AI-augmented HRM: Antecedents, assimilation and multilevel consequences

Highlights

• AI-Augmented HRM (HRM^(AI)) - Antecedents, assimilation and multilevel consequences.

• Multilevel conceptual framework of AI-Augmented HRM (HRM^(AI).

• HRM^(AI) framework based on theories of innovation assimilation and TOP framework.

• Research propositions for advancing scholarship in HRM^(AI) domain.

Abstract

The current literature on the use of disruptive innovative technologies, such as artificial intelligence (AI) for human resource management (HRM) function, lacks a theoretical basis for understanding. Further, the adoption and implementation of AI-augmented HRM, which holds promise for delivering several operational, relational and transformational benefits, is at best patchy and incomplete. Integrating the technology, organisation and people (TOP) framework with core elements of the theory of innovation assimilation and its impact on a range of AI-Augmented HRM outcomes, or what we refer to as (HRM^(AI)), this paper develops a coherent and integrated theoretical framework of HRM^(AI) assimilation. Such a framework is timely as several post-adoption challenges, such as the dark side of processual factors in innovation assimilation and system-level factors, which, if unattended, can lead to the opacity of AI applications, thereby affecting the success of any HRM^(AI). Our model proposes several testable future research propositions for advancing scholarship in this area. We conclude with implications for theory and practice.

Introduction

Scholarly research on the application of information technology (IT) in HRM is rapidly increasing, and different scholars have coined several concepts linking HRM and IT (Bondarouk, Parry, & Furtmueller, 2017; Strohmeier, 2018). These efforts include examples, such as ‘web-based HRM’ (Ruël, Bondarouk, & Looise, 2004), ‘e-HRM’ (Strohmeier & Kabst, 2009), ‘HRIS’ (Kundu & Kadian, 2012), ‘HRM cloud computing’ (Wang, Wang, Bi, Li, & Xu, 2016), ‘HR analytics’ (Vihari & Rao, 2013), ‘Online HRM’, ‘Digital HRM’, ‘Smart HRM’ (Bondarouk, Parry, & Furtmueller, 2017; Strohmeier, 2018) and ad-hoc and strategic use of artificial intelligence (AI) enabled HRM applications (Malik, Budhwar, Patel, & Srikanth, 2020; Malik, Srikanth, & Budhwar, 2020). Access to and generation of structured and unstructured HR-specific databases and an increasing reliance on the use of advanced digitalized HRM and AI applications for generating insights, solving problems and engaging in predictive decision-making for HR functions is on the rise (Saukkonen, Kreus, Obermayer, Ruiz, & Haaranen, 2019; Malik, Budhwar, & Srikanth, 2020; Strohmeier & Piazza, 2015; Tambe, Cappelli & Yakubovich). The proliferation and use of AI-enabled innovative database management is evident in the real world with emerging AI-HR applications and solutions, such as CloudHR, SAP SuccessFactors, BambooHR, GustoHR, OnPay, CakeHR, Trakstar, Deputy, ZohoPeople and so on. Technology giants also developed such applications, such as Google, Microsoft, IBM, and LinkedIn. With increased digitalization and data usage by AI-enabled applications for HRM processes, the role of HR professionals is also being redefined and transformed (Hmoud & Laszlo, 2019; Papageorgiou, 2018). Recent research notes that contemporary AI applications have augmentation and automation functionalities, and in some cases, replaced HRM decision-making that humans typically undertook (Malik, Budhwar, Patel, & Srikanth, 2020; Malik, Budhwar, & Srikanth, 2020; Vrontis et al., 2021).

In general, AI can be referred to as the capacity of machines to make predictions or solve problems using large amounts of data for complex, structured and unstructured environments (Agrawal, Gans, & Goldfarb, 2018). AI uses various techniques and applications, such as neural networks, speech/pattern recognition, genetic algorithms and deep learning (i.e., NLP- Neuro-linguistic Programming, machine learning and machine vision) (Jarrahi, 2018). The three components of AI: high-speed computation, use of big data and advanced algorithms, help AI differentiate from existing legacy IT applications (Cheng & Hackett, 2019; Guenole & Feinzig, 2018) for routine and non-routine decision-making and problem-solving processes (McGovern et al., 2018). The ability of AI applications to process large amounts of data fast assists HR managers in saving time and arrive at accurate HRM decisions (Lindebaum, Vesa, & Den Hond, 2020; Vrontis et al., 2021). There has been a rapid progression from a ‘descriptive and diagnostic’ to a ‘prescriptive and predictive’ approach to AI (DiClaudio, 2019; HRPA, 2017).

A critical review of current HRM literature suggests that AI-enabled HRM applications are becoming part of strategic HRM discussions (Upadhyay & Khandewal, 2018; Kwan, Hermawan, & Hafizhi, 2019) in almost all sub-functional HRM domains, such as recruitment and selection (Van Esch, Black, & Ferolie, 2019), training and development (Maity, 2019), performance management (Leicht-Deobald et al., 2019), talent management (Malik, De Silva, Budhwar, & Srikanth, 2021), employee turnover (Li, Li, Bonn, & Ye, 2019), reward management (Escolar-Jimenez, Matsuzaki, Okada, & Gustilo, 2019), job design (Huang, Rust, & Maksimovic, 2019), employee satisfaction (Nguyen & Malik, 2021) and employee engagement (Burnett & Lisk, 2019). Inherent in the use of AI applications in HRM are algorithms that augment HR practitioners' experience and HR decision-making and problem-solving processes. Despite the high adoption rate of various AI techniques (such as algorithms) in distinct functions of HRM, there is a limited understanding of an integrated and processual framework for the adoption of AI in HRM.

