Artificial Intelligence-Driven Risk Management for Supply Chain Agility PDF

International Journal of Production Research ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/tprs20 Artiﬁcial intelligence-driven risk management for enhancing supply chain agility: A deep-learning- based dual-stage PLS-SEM-ANN analysis Lai-Wan Wong, Garry Wei-Han Tan, Keng-Boon Ooi, Binshan Lin & Yogesh K. Dwivedi To cite this article: Lai-Wan Wong, Garry Wei-Han Tan, Keng-Boon Ooi, Binshan Lin & Yogesh K. Dwivedi (2024) Artiﬁcial intelligence-driven risk management for enhancing supply chain agility: A deep-learning-based dual-stage PLS-SEM-ANN analysis, International Journal of Production Research, 62:15, 5535-5555, DOI: 10.1080/00207543.2022.2063089 To link to this article: https://doi.org/10.1080/00207543.2022.2063089 © 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 23 May 2022. Submit your article to this journal Article views: 21950 View related articles View Crossmark data Citing articles: 75 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=tprs20 INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 2024, VOL. 62, NO. 15, 5535–5555 https://doi.org/10.1080/00207543.2022.2063089 RESEARCH ARTICLE Artificial intelligence-driven risk management for enhancing supply chain agility: A deep-learning-based dual-stage PLS-SEM-ANN analysis Lai-Wan Wong a , Garry Wei-Han Tan b,c , Keng-Boon Ooi b,c,d , Binshan Lin e and Yogesh K. Dwivedi f ,g a School of Computing and Data Science, Xiamen University Malaysia, Sepang, Malaysia; b UCSI Graduate Business School, UCSI University, Kuala Lumpur, Malaysia; c Nanchang Institute of Technology, Jiangxi, People’s Republic of China; d College of Management, Chang Jung Christian University, Tainan City, Taiwan; e College of Business, Louisiana State University, Shreveport, LA, USA; f Emerging Markets Research Centre (EMaRC), School of Management, Swansea University – Bay Campus, Swansea, UK; g Department of Management, Symbiosis Institute of Business Management, Pune & Symbiosis International (Deemed University), Pune, Maharashtra, India ABSTRACT ARTICLE HISTORY This study posits that the use of artificial intelligence (AI) enables supply chains (SCs) to dynamically Received 15 March 2021 react to volatile environments, and alleviate potentially costly decision-makings for small-medium Accepted 30 January 2022 enterprises (SMEs). Building on a resource-based view, this work examines the impact of AI on SC KEYWORDS risk management for SMEs. A structural model comprising of AI-risk management capabilities, SC re- Supply chain agility; engineering capabilities and supply chain agility (SCA) was developed and tested based on data col- re-engineering capabilities; lected from executives, managers and senior managers of SMEs The main methodological approach risk management; artiﬁcial used in this study is partial least squares-based structural equation modelling (PLS-SEM) and arti- intelligence; ANN; PLS-SEM ficial neural network (ANN). The results identified the use of AI for risk management influences SC re-engineering capabilities and agility. Re-engineering capabilities further affect and mediate agility. PLS-SEM and ANN were compared and the results revealed consistency for models A and B. Cur- rent levels of demand uncertainties in the SC challenges managers in making complex trade-off decisions that require huge management resources in very limited time. With AI, it is possible to model various scenarios to answer crucial questions that archaic infrastructures are not able to. This study combines a multi-construct agility concept and identified non-linear relationships in the model. 1. Introduction must oversee the sustainability of their SCs (Carter, Kauf- In the context of the supply chain (SC), agility can refer mann, and Ketchen 2020). Despite having received vast to the firm’s ability to (1) calibrate tactics and opera- attention among scholars, there does not exist a univer- tions within its SC in response or adapt to fluctuations, sally accepted concept for supply chain agility (SCA) (D. opportunities or environmental threats (D. M. Gligor, M. Gligor, Holcomb, and Stank 2013; D. Gligor et al. Holcomb, and Stank 2013); (2) respond to short-term 2019; D. Z. Zhang 2011). The theoretical understand- market fluctuations (Aslam et al. 2018); (3) and exploit ing of SCA is fragmented due to its broad and multi- opportunities while addressing risks through market sen- dimensional concept that spans multiple disciplines. Ear- sitivity, network-based flexibility and process integration lier research concentrated on firms’ ability to succeed in (Brusset 2016). These fluctuations include demand pat- an environment of continuous and uncertainties. This terns change such as quality, quantity and variety and concept evolved into a paradigmatic view of firms’ capa- supply patterns change like shortages and disruptions bility to respond to customers’ dynamic demands and (Blome, Schoenherr, and Rexhausen 2013). According has expanded to multiple business challenges of turbu- to Choudhary and Sangwan (2018), managers are pres- lent environments. Other researchers have defined agility sured to instil agility to ensure sustained SC performance. in terms of a network of different integrated companies Improved SC performance is crucial as competition has to streamline material, information and financial flow shifted to SCs (Abdallah, Abdullah, and Mahmoud Saleh (Costantino et al. 2011). The focus then was on flexi- 2017; D. Gligor et al. 2019; Queiroz et al. 2021) and firms bility and performance while Braunscheidel and Suresh CONTACT Yogesh K. Dwivedi [email protected] Emerging Markets Research Centre (EMaRC), School of Management, Swansea University – Bay Campus, Room #323, Swansea SA1 8EN, UK © 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 5536 L.-W. WONG ET AL. (2009) defined agility as a multi-dimensional construct of The use of Industry 4.0 technologies such as artifi- demand response, joint planning, customer responsive- cial intelligence (AI), big-data analytics, cloud comput- ness and visibility. ing, distributed ledgers can bring about SC efficiencies However, a firm’s ability to acquire timely, relevant and in many ways (Hastig and Sodhi 2020; Sundarakani, up-to-date information can ensure enhanced SC visibil- Ajaykumar, and Gunasekaran 2021; Akter et al. 2020; ity, agility, performance and competitiveness (Baah et al. Belhadi, Mani, et al. 2021). For example, AI enables 2021). On the contrary, poor information sharing creates predictive approaches for risk assessment and minimi- conflicts and can lead to inferior performance (Singh, sation of disruptive events throughout the SC (Riahi Acharya, and Modgil 2020). A flexible SC needs to be et al. 2021), develops models to enable managers to dis- able to sense external stimuli and decide on the appro- cover improvement areas (Ni, Xiao, and Lim 2019) and priate response in a timely manner. And to carry out in the case of Alibaba (2020), a digitalised cloud-based these, the SC relies on the presence of the right metrics production SC powered by AI technologies can allow to inform stimulus, availability of strategic responses and SMEs to identify new opportunities in the shift of cus- scalability of resources and to measure the effectiveness of tomer preference and predict risks through data analytics responses. Information sharing thus is essential to inform capabilities. For this to work, two main components are SC coordination activities for overcoming SC dynam- essential: live data and feedback loop (Leavy, 2019). The ics and flexibility performance (Chan and Chan 2008; value chain is re-configured and modularised into tech- Huo, Haq, and Gu 2020). Untimely sharing of informa- nologically optimised networks and retail activities are tion creates uncertainty in operations due to ambiguity coordinated real-time with machine-made operational about materials supply, supply capacity (G. Zhang, Shang, decisions through the use of machine learning (ML) tech- and Li 2011) and lead time (Osman and Demirli 2012), niques (Leavy 2019). All business activities such as sales, as well as internal operations. Many studies have found marketing and production are decentralised, scalable and that timely information sharing greatly impacts SC per- optimised. Data generated can be collected continuously formance particularly in reducing the bullwhip effect via from real-time interactions and online processes for an better coordination (Jeong and Hong 2017; Ouyang 2007; ongoing feedback loop into the system providing ‘more Tang