that visualize trends to inform managerial decisions (Davenport et al., 2020). Autonomous approaches include algorithmic scoring systems and predictive attrition models that generate evaluations or risk profiles without direct human intervention (Tursunbayeva et al., 2019). As Figure 2 depicts, augmentative approaches can reinforce employee trust by preserving the human role in appraisal, while autonomous approaches raise challenges related to fairness and employee voice (Raisch & Krakowski, 2021).

A cross-cutting theme across all domains concerns ethics, fairness and contextual variation. Studies highlight risks of bias embedded in training data sets (Caliskan, Bryson, & Narayanan, 2017), as well as concerns over opacity and accountability in algorithmic decisions (Lepri, Oliver, Letouzé, Pentland, & Vinck, 2018). Cultural context further shapes the acceptance and impact of AI in TM, with Western-centric studies dominating the field and little attention paid to non-Western perspectives (Budhwar et al., 2023). These gaps underscore the importance of situating AI adoption within broader institutional, ethical and cultural frameworks.

Figure 2 integrates these findings into a conceptual framework that distinguishes between augmentative and autonomous logics of AI adoption in TM. Across domains, augmentative systems are associated with outcomes such as enhanced judgment, managerial support and employee trust, whereas autonomous systems are linked to automated decisions, efficiency and scale and risks of bias and fairness concerns. By clarifying these distinct pathways, the framework provides a foundation for theory development and for guiding practitioners in balancing efficiency with fairness and trust when adopting AI in TM.

The findings of this review highlight that the adoption of AI in TM is characterized by two distinct logics, augmentative and autonomous, that cut across functional domains. This distinction provides an integrative lens that addresses the fragmentation of the existing literature and clarifies pathways for theoretical and practical advancement.

The framework advances TM scholarships in three ways. First, it positions AI not as a monolithic phenomenon but as a dual-logic system with distinct implications for decision-making, fairness and strategic outcomes. This perspective builds on prior calls to theorize digitalization as a heterogeneous set of technologies with varied organizational effects (Raisch & Krakowski, 2021; von Krogh, 2018). Second, by integrating micro-level concerns about fairness and employee trust with macro-level outcomes such as efficiency and scalability, the framework responds to the persistent divide in management research between individual and organizational-level analyses (Aguinis et al., 2018). Third, the framework provides a conceptual anchor for advancing theory in TM, which has often been criticized for descriptive utility rather than cumulative development (Gallardo-Gallardo & Thunnissen, 2016; King, 2015). In this sense, the augmentative–autonomous distinction contributes to building a more theoretically robust foundation for studying AI-enabled HR systems.

For practitioners, the framework offers a tool for evaluating AI adoption choices in TM. Augmentative systems can be leveraged to enhance managerial judgment, provide transparency and build employee trust, although they may be constrained by limits to scalability. Autonomous systems promise efficiency and predictive power, yet they heighten risks of bias, opacity and reduced employee agency. By clarifying these tradeoffs, the framework equips HR leaders and policymakers with a structured approach to aligning AI adoption with organizational values and strategic priorities. In practice, organizations may consider hybrid strategies, deploying augmentative AI in domains where trust and transparency are paramount, such as performance evaluation, while experimenting with autonomous AI where efficiency gains are critical, such as initial screening in high-volume recruitment.

This review also underscores the need for a more cumulative and theoretically grounded research agenda. Three priorities emerge. First, ethical AI and fairness require systematic theorization beyond case-specific observations, particularly regarding accountability mechanisms and employee perceptions. Second, cross-cultural perspectives remain underexplored, with current scholarship dominated by studies in Western and Asian contexts. Comparative research is needed to assess how institutional environments shape the acceptance and consequences of AI in TM (Budhwar et al., 2023). Third, longitudinal and interdisciplinary approaches are essential for understanding how AI adoption evolves over time and interacts with existing TM systems. Integrating insights into information systems, organizational behavior and strategic HRM will be critical for advancing theory and practice. The review adhered to PRISMA 2020 reporting standards to ensure transparency and replicability.

This review highlights not only the growing body of work on AI in TM but also the critical gaps that remain. Building on the augmentative–autonomous framework, we identify four avenues for future research that can move the field toward greater theoretical coherence and practical relevance.

Ethics has emerged as one of the most pressing concerns in the integration of AI into TM. Existing studies raise questions of bias embedded in data sets (Caliskan et al., 2017) and the opacity of algorithmic decision-making (Lepri et al., 2018), yet conceptual engagement remains limited. Future research should examine how augmentative and autonomous systems differentially affect perceptions of fairness, accountability and trust. Scholars could, for example, draw on justice theory (Colquitt, Conlon, Wesson, Porter, & Ng, 2001) to understand employee reactions to AI-enabled evaluations or on institutional theory to analyze how organizations construct accountability structures around algorithmic systems. Comparative studies of governance models across industries and jurisdictions would further advance knowledge on how ethical risks are managed in practice.