The existing conceptual frameworks and theoretical models around the use of AI in HRM reflect a comparatively cautious and conservative outlook (Cheng & Hackett, 2019; Tambe, Cappelli, & Yakubovich, 2019). Perhaps, Strohmeier and Piazza (2015) were the first to develop a conceptual base for identifying a range of areas for using AI-augmented applications for HRM, followed by Jia et al. (2018), who highlighted several strategic and operational aspects of HRM AI applications that could assist in these activities. However, although both of these studies offered a nuanced approach for using AI applications for different HRM functions, these studies lack an explanation of the theoretical basis for AI assimilation in HRM. Moreover, prior literature concerning technology-driven HRM (i.e., HRIS, e-HRM) considers adoption as a single point event (see, e.g., Bondarouk, Parry, & Furtmueller, 2017; Burbach & Royle, 2014; Troshani, Jerram, & Hill, 2011), yet adoption is only a part of the assimilation process. The migration from initial adoption to the diffusion of technological innovation is complex and involves additional steps of diffusion, routinization and extension of assimilation (Chatterjee, Grewal, & Sambamurthy, 2002; Zhu, Kraemer, & Xu, 2006). Such an approach is evident in technological innovations from different businesses, as noted in Hossain, Quaddus and Islam's (2016) four-stage process model explaining adoption, diffusion, routinization, extension for RFID- Radio Frequency Identification (RFID) technology assimilation. Similarly, Basole & Nowak (2016) and Nam, Lee, and Lee (2019) used a three-stage process for supply chain adoption to examine drivers for each stage of business analytics adoption. Therefore, a processual conceptual framework comprising a stage-based process to understand the assimilation of AI in HRM is much warranted.

Thus, observing these limitations and subsequent call for comprehensive AI-HRM adoption framework incorporating interdisciplinary collaboration between HRM and other functional areas (Fountaine, McCarthy, & Saleh, 2019; Kiron & Schrage, 2019), this research proposes a new term ‘AI-Augmented HRM’ (HRM^(AI)), and a processual HRM^(AI) assimilation framework comprising its antecedents and outcomes. Augmentation in the context of HRM^(AI) implies a mindset of symbiotic enhancement (Davenport & Kirby, 2016) to develop AI-enabled HRM interactive systems, using a collaborative partnership between humans and machines to deliver productivity improvements, process efficiency, and human interactions at the interface of AI-human exchanges that augments human-machine cognition (Wilson & Daugherty, 2018).

Thus based on these observations, this research aims first to develop comprehensive and contemporary definitions of HRM^(AI) (AI-augmented HRM) and HRM^(AI) assimilation to highlight and explain the use of AI techniques in the context of HRM. These fine-grained definitions will help to distinguish HRM^(AI) from these similar-sounding concepts (i.e. web-based HRM, e-HRM, HRIS, HRM cloud computing, HR analytics, Online HRM, Digital HRM and Smart HRM), and at the same time will allow several research questions – at multiple levels – to be generated and extended in future HRM^(AI) research. Secondly, given the observation that migration of AI adoption for HRM to diffusion might involve different stages (Chatterjee et al., 2002; Zhu et al., 2006), we propose a four stage process-view approach for HRM^(AI) assimilation based on theory of assimilation innovation. Thirdly, while considering HRM^(AI) assimilation as a full life-cycle that includes not only an organisations' initial evaluation of AI and its adoption but its full-scale deployment (i.e., stages of HRM^(AI) process), this research aims to identify antecedents and consequences of HRM^(AI) assimilation and provide future research propositions based on identified antecedents and consequences.

The essence of this research can be captured in three main contributions to the existing literature in the domain of adoption of AI in HRM. First, this research introduces the concept of HRM^(AI) to differentiate use of AI in HRM from similar technology-enabled HRM terms. Second, this research proposes an HRM^(AI) assimilation mechanism based on a four-stage process comprising the stages of initiation, adoption, routinization and extension in line with the theory of assimilation innovation. Last but not least, despite the current trend in an increased evaluation of research regarding the use of AI for a range of HRM activities and benefits linked to the adoption of AI technology in HRM (Charlier & Kloppenburg, 2017; DiClaudio, 2019; Gulliford & Dixon, 2019; Hmoud & Laszlo, 2019; HRPA, 2017; Marr, 2018; McGovern et al., 2018; Robinson, Sparrow, Clegg & Birdi, 2007; Rodney, Valaskova, & Durana, 2019; Samarasinghe & Medis, 2020; Saukkonen et al., 2019; Tambe et al., 2019), the extant efforts are limited in terms of development of a comprehensive framework that provides an understanding of antecedents and potential consequences of AI adoption in HRM. In order to conceptualise an integrated processual HRM^(AI) assimilation framework, we identify antecedents for HRM^(AI) assimilation based on Technology-Organisation-People (TOP) framework (Brandt & Hartmann, 1999), and then further use operational, relational and transformational consequences framework (Bissola & Imperatori, 2020; Lepak & Snell, 1998; Obeidat, 2016; Parry & Tyson, 2011; Ruël et al., 2004; Strohmeier, 2007) for proposing multilevel consequences of HRM^(AI) assimilation.

The paper first offers a theoretical background informing the study, followed by definitions of the newly conceptualised terms HRM^(AI) and HRM^(AI) assimilation. Next, a detailed theoretical model and future research propositions for the antecedents and consequences of HRM^(AI) assimilation. Finally, the paper highlights the critical theoretical and practical contributions, limitations, and future research possibilities.