et al. 2021) and allow firms to better manage their context, greater opportunities for stakeholders’ for inno- decision-makings leading to improved resource utilisa- vative purposes (Dwivedi et al. 2017, 198). Through tion and lower supply costs. To date, SCA is considered AI technologies, customer issues can be diagnosed and one of the main issues of present-day SC management fixed automatically with customer confirmation (Ming and is the suggested means by which firms master mar- 2018; Kumar, Singh, and Dwivedi 2020). Trends and sales ket turbulence and handle disruptions (D. Gligor et al. forecasts can be generated fast to provide much-needed 2019). insights into customer preferences and this enables busi- In this regard, various studies have examined the nesses to react timely to meet customer personalisation effects of disruptions on the SCs; particularly, the demands. In this manner, firms can adapt dynamically COVID-19 pandemic has been extensively studied by and rapidly in response to market changes. Business scholars (Ivanov and Dolgui 2020b; Queiroz et al. 2020; decisions become smarter (Leavy, 2019). Additionally, Nikolopoulos et al. 2020; Chowdhury et al. 2021; Ivanov manufacturers can reduce inventory levels and minimise 2020; Gupta et al. 2021). However, existing research has wastage including improving profit margin. And in the yet to consider how small-medium enterprises (SMEs) case of Taobao, credit risk assessments for the small busi- can identify appropriate strategies during disruptions nesses using the Alibaba platform are decidedly hard and assess the effectiveness of their strategies in the con- (Zeng 2018). Here, AI-powered algorithms are trained text of the firm’s capabilities (Gruber, Kim, and Brinck- based on transaction data to gauge the competitiveness mann 2015; Papadopoulos, Baltas, and Balta 2020). of the business offerings and determine the credit ratings Despite the significance of SMEs, there is scant informa- of these small businesses. In this manner, the financ- tion on SME disaster planning and recovery (Helgeson ing needs of an ecosystem are taken care of via auto- et al. 2020; Papadopoulos, Baltas, and Balta 2020). Yet, the matic analysis of all actions taken on the platform. At pressure to remain efficient amid the demands of frugal the same time, these algorithms continuously learn and capacity management persists. The challenge is, there- improve the quality of decisions. The result is a highly fore, for SME managers to ensure SC resilience while successful micro-lending business fuelled by live data maintaining their competitiveness and to develop new and automated decision-making. However, these emerg- strategies in their future SC that allow firms to prepare for ing technologies are often underexploited or neglected and build both medium- and long-term SC resilience. by SMEs (Hansen and Bøgh 2020; Moeuf et al. 2018) INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5537 due to various resources and financial constraints, espe- indicate the firm’s capacity to employ its resources and cially during a disruption (L. W. Wong, Leong, et al. 2020; processes to attain desired outcomes (Huo, Han, and Papadopoulos, Baltas, and Balta 2020; Kumar, Singh, and Prajogo 2016). Dwivedi 2020). Despite the challenges, SMEs can also Within the context of capabilities, scholars have con- explore new opportunities and develop business conti- sidered operational capabilities (Kim 2006; Brusset and nuity strategies because of their relatively smaller size Teller 2017) from dynamic capabilities (S. F. Wamba et al. and greater flexibility. But for many, producing econom- 2017; Blome, Schoenherr, and Rexhausen 2013) where ically meant the ability to counter fast-changing trends superior performance can be attained (Cepeda and Vera and customer demand, all of which require speed and 2007). Operational capabilities are processes and routines accuracy in decision-making. Without the ability to rein- related to knowledge (Cepeda and Vera 2007) that allow vent and adapt their business model, it will be hard for firms to respond to unexpected events affecting the SC SMEs to ensure business continuity. Although the role of performance (Barreto 2010). Here, resilience is an oper- technologies has long been studied as a capability, secur- ational capability critical to maintaining continuity of ing competitive advantage during disruptions depends operations (Brusset and Teller 2017) that allows the SC on how resources are configured and re-configured. to absorb or recover from disruptions (Bhamra, Dani, And this is done via ‘sensing’, ‘seizing’ and ‘orchestrat- and Burnard 2011). By contrast, the dynamic capability ing’ resources and capabilities for business continuity is context-dependent and is a learned pattern of coopera- (Papadopoulos, Baltas, and Balta 2020, 3). Grounded tive activities and strategic procedures that enabled firms on the resource-based view (RBV), the objective of this to accomplish new resource configurations and improve study henceforth is to examine the interplay between competitiveness (Teece 2007). In other words, dynamic the use of AI technologies for risk management, SC Re- capabilities indicate the firm’s integration and reconfigu- engineering capabilities and SC Agility; as well as to fur- ration abilities in addressing rapid changes both within ther examine the effect of SC Re-engineering capability and outside the firm (Huo, Han, and Prajogo 2016). on SCA. According to Brusset and Teller (2017), dynamic capabil- This study begins with a discussion of extant litera- ities enable firms to characterise operational capabilities ture followed by the conceptual model and hypothesis that they wish to augment and across SCs. Similarly, development in the following sections. Subsequently, the Brusset (2016) considered how agility could be enhanced analysis of data and findings are presented. The final sec- based on practices and processes. Taken together, RBV tions of this work include a discussion of the results and provides a theoretical angle to comprehend how firms also a review of managerial and practical implications. can leverage their reserves and capabilities to enhance Lastly, this work concludes with a few general outlines for the performance of SCs during disruptions (Queiroz et al. future research. 2020). Despite RBV being widely employed in SC manage- ment theorising, RBV has also been heavily criticised. 2. Literature Review First, RBV is considered to be lacking in terms of man- agerial implications and operational validity (Kraaijen- 2.1. RBV and dynamic capabilities brink, Spender, and Groen 2009). Managerial leaders The RBV is grounded on the view that firms in posses- are informed of the resources to acquire and develop; sion of valuable and rare resources that are non-imitable however, there is no information on how the resources can achieve sustainable competitive advantage by strate- should be acquired. This creates a gap between what is gically leveraging these resources (Yu et al. 2018). Many theorised to be useful and a prescription on how man- scholars have considered RBV in the context of opera- agers should go about obtaining the resources (Lado tions and SC management (Shibin et al. 2020; Y. Yang, et al. 2006). A second criticism arises in regard to its Jia, and Xu 2019; Huo, Han, and Prajogo 2016; Yu et al. limited applicability. A resource that is rare, inimitable 2018; Barney 2001) due to its suitability in explaining how and nonsubstitutable denies the firms any potential to organisation strategic resources can enable organisations generalise the resource. Further, this could lead to an to gain competitive advantage (Shibin et al. 2020). The infinite search for the ultimate resources and capabili- RBV affords superior performance resulting from effec- ties in the quest for a sustainable advantage that exceeds tive use of resources and organisational capabilities to other firms’ capacity to replicate (Priem and Butler 2001). build strategic relationships with customers, SC partners; The critiques mainly concern the lack of demarcation and flexible and speedy responses to market demands and definition of resource and value, in addition to a (Yu et al. 2018). In general, resources can be tangible or narrow explanation of sustained competitive advantage intangible (Kwak, Seo, and Mason 2018); and capabilities (Kraaijenbrink, Spender, and Groen 2009). Without an 5538 L.