Assessing the Hype, Hazards, and Hypothetical Value of AI in Supply Chain Management

Ever since OpenAI launched ChatGPT in November 2022, much of the world has been varyingly captivated, unnerved, and dumbfounded by the technological breakthroughs and seemingly infinite potential of artificial intelligence. In the 18 months since then, companies have rushed to incorporate AI platforms and models into their products, internal processes, and overarching corporate strategies. As a result, what was once a steadily growing trend has rapidly swelled into an investment stampede.

According to The Wall Street Journal, businesses are on track to spend nearly $39 billion on generative AI this year, exactly double the spend from 2023. Projections published last year by Goldman Sachs speak to an even larger spike in AI bets. The financial firm’s economists believe that organizations’ investments in AI could approach $100 billion in the U.S. by 2025 (the research note did not distinguish between generative AI—the form embodied by ChatGPT—and artificial intelligence technologies more broadly). 

While many of the world’s largest corporations continue to be ensorcelled in the honeymoon phase with this revolutionary new technology, there is already some credible data establishing a track record of use-cases for AI in various industries. A December 2022 McKinsey & Company study surveyed nearly 1,500 companies all over the world to gauge how artificial intelligence was being implemented among corporations, and the extent to which it was adding value to those organizations. The consulting firm found that businesses were generally experiencing modest revenue increases and cost reductions across a variety of functions, including manufacturing, sales and marketing, and product development. One of the activities that appeared to be benefiting more from AI integration than any other, however, was supply chain management. Fifty-two percent of respondents reported that the implementation of AI was leading to cost decreases in their supply chain management processes. Nearly 60%, meanwhile, said that it was increasing revenue. 

While many of the world’s largest corporations continue to be ensorcelled in the honeymoon phase with this revolutionary new technology, there is already some credible data establishing a track record of use-cases for AI in various industries.

In some ways, the idea that AI and its latest, most riveting incarnation, generative AI, would prove beneficial to supply chain management should come as no surprise. The operational discipline relies on effectively parsing immense amounts of data, deploying complex analytics to guide strategies and decision-making, and detecting and analyzing hypothetical risks—all functions that can be meaningfully bolstered with the unique capabilities offered by AI. 

But for a field as practical as supply chain management—and, to an arguably equal extent, the related supply chain risk management (SCRM)—it’s not sufficient to be dazzled by the possibilities of incorporating these cutting-edge platforms into daily operations. What’s far more worthwhile is digging deeper and asking what the legitimate use cases for artificial intelligence actually are, and whether or not they’re offset by the hazards of bringing on a nascent, expensive, and potentially disruptive technology.

The Use Cases for AI in Supply Chain Management 

Artificial intelligence is no longer a hypothetical in the world of supply chain management and SCRM. It’s a powerful, multifarious tool accessible to supply chain professionals right now. Further, it’s already being deployed in a diverse variety of ways that are adding tangible value to U.S. manufacturers and other organizations. 

• Predictive analytics for manufacturing equipment 
• Demand forecasting
• Supply chain mapping
• Risk assessments and risk matrices 
• Alternative suppliers

Predictive Analytics for Manufacturing Equipment 

In contrast with the staggering depth of information that feeds it, predictive analytics as a concept is exceedingly cogent and straightforward. Harvard Business School defines it as “the use of data to predict future trends and events.” Such predictions, the definition continues, “could be for the near future—for instance, predicting the malfunction of a piece of machinery later that day—or the more distant future, such as predicting your company’s cash flows for the upcoming year.”

In the context of supply chain management, it’s the former scenario that’s of the most interest. This niche aspect of predictive analytics is sometimes referred to as “predictive maintenance.” In predictive maintenance, AI and machine learning collect vast amounts of data drawn from sensors attached to and monitoring specific equipment or infrastructure. The data, which covers behavioral metrics like temperature, vibration, pressure, humidity, and speed, allows the AI model to learn the regular operating conditions of the machinery and equipment. Once all this data has been comprehensively analyzed, AI is able to immediately recognize deviations from those optimal conditions—thus potentially identifying precursors to a more substantial malfunction, breakdown, or equipment failure. 

