How Does Artificial Intelligence Shape Supply Chain Resilience? The Moderating Role of the CEOs’ Sports Experience

Abstract

In the volatility, uncertainty, complexity, and ambiguity (VUCA) environment, the application of artificial intelligence (AI) technologies is a key engine for shaping supply chain resilience (SCR). This study employs the entropy method to develop an evaluation index system for SCR, incorporating two key dimensions: resistance and recovery capacity. Using a sample of Chinese-listed enterprises from 2009 to 2022, this study reveals that AI significantly enhances SCR, and CEOs’ sports experience can positively moderate the association between AI and SCR. Mechanism examination shows that AI promotes SCR through operational efficiency optimization, information, and knowledge spillover in the supply chain. Heterogeneity analysis reveals that the positive impact of AI is more significant in firms with a high-skilled labor force, firms with high heterogeneity of the executive team’s human capital, high-tech industries, and regions with strong digital infrastructure. Moreover, the AI application has a diffusion effect on the upstream and downstream enterprises of the supply chain, improving AI adoption levels. Our research not only augments the existing literature on the economic ramifications of AI adoption and the strategic value derived from CEOs’ extramural experience but also offers both theoretical frameworks and empirical insights for executive recruitment and fortifying SCR.

1. Introduction

Currently, the world is experiencing unprecedented changes at an accelerated pace. Crisis events with far-reaching impacts, such as pandemics, geopolitical conflicts, and escalating trade disputes, are occurring frequently. A VUCA environment characterized by volatility, uncertainty, complexity, and ambiguity has gradually become the norm, posing more disruption risks to global supply chains. A typical example is the COVID-19 pandemic in 2020, during which 94% of the Fortune 1000 companies experienced supply chain disruptions following the global public health crisis [1]. In 2023, the United States successively established the “Indo-Pacific Economic Framework for Prosperity” supply chain agreement [2] and formed the “White House Council on Supply Chain Resilience (SCR)” [3]. Improving SCR has become the focus of a new round of corporate competition and great power games. SCR mainly represents the capacity of a supply chain to recover to a normal or even more efficient state after facing market risk disruptions [4]. As the micro subjects of economic operation, enterprises undoubtedly face unprecedented supply chain disruption risks in the VUCA environment. Identifying effective means to enhance SCR is critical to achieving sustainable business development and a smooth global economic cycle.

Artificial intelligence (AI), as a revolutionary general-purpose technology, offers a new approach to shaping SCR in the era of the digital economy. AI mimics human intelligence, allowing computer systems to carry out tasks that resemble human thinking and decision-making processes [5]. According to the 2025 Top 10 Supply Chain Trends Report released by the Association for Supply Chain Management, AI ranks first, becoming the most influential supply chain trend in 2025 [6]. Supply chain management in the VUCA environment is increasingly characterized by diversification and complexity, leading to higher disruption risks and more urgent resilience-building needs. AI’s ability to think and act with both technological rationality and human sensibility can fulfill the need for SCR shaping [7]. Through deep learning, automated decision-making, and visualization, AI has the potential to enhance different areas of the supply chain, such as demand forecasting, logistics, production, manufacturing, warehousing, and the design of supply chain networks [8,9,10], driving the intelligent development of the supply chain. Studies related to the association between AI and SCR have primarily concentrated on the theoretical aspects, establishing research frameworks through a literature review [11,12] and interviews [13]. From a quantitative perspective, only a few scholars have used questionnaire analysis and structural equation modeling to confirm the positive influence of AI on SCR [14,15]. Therefore, the existing research lacks a large-sample empirical test on the question of whether AI can enhance SCR.

Integrating AI into supply chain management is essentially a strategic change for the organization. The upper echelons theory emphasizes that the personal traits and psychological structure of executives affect firms’ strategic decisions and organizational performance [16]. Many studies on CEO traits and strategic change are based on this theory, focusing on how traits like regulatory focus [17], temporal focus [18], and overconfidence [19] impact strategic change. Other studies explore the role of CEO career variety [20], international experience [21], and prior board experience [22] in shaping strategic change. However, the specific effects of CEO traits on the economic outcomes of firms adopting AI technology remain unclear. Unlike traditional innovation activities, supply chain intelligence centers on applying AI technologies to manage uncertainty and risk. The resilience and competitiveness shaped by a CEO’s sports experience can drive them to adopt AI technologies proactively. In practice, many successful AI company CEOs who have made a global impact have sports experience. For example, Meta’s CEO Mark Zuckerberg enjoys long-distance running and combat training, while Microsoft’s CEO Satya Nadella is a cricket enthusiast. From a theoretical perspective, sports experience—as an off-the-job activity—helps CEOs develop openness and emotional stability through physical training and teamwork [23]. These qualities enable them to handle uncertainty, giving them an advantage in overcoming strategic inertia [24]. CEOs with sports experience are often more open to ideological innovation and are more likely to encourage the adoption of AI technologies in supply chain management.

Employing data from China, this paper investigates the following questions: (1) Does AI affect SCR, and how? (2) Does CEOs’ sports experience moderate the association between AI and SCR? Our study began by performing panel data regression to explore how AI applications impact SCR. Then, our study constructed the interaction term to examine the moderating effect of CEOs’ sports experience. To explore specific mechanisms, this paper discusses the mediating role of operational efficiency optimization, information, and knowledge spillover. Heterogeneity analysis was also conducted based on the corporate human capital, industry characteristics, and digital infrastructure. Moreover, our research further examined the diffusion effect of AI in the supply chain network.

The primary contributions of our research are outlined as follows.

First, our study confirms the positive role of AI in shaping SCR through large-sample empirical research, providing strong evidence for how AI affects SCR. Currently, studies examining the impact of AI on SCR mainly adopt qualitative interview methods [13] and empirical survey methods [14,15], lacking large-sample empirical tests. On the one hand, our study employed the entropy method to construct the indicator of SCR from the dimensions of supply chain resistance and recovery. On the other hand, this study employed machine learning and text analysis methods to construct AI indicators for companies, establishing the groundwork for future empirical research on AI and supply chains. Additionally, our findings open up the “black box” of the mechanism. This paper overcomes the limitations of existing research on the indirect effects of AI on SCR [25], revealing the internal mechanisms by which AI influences SCR, including operational efficiency optimization, information spillover, and knowledge spillover, thereby deepening the understanding of the relationship between AI and SCR.

Second, this paper identifies the positive moderating effect of the CEOs’ sports experience on the association between AI and SCR, enriching the research regarding the influence of CEOs’ off-the-job experience on corporate performance. Existing research on the economic consequences of AI at the firm level overlooks the impact of human cognitive factors on the application process of AI technology [26]. Although the upper echelons theory emphasizes that the characteristics of top managers influence organizational performance, its framework has insufficient attention to CEOs’ off-the-job experience [27]. In practice, despite differences among successful global CEOs in various aspects, they exhibit common sports-oriented leadership traits. CEOs with early sports experiences help develop a positive attitude and confidence in teamwork and self-management, build psychological resilience, and cultivate flexible psychological adaptability [28]. These traits align with the demands of VUCA environments on leaders. CEOs’ sports experience shapes their positive personality traits and improves their cognitive understanding of AI, which further influences the value creation of AI applications in the supply chain. Our research explores the moderating role of the CEOs’ sports experience, which not only expands the application scope of the upper echelons theory but also provides a new research perspective on the economic consequences of firms’ adoption of AI technologies.