-W. WONG ET AL. appropriate emphasis on bundling resources and involv- to understand the different states, outcomes and possi- ing humans in assessing and creating value, the essence ble transitions of SC. These include the use of reasoning of competitiveness would not be sufficiently captured. Petri Nets and its variants (Rossi and Pero 2012; Asar et al. Additionally, resources development and deployment 2006; Blackhurst, (Teresa) Wu, and Craighead 2008) and should be conceptualised as integrations and applications Bayesian Belief Networks (Qazi et al. 2018; Nepal and (actions/process) instead of capacities owned. Kraaijen- Yadav 2015) to estimate probability distributions of SC brink, Spender and Groen (2009) further suggested dis- loses due to disruptions. More recently, Lima-Junior and tinguishing capacity building (includes resources and Carpinetti (2020) adapted a network-based fuzzy infer- capabilities) and capacity deployment processes so that a ence system to evaluate the performance of SC based on more practical resource-based theory can be developed. SCOR metrics. Their work demonstrated greater predic- Oliver (1997) argues that RBV lacks consideration for tion accuracy, learning ability from historical data and social contexts within which resources are sourced. To suitability to decision-making under uncertainty. address the limitations, the authors proposed a theoreti- Additionally, scholars have also used big-data analyt- cal framework integrating the RBV with the institutional ics in close relation to AIRM for risk identification and theory that provides a better explanation of the motiva- management. According to Yang et al. (2020), unlike tion for the adoption of technology that stemmed from large enterprises, credit-related attributes for SMEs are legitimacy (Shibin et al., 2020). In the context of this usually insufficient and hard to acquire. The social rela- study, RBV is the natural fit theoretical base to under- tions between SMR owners, transactions between SMEs stand capacity building for subsequent capacity deploy- or between SMEs and consumers provide abundant inter- ment as the focus excludes legitimacy generated through active data that could help to predict SMEs’ future credit the inclusion of stakeholders and experts. status. Intuitively, these data contain effective informa- tion that can provide insights on these firms’ financial risks, or it can also be very noisy that the data can be 2.2. AI for risk management irrelevant to analysis. However, exploring SC relation- The role of digital technologies in SC research has been ships can help to comprehensively model the SMEs which considered by scholars in various areas from enhanc- could subsequently improve financial risk analysis for ing forecasting, to production flexibility and SC visibility SMEs. Using neural networks, the authors modelled the (Dubey et al. 2019; Baryannis et al. 2019; Ivanov and Dol- credit topological structure and temporal variations of gui 2020a; Belhadi, Kamble, et al. 2021). According to SMEs and proposed an innovative method of SC min- Baryannis et al. (2019), an approach is artificially intel- ing in a semi-supervised link prediction manner to mine ligent if it can autonomously determine the course of SC relationships using a supervised node classification action that can successfully achieve risk management manner to predict loan defaulters. Overall, their work objectives despite not having complete information on demonstrates that SC relationships improve the accuracy the SC environment. The authors considered knowledge- of predictions significantly. Other scholars such as, Cav- based symbolic AI, fuzzy systems, statistical AI, as well alcante et al. (2019) employed a hybrid approach of simu- as ML-based risk management methods, among others. lation and ML to analyse supplier performance risk pro- They further noted the nascent stage of AI predictive files under uncertainty. Their work, which eliminated the and learning capabilities in the sphere of SC risk man- need to estimate the likelihood of disruptions and fore- agement and the most common application of AI-driven casting performance impacts uses the k-NN algorithm risk management (AIRM) is stochastic parameters for and Logistic Regression classifiers to optimise the selec- modelling. Riahi et al. (2021) considered the distribu- tion of suppliers with the best chance of timely delivery tion of AI techniques across supply chain operations ref- based on past data categorised as either on-time or late. erence (SCOR) areas reported genetic algorithms were The authors utilised the data analytics capability of digital mostly used in the planning process followed by neural manufacturing to explore the conditions of resilient sup- networks. Belhadi, Mani, et al. (2021) investigated AI’s plier performance. Applying k-NN, the authors mapped impact on short-term SC performance during the influ- suppliers’ performance according to date and order quan- ence of uncertainty. Ni, Xiao and Lim (2019) showed tity while Logistic regressions are applied to estimate the that the use of ML in SC management is in a devel- probability of supplier on-time delivery. In this manner, opmental stage and there are insufficient publications. the risk profiles or suppliers are determined according to None of these works was specifically targeted at SMEs the probabilities of success in regard to on-time delivery. and thus further studies are required to extend the gen- Their results suggested that a combination of supervised eralisability of the findings on AI’s applicability to SMEs. ML and simulation can improve delivery reliability by Other researchers have used network-based approaches creating digital SC twins. INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5539 Artificial intelligence-driven risk management is Scott, and Fynes 2014; Liu et al. 2018). Re-adjustment, promising for optimising SC and augmenting its resilience re-designing and re-shuffling SCs to integrate resilience (Riahi et al. 2021). Despite this, the use of big-data ana- in SCs are referred to as re-engineering (Christopher lytics and AI remains largely untapped potential and has and Peck 2004). SC re-engineering comprises the inte- received little attention in SC disruption (Xu et al. 2020; gration of processes and activities that are required to Ivanov and Dolgui 2020a). AI technologies can ‘moni- optimise product and service flow (Liu et al. 2018). SC re- tor and control processes in real-time’, enhancing human engineering approaches that are widely adopted include capabilities rather than replacing it (Dwivedi et al. 2021). (i) incorporating viable alternatives in various situations The gap was also echoed by Dolgui, Ivanov, and Sokolov and having flexibilities that can strengthen the firm and (2018), which envisioned a cyber-SC that overarches tra- (ii) storing a safety stockpile and having backup suppliers ditional SC analytics and several other studies (S. Wamba to create a capacity surplus. and Akter 2019; Akter et al. 2020; Belhadi, Mani, et al. 2021). 2.4. Supply chain agility Supply chain agility is a ‘broad and multi-dimensional 2.3. SC re-engineering concept’ that bridges many disciplines (D. M. Gligor The consideration of SCs’ ability to recover from unex- and Holcomb 2014, 161). According to Yang (2014), pected disruptions is grounded on the notion that some there are two avenues where SCA can be examined: risks cannot be averted (Jüttner and Maklan 2011). Dur- (1) speed and responsiveness to uncertain markets (Van ing the onset of disruption, it would be too late to Hoek, Harrison, and Christopher 2001; Swafford, Ghosh, develop preventive solutions and incorporate prepared- and Murthy 2008); and (2) information-driven relation- ness for an efficient response that will allow resources ships (Huo, Han, and Prajogo 2016). In the work of to be deployed in a manner such that the outcome is Van Hoek et al. (2001), agility encompasses a firm’s as planned (Scholten, Scott, and Fynes 2014; Tomasini responsiveness to dynamic and turbulent market and and Van Wassenhove 2009). According to Schleper et al. customer needs. Swafford et al. (2008) identified agility (2021), SCs need to be prepared for unexpected dis- as an outwardly facing capability that reflects the speed ruptions and environmental fluctuations. The need to at which firms adapt to evolving markets. Firms with rethink conventional SC wisdom and foster future SCs agile SCs are better able to respond to unforeseen circum- with heightened ability to adapt to abrupt disruptions stances. Other researchers defined agility in integrating through digitalisation (F. Li 2020), absorb and withstand various companies into streamlined material and how shocks with resilience and sustainability (Ivanov 2020) flexible information flow and performance (Costantino is imminent. Thus, both robust (proactive) and agile et al. 2012). Li et al. (2009) characterised SCA based (reactive) SC strategies are required to enhance a firm’s on factors surrounding strategic response, operational sustenance capability by preventing risks and enabling response, episodic response as two broad dimensions of resistance to change (Wieland and Marcus Wallenburg alertness and capability. 