Being able to detect an equipment failure before it even materializes—thereby not only mitigating a disruption, but potentially circumventing it entirely—can benefit a range of stakeholders along the electronics supply chain. Electronics contract manufacturers (ECMs) and other suppliers are able to avoid breakdowns, reduce equipment downtime, and maximize productivity in factories and other manufacturing facilities. Parties further downstream, including original equipment manufacturers (OEMs) and other businesses, meanwhile, benefit from fewer manufacturing slowdowns and delays, allowing them to hit production targets more consistently. And in the case of major equipment failures that can’t be seamlessly mitigated, the predictive maintenance offered by AI can at least alert various stakeholders of imminent disruptions, affording them the opportunity to initiate contingency planning and react with prudence and precision.  

Demand Forecasting 

Businesses have been practicing demand forecasting long before the term was coined and popularized over the past few decades. Manufacturers have been at the mercy of the market’s spontaneity and unforeseeable whims for time immemorial. (During the “Tulip Mania” experienced in the Netherlands during the 17th century, for example, demand surged so aggressively that a single tulip bulb could go for as much as $750,000 in today’s dollars.) Historically, their response has been to attempt to extract effective predictions—or “forecasts”—from a slew of quantitative data, including sales trends, seasonal adjustments, and historical patterns. The results have varied, to say the least, in large part because humans are only able to analyze so much information. Further complicated matters is the fact that the conjectures drawn from those incomplete data sets are often imperfect or otherwise colored by internal biases, inaccurate expectations, or even unconscious emotional investments. 

The raft of new AI models now available to organizations don’t bring the same human fallibility to the work of demand forecasting. More importantly, these AI platforms are able to pore over far more information than their human counterparts ever could. Their assessments reach beyond historical sales data and cover more nuanced, indirect metrics, including the performance trends of market competitors, broader macroeconomic factors, and changes to consumer behavior that could presage significant shifts in demand. Certain AI models are even able to detect the early warning signs of panic buying, allowing businesses to respond strategically and opportunistically without interpreting the ephemeral uptick as a permanent shift in demand. 

By leveraging the capabilities offered by AI tools, manufacturers and other companies can meaningfully enhance their capacity to predict the market demand for their products. Because of this, they’re far less likely to be subjected to the devastating opportunity cost represented by supply shortages or have to scramble to implement ad-hoc strategies for rapidly offloading surpluses. 

By leveraging the capabilities offered by AI tools, manufacturers and other companies can meaningfully enhance their capacity to predict the market demand for their products.

Supply Chain Mapping 

One of the largest challenges strategic sourcing and procurement professionals face is understanding the full breadth and depth of their supply chains. A 2021 study, also carried out by McKinsey, illustrates just how difficult extensive, in-depth supply chain mapping is—and how quickly it drops off as organizations delve deeper into their supplier networks. While nearly half of the supply chain executives who responded to the survey reported having good visibility into their tier one suppliers, that figure tumbled to around one-fifth for tier two manufacturers. Any visibility beyond tier two, meanwhile, was all but nonexistent: a mere 2%t of respondents said they were aware of who their tier three suppliers were.

The capacity to conduct deeper, more comprehensive supply chain mapping confers a host of operational benefits. Companies that can see into their second- and third-tier suppliers are able to carry out more complete risk assessments of their supply chains; identify materials, countries of origins and other data points relevant to regulatory compliance; and accurately diagnose the source of delays and shortages in ways that are not available to firms with less visibility. The latter factor is particularly important, because being able to pinpoint the exact source of a supply chain disruption enables manufacturers to strategize and problem-solve much more effectively. When you know that a specific tier three manufacturer is the cause of a shortage in product from a tier one supplier, you can work with that direct supplier to develop targeted resolutions. 

But the central issue plaguing supply chain mapping is not skepticism surrounding its value. It’s the crushing logistical challenges inherent in actually constructing these data-dependent visualizations. This is where AI can demonstrate a substantive use case. These tools are able to scrape the web for the obscure, far-flung resources critical to illuminating transactions between manufacturers—and consequently connecting the businesses that comprise your supplier tiers. AI platforms can procure orders, receipts, customs declarations, freight bookings, and other documentation that create a chain of custody for your product, components, and raw materials. Using these records and documents, businesses can build deeper, more robust supply chain maps that inform and strengthen their supply chain risk management (SCRM). 

AI platforms can procure orders, receipts, customs declarations, freight bookings, and other documentation that create a chain of custody for your product, components, and raw materials.

Risk Assessments and Risk Matrices 

A risk matrix is a visual tool that helps organizations evaluate the various risks to their business and operations. The matrix typically synthesizes two key factors: the likelihood that a risk will occur, and the severity of that risk if it does come to pass. These factors are then given various categories along an x- and y-axis, allowing risk management professionals to assess and contextualize just how serious and threatening a particular risk is. (A cyberattack, for example, may be assessed as having a low likelihood for a company with robust cybersecurity measures in place, but a high severity.)

While drawing on a reliable risk matrix can be a key strategy for SCRM teams, there is a measure of guesswork in the way we assign the respective levels of severity and likelihood to risks like factory shutdowns, extreme weather events, or supplier bankruptcies. It is this specific responsibility that AI can fulfill very effectively. Instead of relying on personal experience, imperfect preconceptions, and limited data sets when using a risk matrix, AI technology can process substantial amounts of information and produce highly objective, accurate assessments of a risk’s applicability to your company and industry. While undoubtedly a niche task, utilizing a risk matrix to evaluate threats with greater precision can help organizations allocate resources more efficiently. 

Identifying Alternative Suppliers 

The work of finding alternative suppliers—also known as multisourcing—is another relatively simple, straightforward function that AI tools can help companies carry out with greater rigor and comprehensiveness. A slew of companies currently offer AI models that can provide businesses with a list of potential suppliers based on specific inputs and parameters. These models not only scour the web and find manufacturers that fit the specified criteria, though. They also sift through other determinative factors like company financials, ESG performance, and customer satisfaction to give organizations a list of thoroughly “vetted” candidates. 

Because of the sheer breadth of research that AI models are able to perform in a short period of time, they are also able to turn up a broader, more all-encompassing roster of companies that might not have been as accessible through more traditional, manual searches. This includes small to medium-sized businesses (SMBs), more diverse suppliers, and other oft-overlooked organizations. 

While this specific function isn’t necessarily transformational for supply chain management teams, it does give organizations the resources to cultivate greater resilience in their SCRM. In an article for the Harvard Business Review, a supply chain management executive with Siemens noted how an AI application was helping his company compile the information on alternative suppliers so pivotal to being able to successfully mitigate future supply problems. While “Technology doesn’t give you visibility to reliably prevent supply disruptions before they happen,” the senior director told HBR, “it can give you information that can help you respond to supply-chain disruptions much faster than human buyers can.” 

The Potential Hazards of AI in Supply Chain Management

The aforementioned use cases—along with the monumental investments from the corporate sector—testify to the extensive promise of artificial intelligence tools and models in supply chain management. Such a powerful, generational technology does not come without pitfalls, however. Manufacturers looking to incorporate AI should be aware of the hazards endemic to the territory. 

• Extensive upfront costs
• Staff skepticism 
• Unreliable, unvetted data

A majority the platforms currently on the market, from both established tech giants and niche start-ups, come with significant upfront costs. Many custom AI solutions, to use just one example, can run well into the six figures. Organizations should be discerning and methodical about such investments, using tools like cost-benefit analysis and potential return on investment (ROI) to gauge the net value of bringing on the technology. 

In addition, these platforms require broad, well-earned buy-in from employees, and executives and managers should not be eager to jump to unilateral decisions about integrating them. Without sufficiently communicating the circumstances, use-cases, and parameters under which these tools are going to be utilized, companies risk alienating staff, compromising employee morale, and even triggering attrition. 

Finally, businesses that start acting on the various predictive analytics offered by AI solutions too quickly risk overestimating the credibility and integrity of these solutions. As anyone who has observed the misstatements and factual inaccuracies produced by ChatGPT knows full well, AI tools are at the mercy of the information being fed to them. Flawed, dated, or “dirty” data can lead to unreliable assessments and analysis, and users should tread carefully in the early stages of implementing AI tools that dredge up inputs from every unvetted recess of the internet.