Third, our research uncovers the bidirectional diffusion effect of the core enterprises’ AI applications on the AI capabilities of upstream and downstream supply chain enterprises. Existing research on the economic effects of corporate AI applications has mainly focused on their impact on the core enterprise itself, overlooking the potential external effects of AI applications. Our research firstly constructs matched sample data of companies and their suppliers, as well as companies and their customers, on a 1:1 basis. Then, our study uses AI-adopting companies as the diffusion source and finds that they have a spillover effect on the AI capabilities of upstream and downstream supply chain enterprises. This paper identifies more positive outcomes of AI in empowering SCR, enriching the study of the spillover mechanisms of core enterprises’ AI applications in the supply chain, and expanding the boundaries of research on the economic consequences of AI.

2. Literature Review and Hypothesis Development

2.1. Economic Effects of AI

In the era of the digital economy, AI, with its powerful data processing and intelligent decision-making capabilities, is gradually penetrating various sectors of economic development. The economic effects of AI applications have become a focal point of academic research.

At the macro level, existing studies have examined the impact of AI development on economic growth, the labor market, and income distribution. Regarding economic growth, numerous empirical studies confirm that AI, as a transformative yet emerging digital technology, plays a key role in driving sustained economic expansion [29,30]. However, some scholars argue that AI adoption may not always lead to increased productivity, showing that AI can have non-linear [31] or even negative effects [32] on economic growth. In the labor market, AI applications in production can replace low-skill, routine jobs, affecting workers across various occupations, wage levels, age groups, and education backgrounds [33,34]. Moreover, new AI-driven technologies can create new types of employment, encourage more flexible labor relationships, and blur the boundaries of traditional employment, offering workers more platform-based job opportunities [35]. Regarding income distribution, some scholars believe that AI’s substitutive nature will affect the distribution of production factors and workers’ incomes, potentially worsening income inequality [36,37]. On the other hand, other studies argue that AI’s application will boost labor productivity, increase national income, stimulate demand, and create new jobs, which could slow or even prevent the widening of income disparity [38].

At the micro level, researchers have focused on how AI affects individual business entities. AI technologies, such as industrial robots and machine learning, are widely used in areas like manufacturing [39], supply chain management [40], customer service [41], and decision support [42] within companies. Studies have shown that AI applications can enhance employment [43], foster innovation [44,45], improve export performance [46], and increase energy efficiency [47]. In terms of corporate non-financial performance, AI also plays a crucial role in reducing greenwashing practices [48] and improving ESG performance [49]. However, not all companies can effectively use AI to create unique economic value. A company’s ability to leverage AI for its competitive advantage depends on how it integrates AI into its operations and positions it within the company’s business model [50,51].

2.2. Influencing Factors of SCR

The concept of “resilience” originates from materials science, where it refers to a material’s ability to return to its original form after deformation, without exceeding its capacity limit [52]. While there is some disagreement about the definition of SCR, most scholars emphasize a company’s ability to respond to and recover from supply chain disruptions [53,54]. Existing research mainly examines the factors influencing SCR from the perspectives of resources and capabilities.

From the resource perspective, companies can mitigate the impact of supply chain disruptions by effectively utilizing their resources. Key corporate resources include digital resources and partnership resources. In terms of digital resources, adopting technologies such as big data, blockchain, and cloud computing allows companies to detect and address risks through data analysis and information processing, thereby enhancing SCR [55,56,57]. Regarding partnership resources, forming strong collaborative relationships among supply chain members can improve information sharing and risk management, thus boosting SCR [58]. Additionally, supply chain integration enables businesses to quickly adapt to customer demand and adjust production plans, strengthening SCR and driving performance growth [59].

From the capability perspective, SCR is reflected in a firm’s dynamic capabilities to respond flexibly and quickly to supply chain disruptions. These dynamic capabilities include innovation, supply chain visualization, collaboration, and agility. Regarding innovation capability, a strong innovation capacity helps companies share knowledge or exchange information with suppliers and customers, improving their flexibility and responsiveness in crises [60,61]. Supply chain visualization refers to the ability to visualize supply and demand, which provides more comprehensive information. With this capability, businesses can analyze changes in upstream and downstream processes in the supply chain, taking timely and targeted actions to ensure rapid recovery after a disruption [62,63]. As for collaboration, stronger cooperation among supply chain members fosters shared trust, facilitates synchronized decision-making, and improves the ability to respond to disruptions effectively [64]. Finally, supply chain agility allows businesses to react quickly and choose appropriate strategies to minimize the negative effects of disruptions, leading to faster recovery [65,66].

2.3. AI and SCR

The application of AI in supply chain management is transforming industry operational models. By using advanced algorithms and machine learning technologies, AI can forecast demand, accurately analyze market trends and consumer behavior, and improve inventory management. This reduces instances of stockouts and overstocking. In research on factors influencing SCR, dynamic capability theory and information processing theory are frequently applied [67,68].

From the perspective of dynamic capability theory, SCR can be seen as a combination of several dynamic capabilities. Dynamic capability theory emphasizes that companies must sense external opportunities and threats, adjust their dynamic capabilities accordingly, and maintain alignment with changes in the external environment. SCR specifically addresses scenarios where supply chain disruptions are caused by external shocks in a rapidly changing environment, creating a natural link between the two [69]. Key corporate dynamic capabilities related to SCR include flexibility and agility. First, in terms of flexibility, AI can enhance the flexibility of supplier selection. By analyzing historical supplier data such as delivery timeliness and quality control, AI helps companies choose the most reliable suppliers [70]. AI also improves flexibility in logistics transportation. In warehouse management, AI can analyze the movement of materials through aisles and recommend optimized floor layouts to speed up inventory retrieval and transportation [71]. In logistics management, AI can monitor traffic patterns, weather forecasts, and real-time road conditions to adjust delivery plans and allocate transportation resources efficiently [10]. Second, in terms of agility, AI helps companies adjust resources and capabilities dynamically by analyzing historical data and building predictive models. Machine learning technology can learn from market and operational data to predict demand fluctuations, optimize production plans, and prevent inventory buildup [40]. This not only improves resource allocation but also speeds up decision-making, enabling companies to adapt more quickly to market changes.

The core concept of information processing theory is aligning an organization’s information processing needs with its capabilities to achieve optimal performance. Information processing theory helps explain how supply chain visibility and uncertainty minimization are achieved through information processing [72]. AI technologies can provide real-time information on supply chain orders, inventory, and transportation, thereby enhancing supply chain transparency. Computer vision technology, supported by advanced learning algorithms, can intelligently recognize and classify complex data in images, optimizing quality control processes on production lines. It also enables precise inventory control by tracking items intelligently [73]. Natural language processing technology can automate contract review and management, reducing human errors and helping businesses make better decisions. AI can also enhance collaboration among supply chain members, improving the overall information processing capacity of the supply chain and effectively managing uncertainty. Companies can use the connectivity features of AI technologies to establish widespread connections among supply chain nodes, improving communication efficiency and transforming partnerships into “strong connections” [13]. For example, downstream supply chain companies, which are closer to consumers, can use AI to capture personalized customer demands and transmit this information upstream. The production side can then use personalized customization and intelligent distribution technologies to meet consumer needs more precisely.