2012). In a reactive approach, SC changes can be mon- Despite lacking a consensus in the definition of agility, itored and strategies relating to customer needs, com- research in SC management has outlined the impor- petitors and techniques. A proactive approach can help tance of developing agility to manage disruption risks identify potential risks and minimise impact before it and ensure service continuity (Braunscheidel and Suresh occurs (Abeysekara, Wang, and Kuruppuarachchi 2019) 2009; C. J. Chen 2019) for firms to take better advan- According to Soni et al. (2014), past literature had tage of changes and synchronise supply with demand. proposed diverse measurements of SCs adaptive capa- In differentiating the concepts of agility and resilience. bility to tackle temporary disruptions; however, there In summary, agility is a key strategy for firms facing is inconsistency regarding the variables that constitute arduous, low probability risk situations because SCs are these measures (Jüttner and Maklan 2011; Liu et al. 2018). required to respond quickly (Abeysekara, Wang, and For instance, some studies considered agility and robust- Kuruppuarachchi 2019). Combined, the dimensions of ness (Wieland and Marcus Wallenburg 2012) while oth- decisiveness, visibility, demand response and customer ers have considered knowledge-management (Scholten, responsiveness form the components of SCA in the con- Scott, and Fynes 2014), SC re-engineering (Scholten, text of this study. Scott, and Fynes 2014), flexibility, redundancy, velocity (Azadeh et al. 2014). While the list presented here is not 3. Research model and hypothesis exhaustive, researchers have outlined the need for firms to employ fitting policies and actions to continually assess Small-medium enterprises need to leverage technologies risks and coordinate efforts of the SC network (Scholten, that allow them to respond to customer requirements 5540 L.-W. WONG ET AL. and improve goods and service quality better to tap into respond efficiently and effectively to events (Soni, Jain, opportunities afforded via seamless and global platforms and Kumar 2014; Liu et al. 2018). The ability to detect (L. W. Wong, Tan, et al. 2020). During the onset of dis- or forecast a potential disruption in advance to reduce ruption, managing risks are not only crucial but more its negative impact (Sheffi 2015). According to Jüttner challenging. This study conceptualises the use of AIRM (2005), effective precautions before the onset of disrup- for two reasons. First, as discussed in earlier sections, tion are possible if risk assessment tools are used to using AI helps firms establish knowledge creation dur- identify weak areas of an SC in advance of a disrup- ing disruptions. Specifically, AI can help firms reduce tion. Along this line, this study suggests that the use of uncertainty by providing insights into firm SC for bet- AIRM to enhance awareness of risk situations in SC at ter predictability and decision-making (Baryannis et al. times of uncertainties enables firms to undertake radical 2019). Secondly, innovation practices that affect SC struc- redesigns such as integration of processes and activities ture can be employed as a means of re-engineering that that can augment and optimise product and service flow. affects the SC performance and is a key factor in SC Further, re-engineering leads to the creation of flexibility management (Sabri, Micheli, and Nuur 2018). Therefore, and redundancy that helps firms recover from disrup- AIRM is a capability that affords SMEs capabilities to tions and build competitiveness (Sheffi 2015; Abeysekara, re-engineer SCs (RP) when necessary and enhances SCA. Wang, and Kuruppuarachchi 2019). Potential disruptions are typically characterised based on the magnitude of 3.1. AI-risk management and SCA impact, the likelihood of occurrence and detection lead time (Sheffi 2015). The earlier the warning of an upcom- According to Chen et al. (2015), firms can capitalise ing disruption, the more a firm can prepare such as relo- on dynamic capabilities to produce ‘cutting-edge knowl- cation of assets, securing backup supplies and in many edge’ (8) amid a dynamic situation. In their work, cases, a sudden disruption such as that of a pandemic may firms’ analytics capabilities can be considered an avenue take weeks for the disruption to hit the firm. Thus, this through which firms can enhance their capability to paper hypothesises that the use of AIRM can potentially process information. This capability enables them to lead to better SC RP. gather, understand and inform appropriate decisions. H2: AIRM has a significant and positive relationship This view is also concurred by Dubey, Bryde, et al. with RP. (2020) in their study on big-data analytics powered by AI. By synthesising information from various sources, AIRM can provide complete visibility with predictive 3.3. SC re-engineering capabilities and SCA data that can significantly reduce cold chain logistics and foster better resource allocation (Myers 2020). Further, According to Wong and Arlbjørn (2008), managing SC insights generated through AI allow firms to better model uncertainties requires firms to be agile, flexible, reliable and predict demand, be more decisive with allocating and fast. Firms that can react and respond to frequently resources with margin-optimisation, can reduce uncer- demands change while continually meeting customer tainty against capacities and supply availability to miti- demands are considered agile. And to be responsive and gate shortages (D. Q. Chen, Preston, and Swink 2015). adaptable, firms need to have the ability to develop flexi- Analytics capabilities used within SC practices are, there- ble practices and operations (Yauch 2011). Gligor (2015) fore, the strategic routes through which firms achieve offered five dimensions of agility, namely (i) the ability new resource configurations. Similarly, this study sug- to quickly detect changes, opportunities and threats, (ii) gests that the use of AIRM offers insights and opportuni- rapid data accessibility within SCs, (iii) resolute decisive- ties for firms to reconfigure resources to adapt to dynamic ness in response to changes, (iv) rapid implementation conditions (Duan, Edwards, and Dwivedi 2019). Collec- of decisions and (v) the ability to ‘modify its range of tively, the use of AIRM can potentially lead to enhanced tactics and operations’ to implement its strategy. All of SCA. Thus, this paper hypothesises that: these abilities are congruent with the definition of SC re-engineering offered by Christopher and Peck (2004). H1: AIRM has a significant and positive relationship with SCA. Earlier sections have also defined RP in the light of flex- ibility and redundancy through AIRM. In this context, RP efforts can be undertaken to identify and manage 3.2. AI-risk management and SC re-engineering SC uncertainties (C. Y. Wong and Arlbjørn 2008) and capabilities accordingly, this study hypothesises that: Firms need to have the knowledge and understanding H3: RP has a significant and positive relationship with of their SC structures to establish a resilient SC that can SCA. INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5541 Figure 1. Conceptual model. H4: RP has a significant and positive mediating effect 4.2. Measurement instrument between SCA and RP. The measurement items were adapted based on estab- The conceptual model representing the above hypothe- lished scales from past literatures with some minor ses is presented in Figure 1 below. wording changes to accommodate the context of the study. The questionnaire items were scored on a seven-point Likert-type scale (1 = strongly disagree and 7 = strongly agree). All constructs were adapted from 4. Methodology past studies: six items for AIRM are adapted from Dubey, 4.1. Data collection and sampling method Bryde, et al. (2020). The construct of SCA is comprised of (i) four items on decisiveness adapted from Gligor The online survey questionnaires were distributed ran- et al. (2013), (ii) five items on demand response, and four domly to Malaysian