Most empirical studies of AI in TM are concentrated in the USA, China and India, leaving large parts of the world underrepresented (Budhwar et al., 2023). This imbalance limits the generalizability of findings and neglects the influence of institutional environments on AI adoption. Future research should investigate how cultural values, labor market institutions and regulatory regimes shape both the acceptance of AI tools and their organizational consequences. Cross-national comparative studies, especially in underexplored contexts such as Africa, Latin America and the Middle East, would enrich understanding of how augmentative and autonomous AI interact with local practices and norms. Such studies could also shed light on whether certain adoption logics are culturally contingent or universally applicable.

Much of the current literature is cross-sectional, offering snapshots of AI adoption without considering its evolution over time. This limits our ability to understand how augmentative and autonomous systems mature, interact or shift in organizational practice. Longitudinal research designs are needed to capture the dynamics of adoption, routinization and employee adaptation. Such studies could explore, for instance, whether augmentative systems tend to evolve into more autonomous configurations or whether hybrid models stabilize as sustainable solutions. Incorporating a temporal perspective would also allow scholars to examine unintended consequences, such as trust erosion or institutional resistance, that only become apparent after extended use.

Finally, advancing theory on AI in TM requires drawing insights from multiple fields. Information systems research provides tools for understanding algorithmic design and user acceptance (Venkatesh, Morris, Davis, & Davis, 2003), while organizational behavior offers perspectives on motivation, trust and fairness (Rousseau, 1995). Strategic HRM contributes frameworks for aligning technology with organizational goals (Collings, Mellahi, & Cascio, 2019). Yet these strands often operate in isolation. Future research should explicitly integrate theories across disciplines, using the augmentative–autonomous distinction as a boundary object that facilitates cross-disciplinary dialogue. For example, combining resource-based views with socio-technical systems theory could generate richer explanations of how AI reshapes the value of human capital in organizations.

This paper has synthesized two decades of research on the intersection of AI and TM through a systematic literature review. By analyzing 124 peer-reviewed studies, we identified both the breadth of AI applications across TM domains and the conceptual fragmentation that has limited theoretical progress. To address this gap, we introduced the augmentative– autonomous AI framework, which distinguishes between systems that support human decision-making and those that substitute for it.

The framework clarifies how different logics of AI adoption shape outcomes across recruitment, development, performance management and retention, as well as ethical and cross-cultural considerations. It also provides a conceptual anchor that links micro-level concerns about fairness and trust with macro-level questions of efficiency, scalability and strategic impact.

Our analysis makes three contributions. First, it offers one of the most comprehensive reviews to date of AI in TM, mapping the field and identifying underexplored themes. Second, it advances theory by introducing a dual-logic framework that integrates disparate findings and highlights tradeoffs between augmentative and autonomous adoption. Third, it establishes a research agenda that prioritizes ethical theorization, cross-cultural inquiry, longitudinal approaches and interdisciplinary integration.

Taken together, these contributions provide both scholars and practitioners with a foundation for understanding and navigating the complex role of AI in TM. As organizations continue to experiment with and expand AI adoption, the need for theoretically grounded, ethically informed and contextually sensitive research is both urgent and enduring.

To enhance transparency and replicability, this appendix reports the exact search strings used in each database.

SCOPUS

[TITLE-ABS-KEY(“artificial intelligence” OR “AI” OR “machine learning” OR “algorithmic decision-making”)]

AND

[TITLE-ABS-KEY(“talent management” OR recruitment OR selection OR “performance management” OR “employee retention” OR “workforce analytics” OR “learning and development”)]

Filters applied:

– Document type: Article

– Language: English

– Publication years: 2000–2025

Web of Science

TS = (“artificial intelligence” OR “AI” OR “machine learning” OR “algorithmic decision-making”)

AND

TS = (“talent management” OR recruitment OR selection OR “performance management” OR “employee retention” OR “workforce analytics” OR “learning and development”)

Filters applied:

– Document type: Article

– Language: English

– Timespan: 2000–2025

EBSCO Business Source

(“artificial intelligence” OR “AI” OR “machine learning” OR “algorithmic decision-making”)

AND

(“talent management” OR recruitment OR selection OR “performance management” OR “employee retention” OR “workforce analytics” OR “learning and development”)

Filters applied:

– Scholarly (peer-reviewed) journals

– English language

– Published 2000–2025

ProQuest ABI/Inform

(“artificial intelligence” OR “AI” OR “machine learning” OR “algorithmic decision-making”)

AND

(“talent management” OR recruitment OR selection OR “performance management” OR “employee retention” OR “workforce analytics” OR “learning and development”)

Filters applied:

– Peer-reviewed journals

– English language

– Publication date: 2000–2025