In summary, enterprises using AI technology can effectively strengthen flexibility, agility, visibility, and supply chain cooperation, which in turn positively affects SCR.

Building upon the preceding analysis, the subsequent hypothesis is posited:

Hypothesis 1.

Ceteris paribus, AI applications can enhance firms’ SCR.

2.4. The Moderating Role of CEOs’ Sports Experience

The Strategic Change Cognition Theory emphasizes that managers’ perceptions of the firm’s internal and external environments are key factors influencing strategic change in the firm [74]. Managers’ perceptions will profoundly affect the organizational production and managerial activities of the firm when there is a major technological update. Therefore, whether AI can shape SCR not only depends on the degree of advancement of AI technology but is also closely related to the awareness of human users towards AI [26].

The combination of AI and supply chain management represents a strategic transformation that brings opportunities as well as challenges, which inevitably require strong involvement and support from the CEO. The upper echelons theory also suggests that managerial traits influence the strategic choices of an organization [27], while managers’ perceptions are influenced by factors such as their educational background, professional experience, and values, thus showing individual differences [75]. In complex and changing business environments, successful leaders need to build confidence in high-pressure situations and manage change based on continuous competition. These behavioral traits share similarities with top athletes [28]. Sports experiences can enhance the construction of CEOs’ psychological capital through challenging physical limit states, which in turn influences their attitudes toward AI. This paper further focuses on the moderating role that CEOs’ sports experience plays in the process of firms utilizing AI to shape SCR.

CEOs’ sports experience contributes to shaping their emotional stability and openness [76]. First, emotional stability is closely linked to the emotional experiences gained through sports. As sports involve competition and contests, individuals experience a wide range of emotions, such as excitement, tension, loss, joy, and frustration. The more sports experience a CEO has, the stronger these emotional experiences become. Engaging in sports also helps individuals recognize and regulate their emotions. Through physical challenges, individuals receive rich sensory stimulation, which enhances the regulation of their nervous system, allowing them to better control their emotions. Moreover, sports are effective in alleviating negative emotions. Second, openness is shaped by an individual’s tolerance for frustration and self-confidence. Sports offer valuable opportunities for developing frustration tolerance. In sports, individuals often face repetitive training or conflicts in team settings. Overcoming these obstacles enhances one’s ability to handle stress. Additionally, sports help build self-confidence. Through participation, individuals receive affirmation from others, which is internalized as self-affirmation over time. Team sports, like soccer and basketball, also foster qualities such as solidarity and a drive for excellence. The enhancement of self-confidence and frustration tolerance encourages individuals to embrace multiple possibilities.

When CEOs have sports experience, their mental traits can strengthen the positive impact of AI on SCR. On the one hand, the emotional stability that CEOs gain from sports helps them handle the challenges of integrating AI into supply chain management [77]. AI is inherently complex, and when building an AI-powered supply chain, companies will face many unfamiliar situations. Only CEOs with high emotional stability can effectively navigate these challenges. They remain optimistic in dynamic situations, make clear judgments about AI integration, and lead their teams to face challenges head-on, thereby accelerating the adoption of AI in the supply chain. On the other hand, CEOs with a high level of openness are more inclined to explore AI applications within the organization [24]. Individuals with high openness tend to have a stronger imagination, aesthetic sensibility, and emotional intelligence, making them more willing to challenge the status quo and embrace innovation. Their strong imagination and discernment help them predict the future potential of AI while evaluating the risks involved in AI integration. Their appreciation for change enables them to see the value that AI brings to the supply chain, while their open-mindedness allows them to overcome organizational inertia and embrace AI more readily.

Building upon the preceding analysis, the subsequent hypothesis is posited:

Hypothesis 2.

Ceteris paribus, CEOs’ sports experience can strengthen the positive impact of AI applications on SCR.

2.5. Mechanism Analysis

The supply chain is an integrated system of logistics, capital flow, and information flow across enterprises. There are interconnections between upstream and downstream enterprises of the supply chain in terms of information, knowledge, and social resources. The application of AI enhances the operational efficiency of node enterprises within the supply chain. In addition, it strengthens the information and knowledge spillover effects among supply chain enterprises, thereby positively influencing SCR. Specifically, this study argues that AI improves SCR through the following three channels.

The first channel is the operational efficiency optimization effect. According to supply chain management theory, effective planning, coordination, and management of the flow of materials, information, and financial transactions between enterprises in the supply chain can significantly improve operational efficiency [78]. AI plays a crucial role in creating a collaborative ecosystem. By utilizing advanced demand forecasting and real-time data-sharing, AI breaks down data silos. This allows upstream and downstream enterprises to communicate promptly across the entire process, from raw material supply to product sales. As a result, product demand, order delivery, delivery status, and inventory information can be shared, alleviating the “bullwhip effect” and improving overall supply chain operational efficiency [79]. The efficient operation of the supply chain also enhances the prediction and recovery capabilities of enterprises. On the one hand, higher operational efficiency enables real-time tracking and monitoring of the supply chain status, helping companies identify and predict potential risks. On the other hand, when faced with unforeseen events, companies can leverage their increased operational speed to adjust inventory and production plans promptly, improve the flexibility of purchasing, manufacturing, and delivery, reduce the risk of capital chain disruptions, and effectively prevent losses caused by supply chain interruptions [80].

The second channel is the information spillover effect. AI expands the communication range of supply chain enterprises, increases the connectivity and richness of the information network, and accelerates the spread of information within the network [81]. Through this information spillover effect, AI helps enterprises achieve better supply and demand matching, thereby enhancing the resilience of the supply chain. On the one hand, AI facilitates supply and demand matching by reducing the cost of information processing and transmission between node enterprises. This alleviates information asymmetry and improves coordination between upstream and downstream companies [72]. On the other hand, AI helps stabilize supply–demand relationships within the supply chain. By analyzing factors such as pricing, purchasing history, and sustainability, AI enables companies to select commercial partners more effectively, reducing the likelihood of contracting with businesses that may default [82]. Partnerships formed through this selective screening tend to result in more stable and reliable supply–demand relationships.

The third channel is the knowledge spillover effect. Mutual learning and communication between upstream and downstream supply chain firms, as well as the use of written contracts, can trigger knowledge spillovers. When AI is integrated into supply chain management, it initially takes over high-frequency, mechanical, and repetitive tasks [83]. This allows upstream and downstream enterprises to enhance their risk resistance and innovation recovery by learning from the production methods and innovation models of companies with more advanced AI applications [81,84]. From a risk resistance perspective, AI’s knowledge spillover effect substitutes skilled labor for unskilled labor across the supply chain, concentrating high-end expertise within the system. This increases the success rate of technological R&D in key areas, enabling a quicker response to supply chain disruptions. In terms of innovation recovery, AI’s ability to analyze and synthesize fragmented information helps formalize and disseminate tacit knowledge. Knowledge spillover significantly reduces the cost of knowledge transfer and technological innovation for upstream and downstream supply chain companies. As a result, these companies can borrow cutting-edge knowledge from AI-advanced firms at a low cost, allowing them to engage in incremental innovation and build a more sustainable supply chain ecosystem.

Building upon the preceding analysis, the subsequent hypotheses are posited:

Hypothesis 3a.

AI applications can promote firms’ SCR through the operational efficiency optimization effect.

Hypothesis 3b.

AI applications can promote firms’ SCR through the information spillover effect.

Hypothesis 3c.

AI applications can promote firms’ SCR through the knowledge spillover effect.

Figure 1 presents our theoretical framework to illustrate the hypothesized relationship in the graph.

3. Research Design

3.1. Sample

Our study selects China’s A-share listed companies in the Shanghai and Shenzhen Stock Exchange from 2009 to 2022 as the initial sample. Then, the sample is filtered as follows: (1) exclude companies in the finance and insurance industries; (2) exclude the specially treated companies with ST, *ST, and PT; (3) remove companies with an asset–liability ratio greater than 100%; and (4) eliminate companies with a missing value. We finally obtained 11,892 firm-year observations. All continuous variables are winsorized at the top and bottom 1%.

3.2. Variables

3.2.1. SCR

SCR consists of the ability of the supply chain to resist and recover from external shocks [4,85].

Supply chain resistance (SCR) represents the stability of the supply chain’s operational condition. From the “process” perspective, the stability of supply chain relationships depends on the capital occupied by downstream customers with upstream suppliers. Referring to related research [86], we measure capital occupation using the natural logarithm of the ratio between accounts receivable and revenue (RESIS1).

From the “results” perspective, whether the supply chain relationship is stable is also reflected in the continuity of the supply–demand relationships between companies. Following related research [87], this paper measures the stability of supply chain relationships by dividing the number of the company’s top five customers that appeared in the previous year by five (RESIS2).

Supply chain recovery (RECOV) refers to the ability of the supply chain to swiftly return to its original path after being disrupted by external factors. From the “supply and demand” perspective, external shocks to the supply chain cause an imbalance between supply and demand in the short term. Referring to related studies [88], our research uses the degree of deviation of production fluctuations from demand fluctuations as the first indicator (RECOV1) for supply chain recovery.

From the economic perspective, when the supply chain is affected by an external shock, corporate performance deviates from the expected trajectory and gradually adjusts to its pre-shock state. The degree of deviation in corporate performance reflects the recovery ability of the supply chain after the shock. Our research constructs Equation (1) and uses the residuals to capture the changes and responses of corporate economic performance over different periods (RECOV2). The larger the value, the stronger the supply chain recovery.

$$Perfor{m}_{i,t}=\alpha +{\beta }_{1}SIZ{E}_{i,t}+{\beta }_{2}LE{V}_{i,t}+{\beta }_{3}GROWT{H}_{i,t}+{\beta }_{4}AG{E}_{i,t}+{\beta }_{5}BOAR{D}_{i,t}+{\displaystyle \sum Frim}+{\displaystyle \sum Year}+{\epsilon }_{i,t}$$

(1)

where Perform represents corporate economic performance, quantified by the ratio of earnings before interest and taxes (EBIT) to the number of employees. The control variables include SIZE, LEV, GROWTH, AGE, and BOARD.

Finally, we apply the entropy method to compute a composite score for four indicators (RESIS1, RESIS2, RECOV1, RECOV2) across the two dimensions of supply chain resistance and recovery, with the resulting value representing supply chain resilience (SCR), as shown in

Table 1. Evaluation index system for SCR.

Decision Level	First Dimension Level	Second Dimension Level	Interpretation	Weight
Supply chain resilience (SCR)	Supply chain resistance (RESIS)	The capital occupied by downstream customers with upstream suppliers	The natural logarithm of the ratio between accounts receivable and revenue	0.0064
		The continuity of the supply–demand relationships between companies	The number of the company’s top five customers that appeared in the previous year divided by five	0.9711
	Supply chain recovery (RECOV)	Imbalance between supply and demand in the short term	The degree of deviation of production fluctuations from demand fluctuations	0.0005
		The degree of deviation in corporate performance	Equation (1)	0.0221

.

3.2.2. AI Application

The existing research mostly uses the penetration of industrial robots to measure corporate AI applications [89]. However, industrial robots cannot fully reflect the AI technology used by companies. The annual reports of listed enterprises contain relevant information about the adoption of AI technologies. Drawing on Song et al. (2025) [45], our research collects textual data from corporate annual reports and employs machine learning to generate an AI dictionary containing 73 keywords. Then, we increase the number of AI-related keywords in the annual report by 1 and take the natural logarithm as the corporate AI application indicator (AI).

3.2.3. CEOs’ Sports Experience

This paper uses whether the CEO has participated in sports or worked in sports-related fields as the indicator of CEOs’ sports experience (SPORTS). If the CEO has such experience starting from a certain year, considering the continuity of corporate behavior, SPORTS equals 1 for that year and all subsequent years; otherwise, it is set to 0. Specifically, our study employs text analysis to collect data on whether the CEO has participated in sports. We use Python 3.7 to scrape all relevant news about CEOs’ involvement in sports. The scraped keywords include 42 sport-related terms such as basketball, running, and marathon. Whether the CEO has worked in a sports-related field is determined through an analysis of the CEO’s resume provided by the annual report.

3.2.4. Other Variables

According to the existing literature [45,89], the following control variables are added: size of the company (SIZE), financial leverage (LEV), company age (AGE), return on assets (ROA), the growth rate of revenue (GROWTH), R&D investment (RD), board size (BOARD), percentage of shares owned by the largest shareholder (Top1), proportion of independent directors (INDEP), separation of duties (Dual), and nature of ownership (SOE). Appendix A reports the specific variable definitions.

3.3. Models

To empirically evaluate the influence of AI integration on SCR, we delineate the regression model specified in Equation (2). Moreover, to investigate the moderating effect of CEOs’ sports experience, we present the moderation model articulated in Equation (3), which includes a CEOs’ sports experience variable (SPORTS) and the interactions between this variable and the AI application measures (AI).

$$SC{R}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}Control{s}_{i,t}+{\beta }_3{\displaystyle \sum Year}+{\beta }_4{\displaystyle \sum Industry}+{\epsilon }_{i,t}$$

(2)

$$\begin{array}{ll}SC{R}_{i,t}& ={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_2{\mathrm{SPORTS}}_{i,t}+{\beta }_{3}A{I}_{i,t}\times {\mathrm{SPORTS}}_{i,t}+{\beta }_{4}Control{s}_{i,t}\\ & +{\beta }_5{\displaystyle \sum Year}+{\beta }_6{\displaystyle \sum Industry}+{\epsilon }_{i,t}\end{array}$$

(3)

where i and t represent the firm and year, while ${\epsilon }_{i,t}$ refers to the random error term. Specifically, AI represents the AI application of firm i in year t. SCR stands for the level of SCR of firm i in year t. AI × SPORTS is the centralized interaction term representing the moderating effect of CEOs’ sports experience on the association between AI and SCR. Controls represent control variables covering corporate financial and governance characteristics. In addition, industry (Industry) and year (Year) fixed effects are controlled. The standard errors are clustered at the firm level.