manufacturing firms listed in the items on customer responsiveness adapted from Braun- Companies Commission of Malaysia (also known as scheidel and Suresh (2009), (iii) two items each on visi- Suruhanjaya Syarikat Malaysia (SSM)) directory between bility adapted from Braunscheidel and Suresh (2009) as September and October 2020 using a team of experi- well as Abeysekara, Wang, and Kuruppuarachchi (2019). enced data collectors due to COVID-19 restrictions. SSM Last but not least, five items for RP are adapted from is an agency responsible for sharing and registering incor- Abeysekara, Wang, and Kuruppuarachchi (2019). porated companies and businesses to the public. Klang Valley, which lies between the state of Selangor and Kuala Lumpur, the capital city of Malaysia, was chosen as the 5. Data analysis sampling location because the region contributed to the 5.1. Respondents’ profile most income of Malaysia’s Gross Domestic Product and has been regarded as the most advanced region for the The profile of the respondents is presented in Table 1. manufacturing sector (Tan et al. 2018; L. W. Wong, Tan, et al. 2020). All respondents were required to evalu- 5.2. Statistical analysis ate the questionnaire using a seven Likert scale rang- ing from strongly agree to strongly disagree. Out of the For this purpose of study, partial least squares-structural 400 questionnaires distributed, 270 responded and after equation modelling (PLS-SEM) was adopted through omitting incomplete data and straight-lining responses, SmartPLS version 3.2.8. The choice of using PLS-SEM is only 252 valid questionnaires were used for this anal- mainly due to the unfulfilment of normal data distribu- ysis. Thus this translates to a response rate of 93.3%. tion. One-sample Kolmogorov–Smirnov test shows that Prior to data distribution, the survey instrument was also all p-values of the indicator construct are less than.05 evaluated for content and face validity with 10 indus- which shows that the distribution of data is non-normal try practitioners who were experts in SC and AI. Some (Dalvi-Esfahani et al. 2020; Tew et al. 2021). Addition- modifications were made to the questionnaire based on ally, PLS-SEM is also suitable for the complicated model the feedback such as the removal of jargon and vogue with the presence of second-order constructs (Ooi, Hew, statements. and Lin 2018). Since SCA is modelled as a second-order 5542 L.-W. WONG ET AL. Table 1. Demographic analysis. judgement, common method bias (CMB) was performed Demographic characteristics Frequency Per cent statistically through the approach developed by Liang Gender Female 131 52.0 et al., (2007) as shown in Table 2. Since the majority Male 121 48.0 of the method, factor loading (FL) is insignificant and Age (years) 30 and below 90 35.7 Between 31 and 40 107 42.5 the substantive variance higher than the corresponding Between 41 and 50 45 17.9 method variance, the results indicated that CMB does 51 and above 10 4.00 not pose a problem (Lee et al. 2020). Several procedu- Number of years with Less than 1 17 6.70 organisation (years) ral remedies were also adopted to restrain CMB such as 1–2 65 25.8 using simple language, assuring maximum confidential- 3–5 88 34.9 6–10 45 17.9 ity and anonymity of participants, informing that there 11–20 22 8.70 are no right or wrong answers and listing the exogenous Above 20 Years 15 6.00 Job position Executive (e.g. Oﬃcer, 127 50.4 construct items before the endogenous construct items Accountant, Senior during the development and administration of the ques- Accountant, Engineer, tionnaire (Philip M. Podasakoff et al. 2003; Adhikari and Senior Engineer, Staﬀ Engineer, System Panda 2020). Analyst, Assistant Manager, etc.) Senior Staﬀ 86 34.1 Engineer/Principal Engi- neer/Manager/Senior Manager/Head of 5.4. Assessing the outer measurement model Department General Man- 20 7.90 ager/Director/Senior Dijkstra Henseler’s (rho_A) and composite reliability Director/Executive (CR) were used to measure internal reliability. In Table 3, Director/Managing all rho_A and CR values for the first- and second-order Director/Chief Executive Oﬃcer/Vice constructs were above the limit of 0.60, exhibiting strong President/President/Chairman internal reliability (Cachón Rodríguez, Prado Román, Other 19 7.50 Age of ﬁrm (years) < 5 Years 22 8.70 and Zúñiga-Vicente 2019; Loh et al. 2020; Bawack, 5 ≤ Years < 10 Years 75 29.8 Wamba, and Carillo 2021). Convergent validity (CA) is > 10 Years 154 61.1 Category of Electrical and electronics 48 19.0 assessed by FL and average variance extracted (AVE). organisation All FL shown in Table 3 is greater than the threshold of product 0.70 except for RP1, RP4, CR4, DR5 and VI2 (Loh et al. Chemical 19 7.50 Textile 17 6.70 2022). Hair et al., (2017) concluded that FL values rang- Food 63 24.9 ing between 0.40 and 0.70 are acceptable if AVE > 0.5 Rubber and plastic 34 13.4 Machinery and hardware 34 13.4 and CR > 0.70 and should be considered for removal if Others 38 15.0 the values are below 0.40. RP1 and RP4 FL were removed Number of employees Less than 5 26 10.3 5 to < 75 134 53.0 from Table 3 due to poor-loading CR4, DR5 and VI2 75 to ≤ 200 57 22.5 on the other hand were retained as the FL can explain > 200 36 14.2% about 50% of the AVE and above the minimum thresh- old of 0.70 for CR. Additionally, AVE for each construct ranged from 0.533 and 0.715 for both first- and second- construct with DR, DE, VI and CR as its first-order con- order and has exceeded the minimum threshold of 0.50 struct, the suitability of PLS-SEM is warranted in this (L. W. Wong, Tan, et al. 2020; Dubey et al. 2021). Both study. G∗Power was further employed to estimate the criteria denote that the measurement model has a good minimum sample size using an effect size, f 2 of 0.15, CV. Next, the discriminant validity (DV) was assessed by probability of error, α =.05 and power level, (1 – β) =.8 the means of the Hetero-Trait-Mono-Trait (HTMT) ratio with 2 as the number of predictors. The actual sample size of correlations. Tables 4 and 5 show that all HTMT val- of 252 was more than the minimum sample size of 68 to ues for first- and second-order are below the minimum assess the proposed conceptual framework. value of 0.85. In addition, the HTMT inference test in Tables 4 and 5 has also been established as the results show the confidence interval did not show a value of 1 5.3. Common method bias for any of the constructs for both 2.5% (lower bound) and Since the data were collected via a self-reported question- 07.25% (higher bound) suggesting that there is adequate naire and in particular, the exogenous and endogenous DV throughout the model (Henseler, Ringle, and Sarstedt constructs were measured by respondents’ perceptual 2015). INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5543 Table 2. Common method factor analysis. Latent Construct Indicators Substantive factor loading (Ra) Ra2 Method factor loading (Rb) Rb2 AIRM AIRM1 0.65∗∗∗ 0.42 0.19∗ 0.04 AIRM2 0.87∗∗∗ 0.76 −0.09NS 0.01 AIRM3 0.95∗∗∗ 0.90 −0.17∗∗ 0.03 AIRM4 0.99∗∗∗ 0.98 −0.17∗ 0.03 AIRM5 0.78∗∗∗ 0.61 0.04NS 0.00 AIRM6 0.57∗∗∗ 0.32 0.20NS 0.04 CR CR1 0.83∗∗∗ 0.68 0.02NS 0.00 CR2 0.83∗∗∗ 0.69 −0.02NS 0.00 CR3 0.90∗∗∗ 0.81 −0.09NS 0.01 CR4 0.48∗∗∗ 0.23 0.13NS 0.02 DE DE1 0.67∗∗∗ 0.45 0.12NS 0.01 DE2 0.95∗∗∗ 0.91 −0.14∗ 0.02 DE3 0.86∗∗∗ 0.74 −0.01NS 0.00 DE4 0.77∗∗∗ 0.60 0.04NS 0.00 DR DR1 0.75∗∗∗ 0.57 0.05NS 0.00 DR2 0.39∗∗∗ 0.16 0.37∗∗∗ 0.14 DR3 0.86∗∗∗ 0.75 −0.15∗ 0.02 DR4 0.91∗∗∗ 0.82 −0.14∗ 0.02 DR5 0.72∗∗∗ 0.52 −0.13NS 0.02 RP RP2 0.72∗∗∗ 0.51 0.08NS 0.01 RP3 0.86∗∗∗ 0.73 −0.09∗ 0.01 RP5 0.76∗∗∗ 0.57 0.00NS 0.00 VI VI1 0.89∗∗∗ 0.79 −0.11NS 0.01 VI2 0.70∗∗∗ 0.48 −0.07NS 0.00 VI3 0.70∗∗∗ 0.50 0.11NS 0.01 VI4 0.730∗∗∗ 0.53 0.05NS 0.00 Average 0.77 0.62 0.00 0.02 Note: ∗∗∗p <.001; ∗∗p <.01; ∗p <.05, NS insigniﬁcant. 