Beyond cargo, supply chain also transfers AI

Machine learning has enhanced quality management and cost efficiency across the global supply chain, but how did it spread?

Featured faculty: Cen Ling and Wu Jing
Written by Pan Jingyi

In recent years, artificial intelligence (AI) has become a key tool for companies, helping with everything from demand forecasting and procurement to streamlining and optimising processes. For global supply chains that have been under pressure lately due to geopolitical uncertainties, trade conflicts, sanctions, and environmental concerns, AI could be a game changer.

According to a recent EY report, around 40 per cent of supply chain organisations are investing in generative AI for managing knowledge and information. Technologies tend to spread along economic networks, either horizontally among competitors or vertically along the supply chain. However, the diffusion of emerging technologies such as AI remains underexplored.

“We uncover a clear pattern of AI diffusion along supply chains, where AI adoption among downstream sectors leads to subsequent adoption among their upstream suppliers,” says Cen Ling, Associate Professor at the Department of Finance of the Chinese University of Hong Kong (CUHK) Business School.

Downstream industries are those closer to the final stage of production and delivery to end customers, including service and manufacturing sectors, while upstream industries refer to sectors providing raw materials or basic inputs, like mining and agriculture, for downstream industries.

Professor Cen highlights Foxconn, a key supplier for Apple and Nvidia, as a notable example. The rapid AI applications of its major customers have turned Foxconn from a labour-intensive company to one that produces AI-driven electric vehicles and AI servers to house Nvidia chips.

Along with Wu Jing, Associate Professor at the Department of Decisions, Operations and Technology of the School, as well as Han Yanru of Stevens Institute of Technology (A Graduated PhD student of CUHK Business School) and Qiu Jiaping of Shanghai University of Finance and Economics, Professor Cen conducted a study titled Artificial intelligence along the supply chain to delve deeper into critical questions like whether supply chain can serve as a diffusion channel for AI technologies and what economic mechanisms propel this diffusion.

Lead-lag patterns

To measure how American public companies are adopting AI, the team tracked the hiring of AI-skilled employees using data from Revelio Labs, an analytics firm that gathers workforce information from employment platforms like LinkedIn and Indeed, from 2009 to 2019.

The results show a significant increase in AI employees over the past decade, with AI use growing in almost all industries. However, considerable variation exists among various sectors, with downstream industries adopting AI technologies much faster than upstream industries. This is because, Professor Cen explains, downstream companies in the supply-chain data are typically large and reputable industry leaders with direct access to big data of customer profiles.

“We uncover a clear pattern of AI diffusion along supply chains, where AI adoption among downstream sectors leads to subsequent adoption among their upstream suppliers.”

Professor Cen Ling

Further analysis unveils that the increase in AI adoption by main customers precedes and potentially causes an increase in AI adoption among their suppliers in the following year. The researchers call this a “lead-lag” within firm-pair relationships and found that this pattern is not driven by market-wide or industry-specific trends.

In a controlled experiment where the team replaced the suppliers with those who have no prior connections to customers, the results showed that the AI adoption at customer firms does not impact that of suppliers. This suggests the diffusion is indeed driven by firm-to-firm interactions.

Learning and catering mechanism

Suppliers may enhance their AI adoption in response to their primary customers through two key channels: learning and catering. Under the learning scenario, primary customers, often large firms or industry leaders, typically adopt AI technologies before their suppliers, which then absorb and apply these technologies in their operations through routine supply chain interactions.

“Under this mechanism, close strategic relationships with customers equipped with AI technologies reduce suppliers’ costs of learning and adopting AI, which may improve suppliers’ own operational activities and performance,” Professor Cen says.

In the catering scenario, suppliers respond to the needs of major customers who have embraced AI technologies to sustain crucial partnerships. “The effectiveness of the catering channel depends on the relative bargaining power of customers against their suppliers,” he adds. “The learning channel is influenced by the relative size of supply-chain partners, which affects the applicability of the knowledge transferred.”

Based on the collected data, Professor Cen and his collaborators found that the learning mechanism is the primary driver. Suppliers mostly adopt AI to enhance their own capabilities rather than just to cater to customers’ demands.

What makes AI spread faster?

To examine the factors that can affect the suppliers’ learning mechanism, the team examined the employee mobility and geographic distance between customers and suppliers and found that when suppliers hire managers who previously worked for their main customers, the AI adoption is notably higher. Managers normally have a broader view of their company compared to rank-and-file employees, who may lack knowledge of AI advancements unless they work in AI-related roles.

“Our results validate that labour mobility from customers to suppliers, particularly employees with AI-related visions or skills, promotes the AI learning along the supply chain,” Professor Cen adds.

The diffusion of AI technology is also stronger when the geographical distances are shorter. Moreover, when customers relocate farther from suppliers, the suppliers’ AI adoption becomes less influenced by their customers.

RELATED ARTICLE

AI vs. humans: Who wins in handling service rejections?

“A shorter distance between supplier and customer facilitates more frequent interactions and leads to faster transfer of knowledge and information,” Professor Cen says, highlighting the crucial role of geographic proximity in stimulating learning.

Positive economic outcomes

Finally, the team examined whether AI diffusion from customers to suppliers actually improves business outcomes. The result confirms that suppliers that have learned AI from their customers are more likely to enhance product quality and manage costs more effectively. More specifically, every standard increase in AI hiring is linked to a 0.74 per cent higher chance of boosting product quality.

As AI continues to spread across industries and borders, understanding how it diffuses along supply chains offers valuable lessons for both business leaders and policymakers.

Professor Cen suggests corporate managers can mitigate risks and maintain competitiveness by tapping into AI knowledge within their supply chain partners. For instance, they can identify the optimal point to acquire such knowledge from trade partners or hire AI experts from these partners.

Compared to traditional technologies, AI requires a significant initial setup cost. However, once established, the ongoing operational costs of using it are comparatively low. Professor Cen argues that government subsidies for downstream customer firms could help kickstart the adoption, “then the positive externalities will diffuse along economic networks to achieve a socially optimal level of AI adoption.”