4. Empirical Results

4.1. Descriptive Statistics

Table 2. Industry distribution of sample enterprises.

Industry Name	Observations	Proportion (%)
Agriculture, forestry, animal husbandry, and fishery	105	0.88
Mining industry	120	1.01
Manufacturing industry	9772	82.17
Industry of electric power, heat, gas, and water production and supply	100	0.84
Construction industry	303	2.55
Wholesale and retail industry	86	0.72
Transport, storage, and postal service industry	36	0.30
Accommodation and catering industry	1	0.01
Industry of information transmission, software, and information technology services	931	7.83
Real estate industry	14	0.12
Leasing and commercial service industry	46	0.39
Scientific research and technical service industry	188	1.58
Water conservancy, environment, and public facility management industry	157	1.32
Industry of resident service, repair, and other services	1	0.01
Education	2	0.02
Health and social work	10	0.08
Industry of culture, sports, and entertainment	19	0.16
Diversified industries	1	0.01

reports the industry distribution of sample enterprises. Among them, the manufacturing industry has the highest proportion, accounting for 82.17%.

Table 3. Descriptive statistics.

Variables	N	Mean	SD	Min	Max	VIF
SCR	11,892	0.7638	0.370	0.02	0.99	-
AI	11,892	1.1410	1.352	0.00	4.78	1.26
SPORTS	11,892	0.0410	0.198	0.00	1.00	1.02
SIZE	11,892	22.0311	1.052	20.19	25.13	1.59
LEV	11,892	0.3849	0.182	0.04	0.81	1.83
AGE	11,892	2.9147	0.298	1.61	3.53	1.07
ROA	11,892	0.0398	0.061	−0.29	0.19	1.50
GROWTH	11,892	0.1622	0.297	−0.45	1.52	1.18
RD	11,892	0.0584	0.050	0.00	0.33	1.43
BOARD	11,892	2.2113	0.162	1.79	2.77	1.84
TOP1	11,892	0.3433	0.139	0.08	0.74	1.07
INDEP	11,892	0.3762	0.049	0.33	0.50	1.69
Dual	11,892	0.3475	0.476	0.00	1.00	1.13
SOE	11,892	0.2133	0.410	0.00	1.00	1.29

reports the descriptive statistics of the main variables. SCR has a standard deviation of 0.370, indicating that there is a certain degree of variation in SCR among sample enterprises. The maximum value of AI is 4.78, and the mean is 1.4140, indicating that the overall level of AI technological application among the sample enterprises is relatively low. Further analysis of the variance inflation factors (VIFs) for the main variables is conducted as well. With a mean VIF of 1.38 and no values exceeding 10, it indicates a low degree of multicollinearity.

4.2. Main Regression Results

Firstly, the impact of AI application on SCR is examined based on Equation (2). Column (1) of

Table 4. Main regression results.

	(1)	(2)
	SCR	SCR
AI	0.014 **	0.012 *
	(2.00)	(1.75)
SPORTS		−0.063
		(−1.50)
AI × SPORTS		0.039 **
		(2.43)
SIZE	−0.001	−0.001
	(−0.04)	(−0.07)
LEV	−0.006	−0.007
	(−0.12)	(−0.14)
AGE	0.077	0.083
	(0.77)	(0.83)
ROA	0.044	0.039
	(0.50)	(0.45)
GROWTH	−0.044 ***	−0.043 ***
	(−2.97)	(−2.91)
RD	−0.375 *	−0.378 *
	(−1.75)	(−1.77)
BOARD	0.020	0.019
	(0.33)	(0.33)
TOP1	−0.023	−0.020
	(−0.29)	(−0.26)
INDEP	−0.118	−0.122
	(−0.68)	(−0.70)
Dual	0.027 **	0.026 **
	(2.10)	(2.02)
SOE	0.013	0.015
	(0.47)	(0.52)
Constant	0.398	0.400
	(0.91)	(0.92)
Year FEs	Yes	Yes
Industry FEs	Yes	Yes
N	11,892	11,892
adj. R2	0.051	0.052

Note: (1) Standard errors are clustered at the firm level. (2) ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively (same as below).

demonstrates the significantly positive coefficient of AI, indicating that AI application promotes SCR, thus validating hypothesis H1. In terms of the economic consequence, a 1% increase in AI technology applications leads to a 5.12% improvement in SCR.

Secondly, the moderating role of CEOs’ sports experience is examined based on Equation (3). Column (2) of Table 4 demonstrates the significantly positive coefficient of AI × SPORTS, indicating that the CEOs’ sports experience strengthens the association between AI application and SCR. Therefore, H2 is supported.

4.3. Endogeneity Test

4.3.1. Propensity Score Matching Approach

The adoption of AI by companies is not random but is determined by corporate characteristics such as human capital, management practices, and technological levels, as well as changes in the external environment. Therefore, there may be sample self-selection bias in our research. Our research employs the propensity score matching (PSM) approach. Specifically, we divide the sample into an experimental group and a control group according to the presence of AI-related keywords in the annual reports. Then, we use the control variables introduced in Equation (2) as covariates and match the sample by sequentially employing the 1:1 nearest neighbor matching method, radius matching method, and kernel matching method. Finally, we re-estimate the benchmark model utilizing the matched samples.

Table 5. PSM method.

	(1)	(2)	(3)
	1:1 Matching	Radius Matching	Kernel Matching
AI	0.013 *	0.015 **	0.015 **
	(1.68)	(2.14)	(2.21)
Constant	0.450	0.284	0.294
	(0.85)	(0.67)	(0.69)
Controls	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes
Industry FEs	Yes	Yes	Yes
N	9432	11,959	11,947
adj. R2	0.045	0.052	0.051

Note: ** and * indicate significance at the 5% and 10% levels, respectively.

shows the results of the PSM approach. As reported in columns (1) to (3), the coefficients of AI are all significantly positive, which further supports our hypothesis.

4.3.2. Instrumental Variable Analysis

This paper further employs instrumental variable analysis to address potential endogeneity issues. Our research utilizes the installed density of industrial robots each year across provinces in China (Robot) to construct the instrumental variable for AI applications. Using the data on the installation of industrial robots in each industry of the United States released by the International Federation of Robotics (IFR) and the employment data in each industry of each province in China, we employ the Bartik method to construct the indicator of the installed density of industrial robots in each province. The specific measure is as follows:

$$Robo{t}_{j,t}=({\displaystyle \sum _{m=1}^M{E}_{j,m,t}\times }Robot\_Stoc{k}_{m,t})/Em{p}_{j,t}$$

(4)

where $E_{j,m,t}$ represents the share of province j’s employment in industry m in year t in China’s total employment in that industry in that year, $Robot\_Stoc{k}_{m,t}$ refers to the quantity of robots installed in industry m in year t provided by the IFR, $Em{p}_{j,t}$ refers to the total national employment in province j in year t in China, and M refers to the industries that need to be summed.

The 2SLS regression results are presented in

Table 6. Instrumental variable analysis.

	(1)	(2)
	2SLS First Stage	2SLS Second Stage
	AI	SCR
Robot	0.026 ***
	(6.33)
AI		0.124 ***
		(3.54)
Constant	−1.263 *	0.613 ***
	(−1.88)	(3.41)
Kleibergen–Paap rk LM	38.24 ***
Kleibergen–Paap rk Wald F	40.05 ***
Controls	Yes	Yes
Year FEs	Yes	Yes
Industry FEs	Yes	Yes
N	11,932	11,932

Note: *** and * indicate significance at the 1% and 10% levels, respectively.

. Column (1) demonstrates that the coefficient of Robot is significantly positive, suggesting that the density of industrial robots installed in the enterprise’s location significantly promotes the application of corporate AI technology. In the second stage, column (2) reveals that the coefficient of AI is still significantly positive after considering potential endogeneity problems. The instrumental variable analysis shows the causal link between AI application and SCR, which provides robust evidence for our hypothesis.

4.3.3. Quasi-Natural Experiment Analysis

Our research further employs the policy of China’s national AI innovative development pilot zones as a quasi-natural experiment. Since May 2019, China’s Ministry of Industry and Information Technology has successively approved 11 national AI innovative development pilot zones, and this policy has provided favorable conditions for companies in the pilot zones to apply AI technology. The DID model is shown in Equation (5).

$$SC{R}_{i,t}={\beta }_0+{\beta }_{1}TREA{T}_i\times POS{T}_t+{\beta }_{2}Control{s}_{i,t}+{\beta }_3{\displaystyle \sum Year}+{\beta }_4{\displaystyle \sum Industry}+{\epsilon }_{i,t}$$

(5)

TREAT is whether the location of firm i lies in the pilot zone, taking 1 if it belongs to the national AI innovative development pilot zone. The sample from three years before and after the policy issuance is selected as the test data. Samples before the policy issuance are defined as POST = 0. Control variables set in Equation (5) are used as the matching criteria, and the 1:1 nearest neighbor matching method is applied. We perform regression on the matched sample, and the results are reported in

Table 7. Quasi-natural experimental analysis.

	(1)
	SCR
TREAT × POST	0.072 *
	(1.83)
Constant	3.873 *
	(1.76)
Controls	Yes
Year FEs	Yes
Industry FEs	Yes
N	1668
adj. R2	0.080

Note: * indicates significance at the 10% level.

. The coefficient of TREAT × POST is significantly positive, suggesting that the positive role of AI in shaping SCR remains robust after addressing the endogeneity issue.

4.4. Robustness Test

4.4.1. Alternative Measures of AI Application

The MD&A section of the annual report has been the subject of extensive research by scholars [90,91]. It describes the company’s operational performance, financial condition, investment in R&D, and development strategy during the reporting period, all of which directly or indirectly involve discussions on the company’s use of AI technology. Our research further measures corporate AI applications by taking the natural logarithm of the number of AI-related keywords in the MD&A section of the annual report (LnMDA). As shown in column (1) of

Table 8. Robustness tests.

	Alternative Measures of AI Application	Lagging Explanatory Variable	Deleting Samples
	(1)	(2)	(3)
LnMDA	0.013 *
	(1.82)
AI_1		0.014 *
		(1.65)
AI			0.013 *
			(1.87)
Constant	0.270	1.077	0.215
	(0.63)	(1.42)	(0.49)
Controls	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes
Industry FEs	Yes	Yes	Yes
N	11,960	6963	11,025
adj. R2	0.051	0.031	0.050

Note: * indicates significance at the 10% level.

, the coefficient of LnMDA remains significantly positive.

4.4.2. Lagging Explanatory Variable

As the impact of AI technology applications on SCR may take more than one year to materialize, our research treats AI technology applications with a one-period lag (AI_1). As reported in column (2) of Table 8, the estimated coefficient of AI_1 is significantly positive.

4.4.3. Deleting Samples

Companies belonging to the information technology industry may have relatively stronger capabilities in digital technology R&D, resulting in higher levels of AI technology compared to other industries. This paper re-estimates the regression after excluding the “Information Transmission, Software, and Information Technology Services” sector. As reported in column (3) of Table 8, the coefficient of AI remains significantly positive.

4.5. Mechanism Examination

The model for mechanism examination is constructed as follows:

$$MEDI{A}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}Control{s}_{i,t}+{\beta }_3{\displaystyle \sum Year}+{\beta }_4{\displaystyle \sum Industry}+{\epsilon }_{i,t}$$

(6)

where MEDIA is the indicator of mechanisms including operational efficiency optimization (EFFICIENCY), information spillover (INFO), and knowledge spillover (PC). The rest of the symbols are consistent with Equation (2).

4.5.1. Operational Efficiency Optimization Effect

A shorter inventory turnover period indicates faster product circulation and higher operational efficiency of the supply chain. This study measures operational efficiency (EFFICIENCY) by dividing the company’s inventory turnover rate by 100.

Column (1) of

Table 9. Results of the mechanism examination.

	Operational Efficiency Optimization Effect	Information Spillover Effect	Knowledge Spillover Effect
	(1)	(2)	(3)
	EFFICIENCY	INFO	PC
AI	0.174 **	−0.009 ***	0.063 ***
	(2.20)	(−3.52)	(4.85)
Constant	13.718 **	−1.596 ***	−10.065 ***
	(2.30)	(−9.62)	(−13.07)
Controls	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes
Industry FEs	Yes	Yes	Yes
N	22,681	22,014	22,436
adj. R2	0.040	0.171	0.506

Note: *** and ** indicate significance at the 1% and 5% levels, respectively.

demonstrates the examination results for the operational efficiency optimization effect. The significantly positive coefficient of AI suggests that the application of AI can enhance the company’s inventory turnover rate and optimize supply chain efficiency. Therefore, H3a has been confirmed.

4.5.2. Information Spillover Effect

This study uses the KV index to assess the level of information spillover within companies (INFO). The KV value accurately reflects the real effectiveness of information disclosure by listed companies, incorporating both mandatory and voluntary disclosures. A higher KV index indicates a lower quality of corporate information disclosure, suggesting weaker external information spillover.

Column (2) of Table 9 provides the results of the mechanism tests for the information spillover effect. The significantly negative coefficient of AI reveals that AI applications can improve the quality of corporate information disclosure and alleviate information asymmetry among supply chain enterprises, thus confirming H3b.

4.5.3. Knowledge Spillover Effect

The number of citations of a company’s patents indicates the movement and spillover of knowledge. When a company cites patents from other firms, it signifies that the company has gained and absorbed knowledge and innovative technologies from those firms. The citation frequency of a company’s patents (PC) is used to measure the level of knowledge spillover from core enterprises’ AI applications.