5.5. Inspecting the inner structural model and large effects for values above 0.02, 0.15 and 0.35, respectively (Cohen 1988; Hew et al., 2018). The results The model fit was measured using the standardised root of the f2 effect sizes are summarised in Table 8 with val- mean square residual (SRMR) value (Hair et al. 2017). ues ranging from 0.209 to 0.683 indicating medium and Since the SRMR value for both saturated and estimated large effects on the predictor construct RP and SCA. models was below 0.08, Hu and Bentler (1999) indicate The predictive relevance of the structural model was that the model has a good fit. Further, multicollinear- further accessed using Stone–Geisser’s Q2 value (Yuan ity was not a concern as all the inner variance inflation et al. 2021; Yan et al. 2021). Table 9 illustrated that both factor (VIF) values for both first- and second-order con- endogenous construct RP and SCA showed predictive structs were below 5.00 which indicates that the problem accuracy of the model with Q2 values greater than zero. of multicollinearity is not a concern (Cao et al., 2021; Lew et al., 2020). The hypotheses in the structural model were tested using a bias-corrected and accelerated (BCa) 5.7. Artificial neural network analysis bootstrap procedure with 5000 sub-samples. Figure 2 and In view of the limitation of PLS-SEM that can only cap- Table 6 show that AIRM (β =.416, p <.001) has a direct ture compensatory and linear studies (Lim et al. 2021), positive relationship with RP thus supporting H2. Sim- the study further complements the PL-SEM analysis ilarly, AIRM (β =.545, p <.001) and RP (β =.400, by adopting the artificial neural network (ANN) anal- p <.001) are positive and significant with SCA, hence ysis as ANN is able to capture non-linear relationship supporting H1 and H3. Table 7 also ascertained that RP in this study and therefore useful in decision-making partially mediates the path between AIRM and SCA in (Ooi, Hew, and Lin 2018; Wan et al. 2021). Two ANN a complementary manner (Hew et al. 2018). This out- models were constructed for SCA and RP. In order to come support H4. The coefficient of determination, R2 determine the predictive accuracy of models A and B, in Table 8 shows that 63.9% of the variance in SCA is the root mean squared error (RMSE) is calculated for explained by RP and AIRM. the 10 neural networks (Wang et al. 2022). Table 10 shows that all the RMSE value shows high prediction 5.6. The predictive relevance and effect size accuracy as they are relatively small with values rang- ing from 0.079 to 1.996 (Lee et al., 2020). Addition- The f 2 effect size was further assessed using Cohen’s f2 ally, the study also ranks the exogenous based on the whereby the intensity is represented by small, medium 5544 L.-W. WONG ET AL. Table 3. Loadings, composite reliability, Dijkstra Henseler and average variance extracted. Dijkstra Henseler’s Composite Average variance Constructs Items Loadings (p-levels) (rho_A) reliability (CR) extracted (AVE) First order AIRM AIRM1 0.812 (p <.001) 0.891 0.915 0.643 AIRM2 0.792 (p <.001) AIRM3 0.796 (p <.001) AIRM4 0.837 (p <.001) AIRM5 0.824 (p <.001) AIRM6 0.748 (p <.001) RP RP2 0.788 (p <.001) 0.670 0.818 0.600 RP3 0.772 (p <.001) RP5 0.763 (p <.001) CR CR1 0.835 (p <.001) 0.781 0.854 0.596 CR2 0.816 (p <.001) CR3 0.806 (p <.001) CR4 0.610 (p <.001) DE DE1 0.782 (p <.001) 0.833 0.889 0.667 DE2 0.832 (p <.001) DE3 0.845 (p <.001) DE4 0.806 (p <.001) DR DR1 0.797 (p <.001) 0.793 0.850 0.533 DR2 0.725 (p <.001) DR3 0.733 (p <.001) DR4 0.783 (p <.001) DR5 0.596 (p <.001) VI VI1 0.798 (p <.001) 0.751 0.838 0.566 VI2 0.649 (p <.001) VI3 0.799 (p <.001) VI4 0.754 (p <.001) Second order AIRM∗ 1.000 1.000 AIRM∗ 1.000 SCA 0.868 0.909 0.715 CR 0.831 (p <.001) DE 0.857 (p <.001) DR 0.838 (p <.001) VI 0.857 (p <.001) RP∗ 1.000 1.000 RP∗ 1.000 Note: ∗Single-item constructs were excluded from this analysis. Table 4. Hetero-Trait-Mono-Trait Assessment (HTMT) for ﬁrst-order constructs. Latent construct AIRM CR DE DR RP VI AIRM CR 0.761 [0.653, 0.862] DE 0.68 [0.529, 0.809] 0.752 [0.631, 0.88] DR 0.609 [0.487, 0.715] 0.731 [0.618, 0.826] 0.812 [0.711, 0.897] RP 0.528 [0.348, 0.682] 0.61 [0.485, 0.747] 0.76 [0.635, 0.873] 0.839 [0.742, 0.934] VI 0.792 [0.700, 0.871] 0.849 [0.673, 1.027] 0.796 [0.618, 0.948] 0.788 [0.636, 0.921] 0.741 [0.532, 0.933] Note: The values in the brackets represent the lower and the upper bounds of the 95% conﬁdence interval. Table 5. Hetero-Trait-Mono-Trait Assessment (HTMT.85 ) for the second most important predictor of SCA. While second-order constructs. there is only one single neuron model for ANN model Latent construct AIRM RP SCA B, the sensitivity analysis shows a 100% of normalised AIRM importance. Results between PLS-SEM and ANN were RP 0.416 [0.277, 0.537] compared using patch coefficient and normalised relative SCA 0.763 [0.691, 0.822] 0.673 [0.582, 0.75] importance, respectively (Ng et al. 2022), and Table 12 Note: The values in the brackets represent the lower and the upper bounds of the 95% conﬁdence interval. shows that both results are consistent for ANN model A and B. normalised relative importance towards the endogenous 6. Discussion and implication variable in Table 11 (Lim et al. 2021). In ANN model A, AIRM is the most important (100% normalised rel- The results of this study suggest that AIRM is a significant ative importance) predictor of SCA while RP is ranked determinant of RP and SCA. This finding supports past INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5545 Table 8. Eﬀect size (f 2 ). Predictor con- structs/dependent constructs RP SCA AIRM 0.209 0.682 RP 0.367 SCA Table 9. Predictive relevance (Q2 ) and R2. Endogenous construct Q2 Predictive relevance R2 RP 0.169 Q2 > 0 0.173 SCA 0.451 Q2 > 0 0.639 and constraints of operations, understand relationships, provide visibility into operations and support decision- making. Further, the findings of this study also resonate Figure 2. Result of hypotheses testing. with the work of Liu et al. (2018) which suggests that risk management performance is essential in generating literature on the potential benefits of AIRM (Baryannis positive RP and SCA outcomes; and risk management et al. 2019; Hosseini and Ivanov 2020) and SMEs (Hansen culture positively affects the RP and SCA capability of SCs and Bøgh 2020; S. Yang et al. 2020). AI technology is a (Abeysekara, Wang, and Kuruppuarachchi 2019). An ear- critical integration to firms which includes better coor- lier work by Wieland and Marcus Wallenburg (2012) also dination in an uncertain environment, understanding concluded that SCA is essential to deal with customer- and predicting consumer habits, developing personalised related risks. profiles of customers, trust building (Dwivedi et al. 2021; According to Christopher and Peck (2004), SCA com- Dubey, Bryde, et al. 2020) and supplier management prises two key ingredients: visibility and velocity. Simply for decision-making (Borges et al. 2021). The adaptive put, the analytics capabilities of SC driven by big data, capability of SC in dealing with disturbances, respond- AI or machine capabilities provide insights for real-time ing to disruptions and subsequently recovering by means decision-making (S. Wamba and Akter 2019). The use of of maintaining continuity of operations can be greatly AIRM, therefore, allows firms to not only enhance firm enhanced through the use of AI and increased informa- performance but is also a driver for accelerating firm per- tion processing capability. The enhanced operational SC formance through robust agility in operations. Managing transparency as per Dubey, Gunasekaran, Bryde, et al. SC agility requires real-time visibility. Utilising AIRM (2020) by means of proactive communication among to harness fast and big data enable managers to rapidly stakeholders leads to better visibility and traceability predict risks via identification and quantification from in SC operations. Here, we can say that AIRM is a past impact and prescribe mitigating strategies tailored potent instrument for firms to address the opportunities to the particular scenario (Dwivedi et al. 2021). The use Table 6. Outcome of the structural model examination. Standard Bias-corrected Original Sample mean deviation T statistics conﬁdence PLS path sample (O) (M) (STDEV) (|O/STDEV|) P-values intervals Remarks AIRM → RP∗∗∗ 0.416 0.415 0.067 6.191 0 0.277 0.536 Supported AIRM → SCA∗∗∗ 0.545 0.545 0.041 13.401 0 0.464 0.624 Supported RP → SCA∗∗∗ 0.400 0.402 0.048 8.423 0 0.302 0.486 Supported Notes: ∗∗∗Signiﬁcant at p <.001 level. Table 7. Mediation analysis. 95% Cls of the Signiﬁcant Indirect 95% Cls of the Signiﬁcant Mediation Direct eﬀect direct eﬀect t-Value (p <.05)? eﬀect indirect eﬀect t-Value (p <.05)? type Paths Algorithm Bootstrapping Algorithm Bootstrapping Types of mediation AIRM → SCA 0.545 (0.464, 0.624) 13.401 Yes 0.166 (0.116, 0.229) 5.798 Yes Complementary via RP (partial mediation) 5546 L.