Enhancing Supply Chain: Exploring and Exploiting AI Capabilities

ABSTRACT

Drawing upon the Organizational Information Processing theory principles, this study delves into the impact of explorative and exploitative artificial intelligence (AI) capabilities on supply chain (SC) performance, specifically regarding efficiency and resilience. A cross-sectional descriptive study through survey leverages structural equation modeling and data gathered from 267 distributor firms to test the proposed model. Results indicate that Explorative AI capabilities have a positive impact on SC resilience, but they do not significantly impact SC efficiency. Conversely, exploitative AI capabilities have a positive influence on SC efficiency and a negative influence on SC resilience. Furthermore, the importance of the digital divide among supply-chain partners is established, which affects SC outcomes. The study took a broad view of AI capabilities as explorative and exploitative AI capabilities which have different impacts on SC outcomes. Another novelty is that it considers “digital divide” among partner firms presenting the views of an emerging economy.

KEYWORDS:

• Artificial intelligence
• digital divide
• AI capabilities
• supply chain resilience
• supply chain efficiency

Introduction

The global landscape has been steadily progressing toward a digital future, with a growing focus on Industry 4.0 technologies.¹ Artificial intelligence (AI), often described as the ability of machines to interact with and imitate human capabilities,² has been a subject of academic interest for some time. Interestingly, the application of AI is typically an integral part of a firm’s data-driven digital transformation project for enhanced performance.³ Only recently have technological advancements unveiled an extensive range of applications for AI, including its integration into supply chain management (SCM).⁴ For example, FedEx leverages data-driven insights derived from AI and machine learning to assist customers in adapting to supply chain disruptions and market challenges.⁵

The supply chain management market is expected to exceed $62.2 billion by 2030, driven by digital transformation, Industry 4.0, and an emphasis on data-driven decision-making.⁶ The goal is to gather essential data regarding supplier performance, factory locations, product development, orders, shipments, and deliveries along the supply chain. Given the magnified role of AI, there is a need to empirically investigate the contribution of AI to the field of SCM, as highlighted in the literature.^4,7–9 Leaders now understand three additional goals in addition to the conventional SC strategy’s functional focus (1) Resilience as strengthening systems to withstand outside attacks to facilitate speedy recoveries, (2) Agility which involves developing the essential capacity to adapt to new market conditions swiftly, and (3) Sustainability as making decisions that will lead to long-term success while taking into account environmental considerations and responsible resource utilization.¹⁰ Srinivasan and Swink¹¹ stated that there is a growing consensus on the importance of integrating SC information to attain superior SC performance. More precisely, a firm’s proficiency in effectively harnessing IT-related resources and applications, often called IT capability, plays a crucial role in realizing the advantages of integration.¹²

Although previous research has acknowledged the significance of leveraging big data in the form of AI-driven SC analytics to enhance agility and resilience capabilities,^9,12 there remains an unaddressed gap in scholarly literature concerning various AI capabilities and how they influence SC performance. Specifically, there is a gap concerning the roles of explorative and exploitative AI capabilities in enhancing SC resilience and efficiency. Prior research indicates that AI leads to improved efficiency and enhanced SC decision-making.^13,14 However, these studies have examined AI capabilities as one-dimensional construct without delving into distinct mechanisms through which different types of AI capabilities influence SC performance.⁸ Explorative AI capabilities refer to the innovative and adaptive functions of AI such as predictive analytics, machine learning-based trend analysis and advanced simulations.¹⁵ Such capabilities allow firms to determine market changes, optimize supply-chain and adapt to changing market environment proactively.¹⁶ In contrast, exploitative AI capabilities relate to optimizing and refining existing processes using automation and incremental technological improvements to streamline workflows.¹⁷ Thus, explorative AI brings strategic foresight and agility, whereas exploitative AI leads to operational efficiency and consistency.¹⁸

The core problem addressed in this study is the lack of a comprehensive understanding of how explorative and exploitative AI capabilities have a differential impact on SC performance. The distinction between explorative and exploitative AI capabilities is important as it relates to different aspects of SC disruptions. Therefore, this study fills the gap by investigating the impact of these AI capabilities on SC performance through SC resilience and SC efficiency.¹⁹ By recognizing the tension between the two facets of AI capabilities and their potential roles, we provide a more holistic view of the role of AI in SCM, offering meaningful insights for advancing SCM theory and practice.²⁰ To recover effectively from SC disruptions, businesses must exhibit expertise in employing information systems and AI to optimize their established structured processes (exploitation) and explore unstructured processes (exploration).²¹

We utilize Organizational Information Processing (OIP) theory to understand how different AI capabilities can create SC resilience and efficiency. OIP describes how organizations may organize, utilize, and share information for improving their information processing capabilities, which play a vital role in managing uncertainty and supply chain disruptions.^22–24 We contend that OIP relates to organizational structures and capabilities that deal with information processing needs of the firms.²²