Column (3) of Table 9 shows the results of the mechanism tests for the knowledge spillover effect. The significantly positive coefficient of AI reveals that the innovative experience and knowledge related to AI applications by core enterprises can spread throughout the supply chain. Therefore, H3c has been verified.

4.6. Heterogeneity Analysis

4.6.1. Micro Level: Corporate Human Capital

Employees and the executive team are important parts of a company’s human capital, affecting the company’s adaptability and integration ability in AI technology applications. High-skilled labor, referring to employees with high qualifications and skills, can learn and master complex knowledge and technologies more rapidly. Additionally, AI adoption not only increases the corporate demand for highly skilled labor but also encourage companies to train employees in AI-related skills. With the integration of AI technology and employee skills, companies can more effectively speed up the use of AI in supply chain management.

The heterogeneity of the executive team’s human capital also plays a significant role in decision-making regarding AI applications. The complexity and continuously evolving characteristics of AI technology necessitate that the executive team conduct a thorough analysis from multiple perspectives. Members with higher education can offer the latest theoretical perspectives, while those with practical experience can ensure the feasibility of the solutions [92]. Thus, the heterogeneity of the executive team’s human capital allows companies to better recognize opportunities and challenges in AI applications.

Following previous studies [93,94], our research measures a company’s high-skilled labor force by the proportion of employees with a graduate degree or higher (Labor). Regarding the heterogeneity of the executive team’s human capital, the educational background categories include below secondary school, secondary school, junior college, undergraduate, master, other, and MBA/EMBA, with values ranging from 1 to 7. Then, we employ the Herfindahl–Hirschman Index to assess the heterogeneity of the executive team’s human capital (H_heter). The calculation formula is as follows:

$$H=1-{\displaystyle \sum _{i=1}^n{P}_{i,t}^2}$$

(7)

The full sample is split into two groups based on the median of Labor and H_heter. Columns (2) and (4) of

Table 10. Heterogeneity analysis: Micro level.

	(1)	(2)	(3)	(4)
	Low Labor	High Labor	Low H_heter	High H_heter
AI	0.017 *	0.019 **	0.002	0.030 ***
	(1.67)	(1.98)	(0.19)	(2.64)
Constant	0.111	0.693	0.383	0.463
	(0.17)	(1.04)	(0.54)	(0.66)
Controls	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes
Industry FEs	Yes	Yes	Yes	Yes
N	5934	5938	5293	5267
adj. R2	0.065	0.039	0.057	0.060

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

indicate that the coefficients of AI are more pronounced in the high-skilled labor force group and the high heterogeneity of the executive team’s human capital group. The above results confirm the vital role of high-skilled employees and the executive team diversity in building an AI-based intelligent supply chain.

4.6.2. Meso Level: High-Tech Industry

AI technology has particular applicability in different use cases, and its influence on SCR may differ across industries. In high-tech industries, technological innovation serves as their core competitive strength. Compared to low-tech industries, high-tech industries tend to have more skilled labor, advanced technologies, and faster R&D speeds. Due to their industry complexity and dependence on advanced technologies, high-tech industries usually have greater data accessibility and processing requirements, allowing them to fully leverage the smart advantages of AI technology.

To differentiate the heterogeneous impacts of industries, this paper refers to the “Classification of High-Tech Industries” issued by the National Bureau of Statistics of China, dividing industries into high-tech and low-tech sectors. According to the results in columns (1) and (2) of

Table 11. Heterogeneity analysis: Meso level and macro level.

	(1)	(2)	(3)	(4)
	Low-Tech Industry	High-Tech Industry	Low Digital Infrastructure	Strong Digital Infrastructure
AI	0.008	0.017 **	0.011	0.019 **
	(0.63)	(1.99)	(0.97)	(2.00)
Constant	0.437	0.481	−0.255	1.138 *
	(0.58)	(0.87)	(−0.34)	(1.96)
Controls	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes
Industry FEs	Yes	Yes	Yes	Yes
N	3948	8012	4403	7557
adj. R2	0.052	0.050	0.055	0.051

Note: ** and * indicate significance at the 5% and 10% levels, respectively.

, the coefficient of AI is more significant in the high-tech industries group, indicating that the high-tech characteristics of an industry are more favorable for the deployment and application of AI technology.

4.6.3. Macro Level: Digital Infrastructure

The deployment of AI technology requires the digital infrastructure of the ecosystem in which the entity operates as a safeguard. The stronger the digital technological infrastructure in the region where the enterprise is located, the more pronounced the effects of AI technology. On the one hand, a well-developed digital infrastructure ensures the rapid flow of information and provides companies with necessary data analysis and processing capabilities. On the other hand, regions with higher levels of digital infrastructure tend to attract more highly skilled talent and capital, which further drive AI technological innovation and commercialization.

We measure the level of digital infrastructure (DIG) through the number of internet access ports to the population in the province where the enterprise is located. Then, the sample is classified into firms with strong and weak digital infrastructure based on the median of DIG. As reported in column (4) of Table 11, the coefficient of AI is only significantly positive in the strong digital infrastructure group, confirming the complementary effect of digital infrastructure and AI application in improving SCR.

5. Further Analysis

Rogers proposed the innovation diffusion effect, defined as the process by which innovation spreads among members of a social network through specific channels [95]. The supply chain is a vital component of the social network, where enterprises are closely connected and exchange resources such as information with upstream and downstream companies, thus creating the transmission and linkage effects of the supply chain [96]. As digital technology advances, the spillover effects of information and knowledge during AI application among supply chain companies have deepened. The innovation experience and knowledge of companies using AI technology become the starting point of the supply chain diffusion effect [97], resulting in the spread of AI across upstream and downstream enterprises in the supply chain.

Our previous analysis reveals that the application of AI technology enhances SCR through information and knowledge spillover mechanisms between companies in the supply chain. This raises the following question: Will the spillover effect of AI in the supply chain lead to a diffusion effect on the AI capabilities of upstream and downstream companies? To address this question, our research further develops a 1:1 matched database of supply chain companies, collecting 497 sets of enterprise–supplier annual samples and 565 sets of enterprise-customer annual samples. Equation (8) is used to test the diffusion effect of AI technology applications in the supply chain.

$$Y_{i,t}={\beta }_0+{\beta }_{1}AI-X_{i,t}+{\beta }_{i}Controls+{\displaystyle \sum Year}+{\displaystyle \sum Industry}+{\epsilon }_{i,t}$$

(8)

where $Y_{i,t}$ refers to the application of AI technology in upstream supplier companies (S_AI-X) and downstream customer companies in the supply chain (C_AI-X). Controls represent the corresponding control variables for upstream and downstream companies, including core enterprise Controls, upstream supplier S_Controls, and downstream customer C_Controls.

Table 12. The supply chain diffusion effect of AI.

	(1)	(2)
	S_AI-X	C_AI-X
AI-X	0.365 ***	0.333 ***
	(6.06)	(4.79)
Constant	−0.758	−4.420 **
	(−0.49)	(−2.44)
Controls	Yes	Yes
Year FEs	Yes	Yes
Industry FEs	Yes	Yes
N	497	565
adj. R2	0.438	0.417

Note: *** and ** indicate significance at the 1% and 5% levels, respectively.

shows the empirical results of examining the diffusion effect of AI in the supply chain. The coefficients of S_AI-X and C_AI-X are both significantly positive, suggesting that the application of AI technology by core enterprises boosts the AI capabilities of upstream suppliers and downstream customers. The results above indicate that corporate AI adoption has a positive diffusion effect across supply chain networks, triggering an overall improvement in the intelligence level of the supply chain system.

6. Discussion

In an environment where the characteristics of VUCA are increasingly visible, supply chains urgently need to accelerate their own changes to cope with all kinds of complex and severe risks and challenges. How to use AI to enhance SCR has rapidly attracted extensive attention from the academic community. Scholars have studied the use of AI techniques for system development and algorithm modelling, and provided application strategies in demand forecasting, risk management, transportation, supplier selection, and inventory management. There are also studies that systematically sort out how AI technology impacts different aspects of SCR through a literature review and build a theoretical framework of AI-SCR [11,98]. Despite the comprehensive exploration of AI’s positive impact on shaping SCR in the existing literature, much of the research predominantly relies on theoretical analyses [11,98] and survey-based approaches [15,99]. In contrast, this study seeks to offer more timely and empirical evidence through large-scale testing. Specifically, an SCR index for enterprises is developed using the entropy weighting method, encompassing both the resistance and recovery capacity dimensions. Additionally, machine learning and text analysis techniques are applied to construct an AI index, which highlights the beneficial influence of AI applications on SCR. Further, this study elaborates on the operational mechanisms involved, such as optimization of operational efficiency, information spillover, and knowledge spillover. This research not only contributes to advancing the theoretical framework regarding AI’s role in shaping SCR but also provides a practical and quantitative model for future theoretical inquiries.

In addition, this paper analyzes how CEO characteristics influence the process of AI technology application. For traditional enterprises, the application of AI technology to supply chain management is a strategic change that requires top managers to carry out top-level design and resource allocation optimization. This paper overcomes the shortcomings of previous studies of the micro-level economic consequences of AI technology application, which ignore managerial characteristics [26], and explores the moderating role played by CEOs’ sporting experiences. Practice has demonstrated that the sports experiences of many well-known entrepreneurs have a profound impact on the level of corporate AI technology adoption [100]. Examining the moderating role of CEOs’ sports experience can make up for the shortcomings of the existing upper echelons theory framework, which pays insufficient attention to the economic consequences of CEOs’ off-the-job experience.

Finally, most of the existing studies on the economic effects of AI applications focus on the impact of AI on the firms themselves, ignoring the possible external effects of AI applications [101,102]. Studies have been conducted to identify groups of enterprises with vertical networks, and to analyze the spillover phenomenon of digital transformation by intercepting the vertical correlation between enterprises [103,104]. Artificial intelligence technology, as a kind of digital technology, also has high synergy and strong externality characteristics. This paper confirms the diffusion effect of enterprise AI technology applications upstream and downstream of the supply chain, which enriches the research on the factors influencing the level of AI technology.

7. Conclusions and Implications

7.1. Conclusions

This paper discusses the role of AI in shaping SCR through large-sample empirical examination. Based on dynamic capability and information processing theory, our study reveals that AI application significantly improves SCR, and this conclusion remains valid after addressing endogeneity issues. Moreover, the CEOs’ sports experience strengthens the positive impact of AI on SCR.

In terms of the specific channel, mechanism examination shows that AI can drive companies to enhance SCR by improving operational efficiency and promoting information and knowledge spillover in the supply chain. Heterogeneity analysis reveals that such a positive impact is more pronounced in firms with high-skilled labor forces, firms with high heterogeneity of the executive team’s human capital, high-tech industries, and areas with strong digital infrastructure.

In addition, our research successfully identifies the diffusion effect of AI applications in the supply chain network. AI application of core enterprises not only enhances SCR but also boosts the AI capabilities of upstream suppliers and downstream customers.

7.2. Managerial Implications

Our research has several managerial implications for regulators and enterprises. For regulators, this paper reveals that AI application improves SCR. The better the digital technology facilities in the area where the enterprise is located, the stronger the positive effect that AI technology may exert. The above findings indicate that the government should first guide companies to solidify their strategic direction towards intelligence. The government can implement incentive policies at the firm level, providing financial subsidies, tax reductions, and other support for AI innovation and application, guiding companies to seize the opportunities presented by the new round of technological revolution and industrial change. Secondly, the government should accelerate the creation of a “hard environment” to provide fertile ground for AI adoption. Specifically, the government can improve digital infrastructure construction, including building high-speed, stable internet networks and secure, reliable data centers, establishing cloud computing platforms to provide computation, storage, and application services for companies, and promoting the application of IoT technology in logistics and warehousing.

For companies, apart from the vital role of AI in shaping SCR, our results also indicate that CEOs’ sports experience, the heterogeneity of senior executive teams, and high human capital can strengthen the positive association between AI and SCR. AI application has a diffusion effect on the upstream and downstream enterprises of the supply chain. In a VUCA environment, managers must deeply understand the close relationship between digital intelligence transformation and SCR development, formulate reasonable AI application strategies, and leverage AI technology to enhance crisis response capabilities in supply chain management. CEOs’ sports experience can be an important consideration for the board of directors when selecting a CEO. In addition to the usual academic background and work experience, the board of directors can try to focus on the CEOs’ sports experience to assess personal traits such as openness. In addition, companies should build a diverse executive team. Here, companies form their executive team based on the principles of differentiation and diversity, introducing senior managers with varied backgrounds to inject vitality into corporate AI applications. Companies can also adjust their workforce skill structure to better leverage the positive effects of AI. Companies should actively recruit talent with expertise in AI and regularly provide employees with AI-related skills training. Finally, companies should capitalize on the supply chain diffusion effects of AI. Companies should actively build collaborative relationships for digital intelligence across the upstream and downstream supply chain, integrating internal and external resources, data, and services to achieve full-chain upgrades and ecological development of the intelligent supply chain.

7.3. Limitations and Prospects

This study has several limitations that warrant further investigation.

First, this study did not fully capture the technological characteristics of AI when measuring this variable, as it lacked a more detailed classification of AI technologies. AI can be categorized into logical AI and machine learning, with the former emphasizing propositional reasoning and the latter utilizing data-driven learning for prediction and decision-making. Future research could attempt to classify AI technologies more comprehensively and examine how different AI applications influence SCR in varying ways.

Second, in exploring the mechanisms through which AI impacts SCR, our study only considers the CEO’s sports experience as a moderating variable at the individual level, without delving into other executive attributes or traits. Consequently, future research may explore additional factors at the executive level, such as the background in information technology or academic expertise, and investigate their influence on the microeconomic outcomes of AI applications.

Finally, while AI is a key technological factor influencing SCR, other organizational and environmental factors also play crucial roles. Future research should explore these additional dimensions to provide a more comprehensive understanding of SCR. By broadening the focus beyond technology to include organizational and environmental aspects, a more holistic perspective can be developed, offering practical guidance for businesses navigating the increasingly complex and volatile global supply chain landscape.