-W. WONG ET AL. Table 10. RMSE values for SCA and RP. engagement with users and knowledge is acquired from Model A Model B it. While these are standard processes, AIRM offers speed Input: AIRM, RP Input: AIRM and possibly better suggestions by analysing a much more comprehensive range of past scenarios, live data and form Output: SCA Output: RP a feedback loop to the system. It was also observed that SC Training Testing Training Testing re-engineering capability is significantly related to SCA Neural network RMSE RMSE RMSE RMSE and mediates the relationship. Leung et al. (2018) demon- ANN1 0.079 0.079 0.123 2.219 ANN2 0.077 0.081 0.120 2.492 strated that the use of a smart system to re-engineer ANN3 0.079 0.086 0.121 1.387 e-order fulfilment processes can successfully mitigate ANN4 0.079 0.085 0.119 2.457 ANN5 0.080 0.087 0.121 2.339 irregular order arrival patterns, limited time for order ANN6 0.078 0.091 0.123 1.354 processing and overcome logistics challenges thereby ANN7 0.081 0.058 0.123 1.967 preserving customer satisfaction. However, this study ANN8 0.080 0.071 0.126 1.534 ANN9 0.093 0.074 0.124 2.588 deviates from Abeysekara, Wang, and Kuruppuarachchi ANN10 0.080 0.079 0.126 1.623 (2019), whose work considered both RP and SCA as Mean 0.081 0.079 0.123 1.996 SD 0.004 0.010 0.002 0.485 distinct factors of SC resilience and the work of Liu et al. (2018) which outlined that SCA and RP need to be transformed through risk management performance Table 11. Sensitivity analysis. for excellent firm performance. According to Abeysekara, Model B Wang, and Kuruppuarachchi (2019), the capacity for Model A (Output: SCA) (Output: RP) adapting to and coping with uncertainties while ensur- Neural network AIRM RP AIRM ing operations continuity by means of fast adjustments ANN1 0.536 0.464 1.000 is referred to as SC resilience. This term is mainly used ANN2 0.539 0.461 1.000 ANN3 0.510 0.490 1.000 to characterise low probability chronic disruptions while ANN4 0.537 0.463 1.000 smaller but more frequent disruptions such as logis- ANN5 0.560 0.440 1.000 ANN6 0.555 0.445 1.000 tics deliveries, machine and technology caused disrup- ANN7 0.478 0.522 1.000 tions such as in-house disruptions are only recently ANN8 0.505 0.495 1.000 considered in resilience literature. But there exists nei- ANN9 0.490 0.510 1.000 ANN10 0.476 0.524 1.000 ther a consensus on the definition of resilience nor a Average relative 0.519 0.481 1.000 focus on the decisive factors of resilience. Abeysekara, importance Normalised relative 100.000 92.827 100.000 Wang, and Kuruppuarachchi (2019) conceptualised SC importance (%) resilience as comprising of re-engineering, agility, col- laboration and risk management culture (Christopher and Peck, 2004) with risk management culture being of AI has also been shown to improve SCA within stake- the pre-requirement for resilience following Liu et al holder collaborative relationships (Dubey, Bryde, et al. (2018). This work focuses on the agile SC that pur- 2020). Alter (2021) differentiates four categories of smart- sues faster responses fuelled by live data and feedback ness for devices according to specific capabilities. These loop powered by AI technologies to ‘create the ability are information processing, internal regulation, knowl- to respond rapidly and cost-effectively to unpredictable edge acquisition and action outside the organisations changes in markets and environmental turbulences’ (Car- where capabilities such as sensing, actuation, coordina- valho, Azevedo, and Cruz-Machado 2012, 50). An agile tion, communication and control may be augmented by management approach entails consideration of the fol- AI applications. According to the author, every informa- lowing parameters: market sensitivity, customer satisfac- tion processing is present in a work system and includes tion, quality improvement, use of technologies among internal regulation to recognise its state and respond others. accordingly. Subsequently, actions are deployed through Table 12. Comparison between PLS-SEM and ANN results. Ranking (ANN) Original sample ANN results: Ranking (PLS-SEM) [based on (O)/Path Normalised relative [based on Path normalised relative PLS path Coeﬃcient importance (%) coeﬃcient] importance (%)] Remark Model A (Output: SCA) AIRM → SCA 0.545 100.000 1 1 Match RP → SCA 0.400 92.827 2 2 Match Model B (Output: RP) AIRM → RP 0.416 100.000 1 1 Match INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5547 6.1. Theoretical implications value use cases, analytics capabilities and tools among others. As described earlier, RBV has been criticised for its lack Researchers have considered SMEs in a juncture of of applicability as theoretical grounding. This work con- either not having the capability and technological infras- tributes to RBV theory development. The study identified tructure to undertake technological interventions (L. W. that an SC must exhibit characteristics of agility to sense, Wong, Leong, et al. 2020), or having the flexibility to reconfigure and transform in response to uncertainties. change and adapt due to size (Shepherd and Williams The inclusion of AIRM characterised by the use of live 2018). There are four ways in which SMEs need to focus data and feedback loop necessarily complements the on in order to be more resilient (Bak et al. 2020). First, resources of firms in terms of pursuing faster responses is the role of collaboration. Collaboration with a net- to dynamic situations. This is in line with transform- work or specialised partners allows SMEs to maintain the ing resources accordingly to address SC uncertainties. focus on their core competencies so that they can with- Resources remain the core elements of firms to survive stand disturbances. Close relationships with customers disruptions. and partners can help SMEs to strengthen their techno- This study extends the existing literature on the use logical base, and broaden the adoption of technologies of AIRM and its implications on RP and SCA. Further, while investing in building strategic and complementary the concept of SCA is studied as a composition of four relationships. Second, the smaller size of SMEs allows dimensions, namely customer responsiveness, visibility, for greater agility, flexibility and quick response at times demand responsiveness and decisiveness. Extant litera- of disturbances thereby overcoming the lack of capabil- ture has highlighted the multi-dimensional perspectives ities by being more resilient through diverse customer of agility and the unavailability of a standardised con- portfolios and seamless communication systems. Thirdly, ceptualisation of agility (Abeysekara, Wang, and Kurup- by developing a shared pool of technical content and puarachchi 2019; D. Gligor et al. 2019). In this manner, market knowledge, SMEs can enhance their abilities to this study contributed further to the understanding of the withstand turbulence and respond to external changes. agility concept in the context of managing risks through According to Papadopoulos et al. (2020), aligning SME’s the use of AI. In response to RQ1 and RQ2, the find- business strategy with technologies constitutes a research ings show that the use of AI for SC risk management is a avenue and whether the technological alignment is at par strong predictor of SC re-engineering RP and SCA. Fur- with business strategy or an indication of SME’s tech- ther, there is a mediating effect between SC RP and SCA. nological investments. Further, high costs and unclear This work adds empirical evidence to the potentially ben- return-on-investments may hinder SMEs from adopting eficial use of AI for SC risk management as highlighted technologies but on the same aspect (Agrawal, Pandit, from extant reviews (Ni, Xiao, and Lim 2019; Riahi et al. and Menon 2012), SMEs may overcome their financing 2021; Akter et al. 2020; Belhadi, Mani, et al. 2021). AI limitations through the use of technologies as explained algorithms such as neural networks, genetic algorithms earlier in the case of Taobao. Therefore, it is pertinent and support vector machines can address complex SC that SMEs understand the impact of technologies on their management problems such as demand/sales estimation businesses and innovativeness. SMEs need to consider and can accurately predict retailer demands with time if they have a clear and actionable roadmap to achieve lags (Ni, Xiao, and Lim, 2019). The forecasting models the alignment. Is there a clear understanding of and abil- built using methods such as neural networks, fuzzy logic ity to govern the infrastructure? Along this vein, future and data mining are more reliable than traditional mod- research could also consider the formidable issue of tech- els did. According to Riahi et al. (2021), the application nological debt and its implications on competitiveness in of AI in SCs has been studied in diverse sectors – retail, the face of ever-changing business realities that demand automotive, manufacturing, healthcare and several other higher agility. sectors. However, the majority of the papers studied in their work considered a simulation-based approach and does not focus on real case application. One explana- 6.2. Managerial implications tion for this is that AI requires a vast amount of existing data for learning and to achieve the potential of AI- Supply chain risk management remains a significant chal- first requires the use of big-data and applying analyt- lenge that affects firms’ performance (Brusset and Teller ics for subsequent prediction (Akter et al., 2020). This 2017) and in particular, SMEs need to be at the forefront means successful AI transformation hinges on a good of adopting new technologies if they wish to remain com- data ecosystem with strong data governance, business petitive. However, the complexity of AI is a challenge for 5548 L.