Specifically, in the current context, AI exploitation entails the standardization of information and systems to improve information processing capability, facilitating swift decision-making and immediate actions by the focal firm and its SC partners when addressing disruptions. On the other hand, AI exploration encourages extensive interfirm information sharing, fostering a collaborative approach in developing novel solutions to mitigate SC disruptions. Nevertheless, operational managers often grapple with the inherent tension between these two strategies.²¹ The situation varies significantly between developed and emerging economies. To illustrate, the Global AI Adoption Index 2021 report by IBM²⁵ highlights that AI adoption in India lags due to the absence of a cohesive adoption strategy and a shortage of the necessary technology skills. Regarding AI-generated revenue, India reached USD 12.3 billion in 2023, with projections indicating a compound annual growth rate (CAGR) of 42% in 2022. The AI market size in India is expected to expand to USD 71.0 billion by 2027.²⁶ Furthermore, the digital divide among SC partners in India and the Western world will affect how they derive AI benefits and need to be studied formally. Digital divide has evolved into an important discussion point with the rise of digital economy. It refers to disparate access to information communication technology, systems, and skills among people.²⁷ Understanding the digital divide concept “will help in clear technology needs that could lead to the development of more coherent frameworks and policies to address those needs and to intensified efforts to diminish and bridge digital gaps.”²⁸ We consider digital divide as the differences in digital technology adoption and utilization among the SC partners. We cannot assume that access to and use AI and advanced technology will automatically provide all the benefits of the technology. The digital divide between partner firms will affect individual outcomes and benefits²⁹ and hence worth an inquiry while examining the impact of AI capabilities on SC performance. Studies on AI-related factors specifically need to empirically investigate how various AI implementation uses (exploitative and explorative use) influence SC resilience and SC efficiency. Based on OIP, our study seeks to establish connections between AI capabilities, SC resilience, and SC efficiency. This research aims to contribute to recent studies that have employed the Organizational Information Processing theory in the field of supply-chain management (SCM).^27,28,30 Our study aims to explore the following research problems:

1. How do AI capabilities (exploitative and explorative) influence SC outcomes (SC resilience and SC efficiency)?

2. Does the digital divide between SC partners affect the relationship between AI capabilities (exploitative and explorative) and SC outcomes (SC resilience and SC efficiency)?

The research is particularly important as it delves into the dual facets of AI capabilities—explorative and exploitative—and aids in our understanding of how organizations can navigate digital divide in supply chain management. By throwing light on the differential impact of AI capabilities and the digital divide, we contribute novel insights that can lead to strategic decision-making and technological investments in supply chain operations. To address the research problems, we collected survey data from owners and SC managers of small and medium distributor firms in India and tested the hypotheses using Smart-PLS 3 software. Our findings indicate positive effect of exploratory AI capabilities on SC resilience, negative effect of exploitative AI capabilities on SC resilience, positive effect of exploitative AI capabilities on SC efficiency. Regarding the moderating effects, our analysis confirms that the digital divide weakens the relationship between (1) explorative AI capabilities and SC resilience (2) exploitative AI capabilities and SC efficiency, (3) exploitative AI capabilities and SC resilience.

The structure of this paper is as follows: the next section reviews pertinent literature on the theoretical underpinning and the role of AI in supply chain management followed by the development of the conceptual framework. The next section outlines the methodology followed by the findings section. The paper concludes with a discussion section which delineates the implications for theory and practice and future research directions.

Literature review

Among the digital technologies reshaping production and the competitive landscape, AI has emerged as a foundational element of modern business operations, redefining how firms engage with technology.³² AI combines machine learning, big data, and deep learning, representing advanced IT systems and applications with diverse applications in managing distribution networks and business relationships.³³ Consequently, it is crucial to examine how leading companies develop strategies to harness AI’s potential to enhance supply chain performance.³⁴

Globally, the COVID-19 pandemic underscored the necessity for enhanced resilience within supply chains.³⁵ Supply chain resilience is defined as the capacity of a firm to withstand unforeseen and disruptive events, e.g., the COVID-19 pandemic and economic crisis, and quickly restore to its initial performance level or adapt to the changing market conditions to deliver expected firm performance.³⁶ According to the logic of dynamic capabilities, the use of data-driven decision-making through AI applications leads to the development of information processing capabilities within organizations.³⁷ Managers can synthesize and utilize this information to address uncertainties about supply and demand, which helps them better analyze and combine complex information from diverse sources.³⁸ Lack of such skills results in inventories or higher spending on developing supply chains and reduced profit margins.⁸ Thus, AI capabilities can facilitate real-time synchronization and collaborative inter-organizational relationships, ultimately improving SC efficiency and performance.³⁹ Nevertheless, as argued in the literature,⁸ research about the development of supply chain in the digital era is still in its early stages. It should be updated to reflect the impact of the evolving technological landscape and AI capabilities on SC resilience and performance,^40,41 and that becomes the central theme of this study.

OIP posits that firms require more information when dealing with uncertainty and higher levels of interdependence.³⁷ To accomplish tasks efficiently, a company must intentionally extract, transform, and distribute information for enhanced decision-making.⁴² Firms can do so by developing lateral relations or vertical information systems.¹¹ Integrating various Information Technology systems and capabilities enhances information processing capability and improved information flow and collaborative business planning among partners.^11,43

“Ambidexterity” is defined as a trade-off between exploiting certainty and existing capabilities and exploring new possibilities.^44,45 Specifically, exploration involves generating new knowledge, while exploitation focuses on refining existing knowledge.¹² Research indicates that understanding and building organizational ambidexterity and hence the digital capabilities make organizations more resilient and bring competitive advantage.²³ Big data and analytics capabilities are related to organizational routines and practices.^46,47 By changing routines and sources of data efficiency firms can solve current supply-chain issues and identify new opportunities.⁴⁶ Since there is a conflict between exploitative and explorative capabilities in an organization,⁴⁴ a firm needs to learn and maintain a balance between conflicting activities for long-term sustainability.⁴⁸ This implies that both explorative and exploitative capabilities are necessary for supply-chain management. Extending the literature on organizational ambidexterity, rooted in OIP, we introduce the concepts of explorative AI capabilities and exploitative AI capabilities. A company’s capacity to explore new AI resources and practices is referred to as explorative AI capability, and exploiting existing AI resources and practices is denoted as exploitative AI capability.⁴⁹ More specifically, explorative AI capability indicates a company’s aptitude to identify and implement novel AI resources to bolster its business processes and strategies. In contrast, exploitative AI capability refers to the firm’s capacity to refine and extend its existing AI resources to support business processes and strategy. Greater exploitation and exploration bring in SC resilience through agility and redundancy.⁵⁰ Apparently, development of ambidextrous capabilities leads to reconciliation of SC resiliency and efficiency, which leads further deliberation from the methodological and contextual perspective.⁵¹ Hence, our study is focused on comprehensively examining the influence of explorative and exploitative AI capabilities on SC resilience and efficiency.^7,52

When discussing AI capabilities, an important consideration is the digital divide concept. The term “digital divide” is a helpful metaphor describing the perceived disadvantage faced by those who either cannot or choose not to incorporate these technologies into their daily lives.^53,54 Commonly, digital divide involves disparities in physical infrastructural access and the actual use of digital technologies by individuals and companies. Companies, governments, and people across the globe are producing so much data that they cannot be accessed through traditional methods, rather they need analytical techniques to convert that data into meaningful insights.⁵⁵ Recognizing digital divide as a significant barrier to social inclusion, technological progress, and business opportunities, we explore the relationship between the digital divide and AI capabilities in SC management. This investigation is important because limited access to AI, big data, and analytics creates digital divide. Research shows that digital divide is strongly related to analytics and supply-chain management. A high level of internet access and AI expertise will lead to increased electronic trade among firms and make all operations safe, affordable, and safe.⁵⁶ Extant research highlights that digital divide involves at least three factors: the ease of accessing information, the utilization of information, and the capacity to absorb information, all linked to an organization’s AI capabilities.⁵⁷ Our current study considers the moderating effect of the digital divide among partner firms to evaluate its impact on supply chain performance.

Conceptual framework and hypotheses

Technology is pivotal in establishing connections between SC partners, offering a formalized and standardized language and streamlined information flow. This improved information flow strengthens their ability to process information, which enables them to respond quickly to uncertainties in the supply chain ecosystem.¹¹ OIP offers a theoretical framework to understand how firms employ AI capabilities to enhance SC outcomes (resilience and efficiency). We also differentiate whether implementation of AI is mainly related to exploration or exploitation. This study considers explorative AI capabilities with the objective of fostering new opportunities and expanding supply-chain networks, empowering firms to devise innovative solutions (search, innovation, experimentation) in collaboration with their SC partners to maintain long-term competitiveness.^58,59 This study posits that exploitative AI capabilities enhance the efficiency of SC network with refinement of existing AI assets, and execution of SC routines aimed at monetizing established capabilities.⁵⁹ As these different modes of AI capabilities are associated with varying organizational routines,⁶⁰ an investigation into their impact on SC outcomes may have important managerial implications.

Explorative AI capabilities

AI exploration involves pursuing novel problem-solving methods and the evaluation of their effectiveness.⁶¹ Explorative AI capabilities signify a firm’s proactive approach to continually seek new and advanced data-driven and analytical resources to capitalize on business opportunities using these resources and capabilities.⁴⁹ In this regard, research has emphasized the importance of supply chains relying on advanced technology to develop adaptive systems capable of addressing uncertainty and evolving conditions.^20,62 Explorative AI develops foresight and agility in SC systems.¹⁸ Consequently, it is anticipated that explorative AI capabilities will enhance SC resilience. Explorative capabilities have the potential to generate value through collaborative planning, allowing for the assimilation and transfer of new information to meet evolving collaboration needs.⁶³ Furthermore, explorative AI offers support for dynamic capabilities encompassing sensing, seizing, and transformation, which helps to mitigate the risks of actual disruptions and proactively address potential supply chain issues.⁶¹ This information and expertise promote responsiveness and are particularly valuable in dynamic environments.⁶⁴ Therefore, we posit the following hypothesis:

H1:

Explorative AI capabilities are positively related to supply chain resilience.

Supply chain operations have become increasingly data-driven, emphasizing information over physical assets such as inventory, warehouses, and transport equipment.⁶¹ Utilizing AI-powered information technology for exploratory purposes encourages thorough and extensive information sharing among SC partners, especially for unstructured tasks. AI algorithms aid in optimizing the development of efficient supply chain networks.⁸ The demonstrated potential of AI in improving the decision-making processes within supply chains indicates that AI can be leveraged to enhance the long-term performance and competitive edge of supply chains.⁶⁵ Consequently, we posit the following hypothesis:

H2:

Explorative AI capabilities are positively related to supply chain efficiency.

Exploitative AI capabilities

Exploitative AI capabilities denote a firm’s capacity to use existing AI resources to extend and enhance its business processes. However, SC resilience hinges on developing adaptive systems capable of accommodating the evolving conditions within the supply chain over time.²⁰ Evidence shows that exploitative AI capability negates the relationship between information sharing and SC performance.⁶⁶ One possible reason is that a firm concentrating on the exploitative AI capability to build information-sharing routines is focused on identifying opportunities within the shared information. Such firms may fail to develop new “cognitive maps” to accommodate novel meanings and interpretations of shared data and market information.^63,67 AI systems add layers of complexity to supply chains, which bring in more interdependencies and, if not managed well, can lead to cascading effects. Such interconnectivity and interdependence of AI-driven capabilities can increase risks and vulnerabilities in SC management, making them less resilient to disruptions.²⁰ AI exploitation may often indicate over-reliance on AI capabilities for decision-making which may lead to reduced human oversight and intervention. This over-reliance on existing AI systems and capabilities may reduce the possibility of supply chains adapting to uncertainty. Moreover, exploitative AI capabilities are optimized for specific tasks and scenarios which weaken their response to disruptions.⁶⁸ Often, a firm actively engaged in information sharing gains access to enormous amounts of data about products/services and customers, which could result in an overwhelming volume of information that is challenging to interpret, synthesize, and comprehend in changing environment.⁶⁹ Therefore, we propose that:

H3:

Exploitative AI capabilities are negatively related to supply chain resilience.

Exploitative AI capability is vital in enhancing collaborative planning by extending a firm’s dedicated AI systems and resources to facilitate the effective coordination of ongoing business processes.⁷⁰ More specifically, a firm can continually enhance and update its data management architectures, communication networks, and application portfolios by consistently leveraging its current data-driven and analytical resources in collaboration with channel partners.⁶³ Drawing upon OIP theory, these highly integrated systems break down organizational silos and enable firms to more efficiently share and recombine valuable knowledge within and across administrative boundaries. Consequently, exploitative AI capability aids in developing operational competence,¹² which, in turn, allows for more effective resource utilization from channel partners and the execution of collaborative planning. This ultimately leads to an efficient supply chain. As a result, we put forth the following hypothesis:

H4:

Exploitative AI capabilities are positively related to supply chain efficiency.

SC efficiency and SC resilience

A resilient supply chain exhibits a high degree of agility, denoting its ability to absorb changes resulting from disruptions and exhibit responsiveness amidst uncertainty.⁷¹ Resource efficiency can contribute to firms gaining a competitive edge. The impact of supply chain efficiency on enhancing a firm’s resilience stems from the collaborative capabilities of firms. The effective orchestration of the firm’s resources, including intangible assets like leadership, is a pivotal aspect of supply chain management efficiency, which, in turn, positively influences supply chain resilience. Hence, we propose the following hypothesis:

H5:

Supply chain efficiency is positively related to supply chain resilience.

Moderating effect of digital divide

The digital divide refers to the disparities in adopting and utilizing information communication technologies (ICT) across communities and regions. Disparities in internet access are more than just a function of individual capabilities or resources. The term “double Digital Divide,” coined by Chen and Wellman,⁷² describes the divide in internet access and utilization that is segregated by geographical locations and the socioeconomic status of individuals. Regarding the adverse impact of the digital divide on e-commerce, while the literature offers limited studies, there are substantial disparities in adopting electronic commerce, depending on the country’s type and the extent of its digital divide. The digital divide significantly affects technological factors, such as network access, availability of computer equipment, training in internet technology usage, and the practical benefits of its application. In line with this, Zouari et al.⁷³ argue that the digitalization of supply chains is closely associated with digital maturity and integrating digital supply chain tools. Their proposition suggests that digital maturity (or the divide in digital maturity) can positively or negatively influence supply chain resilience. Therefore, we hypothesize the following:

H6:

The digital divide between partner firms weakens the relationship between explorative AI capabilities and supply chain efficiency.

H7:

The digital divide between partner firms weakens the relationship between explorative AI capabilities and supply chain resilience.

H8:

The digital divide between partner firms weakens the relationship between exploitative AI capabilities and supply chain efficiency.

H9:

The digital divide between partner firms weakens the relationship between exploitative AI capabilities and supply chain resilience.

Figure 1 describes our proposed model.

Figure 1. Proposed model.

Research method

We follow a cross-sectional and correlational study design, given that data collection happened within a specified timeframe. A quantitative approach was employed to explore the connections among exogenous, moderating, and endogenous factors within small- and medium-sized distributor firms in India. Data collection was facilitated through self-administered questionnaires. Respondents were asked to rate their responses using a five-point Likert Scale, ranging from “strongly agree” (1) to “strongly disagree” (5). The survey items used for measuring the variables were sourced from existing literature. A summary of the construct items and their corresponding references is provided in the Appendix.

We first pretested the survey questionnaire with 45 SC managers and one academician working in industrial relationships and supply chain domain. Finally, we collected the data through a survey from a random sample of 1732 owners and SC managers of small and medium distributor firms in India, representing the manufacturing, electronics, cement, and metals industries. These managers were identified during a management development program held at author’s institute, which focused on sales and distribution management training for working managers. The participants came from various regions in India. We apprised them of the study and requested their consent to participate in it. For those who responded, we followed the snowball sampling to reach a wider but pertinent audience for the study, as large data collection is often a huge challenge. We received 267 completed surveys, which reflect a 15.4% response rate deemed suitable for such studies.⁷⁴ Table 1 shows the descriptive statistics.

Table 1. Descriptive statistics.

Characteristics of respondents (sample = 267)
Sector	Percentage	Company age	Percentage
Manufacturing	32%	1-5 years	25%
Cement	15%	5-10 years	35%
Electronics	20%	10-15 years	33%
Chemicals	11%	15-20 years	5%
Industrial tools	22%	>20 years	2%
Employee count	Percentage	Respondents’ job titles	Percentage
1-10	2%	Vice President (SCM, Operations,	33%
10-20	7%	Purchasing)
20-30	12%	SC Management Director	12%
30-40	33%	Operations Director	13%
40-50	21%	Purchasing Director	21%
>50	25%	SC Manager	21%

This study’s measurement and structural models were evaluated using Smart-PLS 3 software. PLS-SEM is particularly suitable for cases involving non-normally distributed data or small sample sizes.⁷⁵ We conducted a non-response bias analysis to ensure the data’s validity, following the methods outlined in literature.^31,76 Our findings indicated no significant differences between the early and late responses (with respective sample sizes of N = 127 and N = 140) in terms of industry, company age, firm size, or employee count (t = 0.55; p = .39; t = 0.64; p = .54; and t = 0.81; p = .61, respectively). This confirms that non-response bias is not a significant issue in the present study. Furthermore, we proactively tested for common method bias (CMB).⁷⁷ For the ex-ante analysis, we collected data from respondents with relevant knowledge of logistics and SC management (e.g., executives in operations and procurement). We ensured the anonymity of all respondents. Later, we conducted a post-hoc analysis to examine CMB using Harman’s⁷⁸ one-factor test. The unrotated exploratory factor analysis indicated five factors, giving 35.13% as the highest variance explained by a single factor. Therefore, we conclude that CMB does not pose a problem for further analysis.

Findings

In Table 2, we present the outcomes of the measurement model assessment. We evaluated the measurement model’s convergent and discriminant validity. All factor loadings exceeded 0.5, establishing their reliability. Additionally, all constructs exhibited a Cronbach’s alpha of >0.70, an average variance extracted (AVE) of >0.5, and a composite reliability (CR) of >0.60, confirming the convergent validity of the model.⁷⁹ In terms of discriminant validity, we applied the Fornell and Larcker criterion. The squares of the AVE (diagonal values) are greater than the correlations with other variables, as displayed in Table 3.

Table 2. Estimation of the measurement model parameters.

Construct/Items	Loadings	Composite reliability (CR)	Cronbach’s Alpha (α)	Average Variance Extracted (AVE)
AI exploration		0.89	0.72	0.63
AIExplor1	0.90
AIExplor2	0.89
AIExplor3	0.75
AI Exploitation		0.81	0.87	0.65
AIExplot1	0.72
AIExplot2	0.83
AIExplot3	0.71
AIExplot4	0.79
AIExplot5	0.81
AIExplot6	0.86
AIExplot7	0.91
SC resilience		0.92	0.90	0.75
SCR1	0.73
SCR2	0.74
SCR3	0.80
SCR4	0.88
SCR5	0.93
SC efficiency		.87	.81	.73
SCE1	.81
SCE2	.79
Digital divide		.83	.76	.76
DD1	.82
DD2	.75
DD3	.78
Control
Size	1	1	1	1

Table 3. Fornell and Larcker criterion for discriminant validity.

	AIExplor	AIExplot	SCR	SCE	DD
AIExplor	.75
AIExplot	.43	.72
SCR	.51	.43	.73
SCE	.44	.52	.57	.70
DD	.36	.69	.41	.48	.72

The structural model included two endogenous variables, SC resilience and SC efficiency, with corresponding R-squared values of 0.38 and 0.25, as presented in Table 4. These values establish the robustness and quality of the structural model. Furthermore, we evaluated the predictive relevance of the model, using Stone-Geisser Q2 values as the benchmark (0.13 for SC resilience and 0.19 for SC efficiency). Additionally, we conducted an overall quality assessment through standardized root mean square residual (SRMR).⁸⁰ In our case SRMR = 0.07, indicating a significant model quality. Also, the normed fit index (NFI) registered at 0.92 (>0.90) indicates a good model fit.⁷⁹

Table 4. Relevance and quality of the structural model.

Constructs	R^2	Q^Citation2
SC resilience	0.38	0.13
SC efficiency	0.25	0.19

Table 5 displays the outcomes of the PLS structural model, confirming the statistically significant positive association between Explorative AI capabilities and SC resilience, thus supporting H1 (β = 0.20, t = 3.2, p = .00). However, the findings do not demonstrate a statistically significant positive connection between Explorative AI capabilities and supply chain efficiency, leading to rejecting H2 (β = 0.39, t = 1.3, p = .30). Exploitative AI capabilities exhibit a statistically significant negative relationship with supply chain resilience, thus supporting H3 (β = −0.5, t = 43, p = .00). The results also affirm that Exploitative AI capabilities have a statistically significant positive relationship with supply chain efficiency, supporting H4 (β = 0.29, t = 2.7, p = .01). Furthermore, the findings reveal that SC efficiency is statistically significantly positively related to supply chain resilience, supporting H5 (β = 0.41, t = 2.2, p = .00). Therefore, only hypotheses H1, H3, H4, and H5 among the direct relationships receive support.

Table 5. PLS structural model.

Hypothesis		No.	Beta	SE	T value	P value	Inference
Direct effects
AIExplor → SCR		H1	.20	.05	3.2	.00	Supported
AIExplor → SCE		H2	.39	.03	1.3	.30	Not supported
AIExplot → SCR		H3	−.35	.04	4.3	.00	Supported
AIExplot → SCE		H4	.29	.06	2.7	.01	Supported
SCE → SCR		H5	.41	.05	2.2	.00	Supported
Indirect effects
AIExplor*DD → SCE		H6	.12	.02	1.2	.30	Not supported
AIExplor*DD → SCR		H7	.22	.01	3.1	.00	Supported
AIExplot*DD → SCE		H8	.27	.03	2.7	.01	Supported
AIExplot*DD → SCR		H9	.11	.04	3.7	.01	Supported

The research model posited that the digital divide moderates the relationship between exploratory and Exploitative AI capabilities and SC resilience and efficiency. We performed a two-stage calculation to arrive at the moderation test results recommended by Becker et al.⁸¹ to determine whether the moderating variable significantly influences the relationship between exogenous and endogenous variables. The results of the moderation analysis indicate that the digital divide moderates the relationship between Explorative AI capabilities and SC resilience (β = 0.22, t value = 3.1, p-value = .00), Exploitative AI capabilities and SC resilience (β = 0.11, t value = 3.7, p-value = .01), and Exploitative AI capabilities and SC efficiency (β = 0.27, t value = 2.7, p-value = .01). Consequently, hypotheses H7, H8, and H9 are supported. However, the moderating effect of the digital divide on the relationship between Explorative AI capabilities and SC efficiency is insignificant (β = 0.12, t value = 1.2, p-value = .30), and, therefore, H6 is not supported. These findings indicate that the digital divide significantly moderates the impact of Exploitative AI capabilities on SC resilience and SC efficiency and the effect of exploitative AI capabilities on SC resilience. This suggests that a higher level of digital divide between distribution partners hinders the development of a sustainable distribution network.

Discussion and conclusion

Digital transformation initiatives encompass the adoption and integration of technologies like AI and data-driven decision-making processes, which have a far-reaching impact on various aspects of businesses. AI capabilities can trigger dynamic changes influencing sales and distribution.⁸² This study aimed to investigate how exploitative and explorative AI capabilities affect SC resilience and SC efficiency, aligning with the literature gap on using data and information flows for decision-making processes in supply chain.⁸³ Furthermore, we explored the role of the digital divide between partner firms in explaining how they leverage the differential AI capabilities for expected SC outcomes.⁵⁶

Implications for theory

While literature has examined the potential of AI in supply chain management, there remains a gap in the theoretical explanation of AI capabilities and their impact on SC management.^41,83 Drawing from the emerging literature on OIP, we develop (1) a distinction between explorative and exploitative AI capabilities and (2) examine their differential impact on SC resilience and efficiency. Since the implications of AI implementation among supply chain partners is contingent on digital maturity, the digital divide among firms assumes a critical role in our understanding of SC resilience and SC efficiency through AI capabilities.⁵⁶

Rooted in the organizational information processing theory, we investigate the impact of AI capabilities (explorative and exploitative capabilities) on supply chain resilience and efficiency.⁸⁴ Our results align with the concepts proposed by the theory, which emphasize aligning AI capabilities with the information processing requirements for enhancing the robustness of supply chain networks.^42,69 Further, the unique and comprehensive description of AI capabilities developed in this study finds support from the literature that firms try to strike a balance with AI exploitation and AI exploration for operational excellence and decision-making.⁸⁵ Firms that can integrate exploratory and exploitative efforts gain long-term competitive advantage and success.^60,86 Thus, while studying AI capabilities as a concept, it gives a theoretical justification to study that in a holistic manner. No previous research has provided a model expanding the definition of AI capabilities (explorative and exploitative) to assess its impact on SC outcomes.

Based on survey data collected from small and medium-sized distributor firms in India, our findings indicate a positive relationship between exploratory AI capabilities and SC resilience. Based on the definition of exploratory AI capabilities, given earlier, this finding draws support from previous studies that predictive analytics can forecast disruptions in supply chain, which aids in decision-making, creating resilient supply chain networks.^8,87 However, contrary to the existing literature^8,65 explorative AI capabilities have no significant association with supply chain efficiency. This finding is aligned with organizational strategy literature which mentions that exploratory activities often lead to trial and error and may disrupt the established operations and not support efficiency.^88,89 While data confirmed that exploitative AI capabilities are negatively related to supply chain resilience, exploitative AI capabilities have a positive effect on SC efficiency. Highly optimized supply chains involving exploitative AI capabilities are often less adaptable to sudden shocks. In fact, a focus on efficiency can prevent the development of adaptive capabilities needed to respond to disruptions.^90,91 While exploitative AI capabilities such as optimizing current supply chain operations and standardizing processes lead to efficiency, it can also create vulnerabilities in the system where some failure in the supply chain can have cascading effects damaging resilience. Additionally, results reinforce the findings from the literature that SC efficiency has a positive impact on SC resilience. SC efficiency creates flexibility and responsiveness which are indicative of SC resilience.⁹¹ Efficient information flows in supply chains improves visibility and coordination among SC partners, bolstering resilience.⁹² Regarding the moderating effects of the digital divide, our analysis confirms that there is no significant impact of the digital divide between partner firms on the relationship between explorative AI capabilities and SC efficiency. Also, because the relation between explorative AI capabilities and SC efficiency comes out as non-significant. However, the digital divide does weaken the relationship between explorative AI capabilities and SC resilience. Additionally, the digital divide between partner firms weakens the relationship between exploitative AI capabilities and both SC efficiency and SC resilience. Partner firms with weaker digital maturity may lack the required resources and skills to fully utilize the exploratory AI tools and insights drawn. Digital technology adoption and integration affect SC resilience emphasizing collaboration and shared information.²⁰ Thus, the digital divide exhibits varying effects on AI capabilities and their impact on SC performance.

Implications for managers

The findings uncovered that there is indeed a connection between AI capabilities and supply chain outcomes.⁹³ We posit that a company’s supply chain results rely on how extensively the company engages in exploring novel and advanced AI capabilities and exploiting existing AI capabilities. Explorative capabilities can lead to innovation and the discovery of new SC opportunities, while exploitative capabilities focus on optimizing current SC operations. The choice and balance between these capabilities depends on a firm’s strategic goals.²⁰ For example, in the logistics and transportation sector, explorative AI can be applied to explore novel route optimization algorithms, which may consider real-time traffic data and other dynamic factors to improve delivery efficiency.⁹⁴ Explorative AI can help identify and evaluate potential new suppliers by analyzing a wide range of data sources, including supplier financials, reputation in the market, and social responsibility practices. Companies frequently employ exploitative AI to fine-tune their inventory management, relying on historical data and demand trends.⁹⁵ AI-powered inventory systems continually adjust reorder thresholds and safety stock levels to minimize holding expenses while guaranteeing product availability. Further, it can help oversee supplier performance and streamline procurement processes.

Resilience necessitates sensing and seizing opportunities to adapt current business models to the changing external environment. It has been observed during the COVID-19 pandemic that firms that could sense and analyze data for the rapid and widespread market disruptions and were quick to resolve were in a better position to sustain the business.³⁵ As managing disruption requires companies to be more agile and ramp up or ramp down their SC network, explorative AI capabilities are more useful in such scenarios and promote SC resilience.^9,12 For example, it can detect early warning signs of political unrest in a region where a supplier is located, enabling organizations to take preventive measures or diversify their supplier base. Explorative AI can identify opportunities for dynamic inventory strategies, such as postponement or localized production, to reduce the reliance on centralized warehouses and increase flexibility in the supply chain.²⁰ However, SC efficiency requires collaboration, combined decision-making, and a shared understanding among partner firms. A company that solely depends on its exploratory AI capabilities for collaborative planning might struggle to fully harness the benefits of collaborative operations due to its ongoing experimentation with new AI resources, potentially hindering the promotion of supply chain efficiency.⁹⁰

Exploitative AI capabilities also affect the building of resilient and efficient SC systems. Resiliency requires fathoming new information and creating new systems and requires an innovative, game-changing mind-set; hence, exploitative AI capabilities negate building resilience SC systems.⁶⁸ Leveraging a firm’s existing AI resources may not lead to creating new “cognitive maps” capable of accommodating fresh interpretations of shared knowledge.^37,41 While enhancing efficiency, this approach may not necessarily contribute to improved resiliency. By continuously refining and optimizing processes based on historical data and patterns, organizations can reduce carrying costs and stockouts, making their supply chains more competitive. Thus, the findings confirm that varied AI capabilities—explorative and exploitative—have different implications for supply chain partners.

The digital divide plays a moderating role in influencing AI capabilities’ impact on SC resilience and efficiency. In general, most supply chain decisions are influenced by individuals, and firms may be hesitant to share information, even with long-term partners, due to the profound consequences it can entail, coupled with a lack of technological resources and skills to facilitate communication among firms.⁹⁶ In a country like India, AI adoption is slow, so firms cannot cope when their upstream partners have a higher digital maturity.⁹⁷ A lack of digital tools and technologies can limit supply chain visibility. Companies might encounter challenges in monitoring the flow of products, managing inventory, and promptly addressing disruptions. In areas with limited internet connectivity or digital literacy, such as rural areas, communication with suppliers, customers, and logistics partners may be hindered. This can result in delays, misunderstandings, and difficulty coordinating supply chain activities.^98,99 Businesses that lack digital capabilities may face administrative hurdles and delays in compliance, affecting their supply chain operations.

Limitations and future research directions

The study presents an enhanced understanding of AI capabilities, the digital divide among SC partners, and SC performance rooted in organizational information processing theory. Like any study, our research is associated with certain limitations that open avenues for future research. First, we employ a cross-sectional design and concentrate on the context of small and medium-sized distribution firms within India. Second, our data collection is restricted to a single time point due to the unavailability of longitudinal data required for assessing causality over an extended duration. Third, we focus on supply chain (SC) efficiency and resilience as the dependent variables. However, we can delve into more advanced constructs, such as SC viability. Subsequent research in different regions may offer insights into commonalities and disparities within alternative settings. Therefore, conducting longitudinal investigations could yield valuable perspectives on the interplay between evolving learning processes, emerging AI capabilities, and the performance of supply chains regarding their resilience and efficiency. Some possible research questions could be:

1. How do AI capabilities affect trust and commitment among the SC partners in an SC ecosystem?

2. Do AI capabilities lower the opportunistic behaviors of various SC partners?

3. What types of AI capabilities lead to enhanced SC network performance?

4. What firm and policy-level measures are needed to minimize the digital divide among SC partners?

5. What are the negative implications of the digital divide on the SC ecosystem’s business performance?

6. How can AI capabilities be used as supplier selection criteria and to formulate partner development programs?

Acknowledgments

We thank the reviewers and EIC for their constructive comments, which helped to improve the content and presentation of the paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Indian Institute of Management Lucknow [seed money grant number SM-269].

Appendix.

Measure items

Table

Construct	Items/Measured Questions (5-point Likert scale)	References
AI Exploration	Our firm constantly acquires new AI resources (e.g.,a new generation of AI architecture, potential AI applications, and critical AI skill).	66,100
	Our firm constantly experiments with new AI resources.
	Our firm constantly experiments with new AI management practices.
AI Exploitation	Our firm focuses on utilizing existing AI resources to meet current business needs(e.g., current business models, business processes, and AI service capability).	66,100
	Our firm focuses on reusing existing AI applications and services
	Our firm focuses on reusing existing AI components(e.g., hardware and network resources).
	Our firm focuses on reusing existing AI knowledge and skills.
	Please indicate the extent to which you use AI to communicate with your primary buyer for the following:
	1: Order processing, invoicing and settling accounts
	2: Exchange of shipment and delivery information
	3: Managing warehouse stock and inventories
Supply Chain Resilience	Our firm’s supply chain is well prepared to face constraints of supply chain disruptions	41,101
	Our firm’s supply chain can rapidly plan and execute contingency plans during disruptions
	Our firm’s supply chain can adequately respond to unexpected disruptions by quickly restoring its product flow
	Our firm’s supply chain can swiftly return to its original state after being disrupted
	Our firm’s supply chain can gain a superior state compared to its original state after being disrupted
Supply Chain Efficiency	On a scale from 1 to 5, please rate how you perceive the performance of your supply chain in terms of the following metrics:	102
	Supply chain cycle efficiency Meaning: Percentage of time in which inventory (raw materials/WIP/finished products) is active/moving in the supply chain over total time spent in the supply chain
	Supply chain flexibility Meaning: Average time required for the supply chain to respond to an unplanned 20% increase in demand without service or cost penalty
Digital Divide	Compared to your upstream partner, on a scale from 1 to 5, please rate how you perceive your firm in terms of the following metrics:	103
	Information accessibility is high
	Information utilization is high
	Information receptiveness is high