-W. WONG ET AL. SMEs that lack the knowledge and resources to exploit and putting in place an appropriate infrastruc- the benefits of technology despite being aware of its ben- ture to support the technology. And as with any efits (Hansen and Bøgh 2020). That said, the challenge digitalisation initiatives, management needs to be for SMEs in navigating uncertainties and disruptions prepared for various governance and protection goes beyond technological affordability. Calling man- issues arising therewith. Where necessary, SMEs agers to be open-minded towards emerging technolo- need to explore opportunities for assistance (both gies (L. W. Wong et al. 2021), cautioned that there are educational and financial) that are available to often pre-conceived notions about specific technology them from various entities and plan the journey and users tend to reject innovations without due con- for adoption. Integrating AIRM is a journey that sideration or their potential. Thus, understanding AI is requires careful consideration and planning; an critical for acceptance and subsequently, SMEs need to effort that goes beyond a mere upgrade or a patch. reassess and formulate the right strategies that will see them through these challenging times. Practical impli- 7. Conclusion cations are provided below based on the findings of this study: The purpose of this study was to investigate the use of AIRM for enhancing SCA and RP for securing business (i) Transitioning to the use of AIRM requires man- continuity. The hypothesis that AIRM enhances SC RP agers to adopt a proactive, risk-taking and innova- and SCA is all supported; the mediating effect of RP and tive mindset (Dubey, Gunasekaran, Childe, et al. SCA is also supported. Thus, the use of AIRM allows 2020). Managers need to learn from the disrup- SMEs to better cope with the changes caused by dis- tions and take appropriate measures. They must ruptions. This means SMEs exploring opportunities for make rapid decisions and take immediate actions improving SCA and RP may find AIRM a strong driver. to maintain business operations. They must be Nevertheless, some limitations remain. This study is con- careful in identifying opportunities from threats ducted using AIRM in SCs for SMEs in a specific region. addressing issues that could slow or cripple the SC. Generalisability may be problematic although the liter- Echoing Wong et al. (2021), SCs generally involve ature has presented sufficient support for the model as multiple stakeholders and a prime challenge for discussed in earlier sections. The model in this study managers implies that they must first understand is also tested with encouraging outcomes. However, this the benefits of any emerging technologies in order is also an opportunity for future work to consider a to persuade adoption before any benefits can be broader base of respondents. Further, the model did not reaped. include other influences of technological adoption such (ii) Agility and re-engineering capabilities need to as culture, management commitment and technological be architected into SCs. The use of technologies readiness. We recommend future studies consider these enables scenario planning to help managers antic- factors in understanding the adoption of AI for SC risk ipate expected and worst-case situations to ensure management. As per Li, (2020, 812), ‘in today’s unpre- supply and demand alignment. With the use of dictable digital environment, it is no longer viable to AI and analytical tools, control towers can high- develop a new strategy and then execute it over many light situations where disruptions will result in years’. There is a mismatch between traditional business difficulty meeting demands. SC visibility, espe- models and digital future – strategies must be recali- cially during disruptions is important for firms brated through execution, for digital transformation is to see how SC is affected and takes appropriate an ongoing process. Likewise, emerging technologies and actions. their applicability call for new research work to be con- (iii) Finally, the use of AIRM requires the presence tinually conceptualised and validated to be of signifi- of data. SMEs need to understand and embrace cance. Along with this, future research can also con- the notion that digital transformation is the way sider incorporating the moderating effects of disruptions forward. Therefore, it is important that all lev- and performance outcomes in the research. Finally, the els of stakeholders fully understand the impli- AI ecosystem is a ‘family of overlapping aspects, tech- cations and issues surrounding the adoption of nologies and techniques’ (Stahl 2022) that raises several AIRM. Management needs to take the pivot role governance-related challenges. Future research may con- in encouraging and spearheading the transfor- sider incorporating policy and governance aspects to help mation. This may mean moving from traditional organisations decide on the adoption of responsible AI manual-based recording to digital capture of data algorithms as well as security and data privacy concerns INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5549 that are characteristics of AI techniques and big-data Yogesh K. Dwivedi is a Professor of applications. Despite these limitations, AI-enabled sys- Digital Marketing and Innovation and tems are powerful and promise firms better usage of their Founding Director of the Emerging Mar- kets Research Centre (EMaRC) at the data, optimisation of processes and innovate businesses. School of Management, Swansea Univer- sity, Wales, UK. In addition, he holds a Distinguished Research Professorship at Disclosure statement the Symbiosis Institute of Business Man- No potential conflict of interest was reported by the author(s). agement (SIBM), Pune, India. Professor Dwivedi is also cur- rently leading the International Journal of Information Man- agement as its Editor-in-Chief. His research interests are at Notes on contributors the interface of Information Systems (IS) and Marketing, focusing on issues related to consumer adoption and dif- Lai-Wan Wong is an Assistant Professor fusion of emerging digital innovations, digital government, in the School of Computing and Data and digital and social media marketing, particularly in the Science and Deputy Registrar of Xiamen context of emerging markets. Professor Dwivedi has pub- University Malaysia. She is interested in lished more than 500 articles in a range of leading academic human dynamics specifically in the digi- journals and conferences that are widely cited (more than tisation and transformation of society, 35,000 times as per Google Scholar). He has been named both micro- and macro-levels. To date, on the annual Highly Cited ResearchersTM 2020 and 2021 her works have appeared in notable jour- lists from Clarivate Analytics. Professor Dwivedi is an Asso- nals including IEEE Transactions on Engineering Management, ciate Editor of the Journal of Business Research, European International Journal of Production Research, Supply Chain Journal of Marketing, Government Information Quarterly and Management: An International Journal, International Journal of International Journal of Electronic Government Research and Information Management, etc. Senior Editor of the Journal of Electronic Commerce Research. Garry Wei-Han Tan is an Associate Pro- More information about Professor Dwivedi can be found at: fessor at the Graduate Business School, http://www.swansea.ac.uk/staff/som/academic-staff/y.k.dwivedi/. UCSI University. His research interests include mobile commerce and consumer Data Availability Statement behaviour. Since 2019, he has been rated as one

Artificial Intelligence-Driven Risk Management for Supply Chain Agility PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue