Supply chain professionals know all too well about the “bullwhip” effect—a phenomenon in which small fluctuations at a supply source can cause progressively larger fluctuations downstream. But increasingly complex supply chains and unexpected disruptions have exacerbated this classic problem in recent years.

The pandemic may have been the loudest wake-up call, but new interruptions emerged, warring factions in the Red Sea, a drought in the Panama Canal, and most recently, a container ship crash that felled a bridge and blocked the Baltimore harbor, one of the busiest ports in the United States.

“Planners are in an epic battle against this fundamental supply chain problem, which has grown worse over time,” says Adrian Krueger, associate director at Accenture. “More than two decades ago, we started using data and algorithms to help with this problem, but the growing complexity and disruptions have amplified the issue. They have created a massive strain on a human’s capacity to deal with it.”

Generative AI (GenAI) has the potential to help businesses cope better with a range of supply chain problems. In the near term, however, most companies are limited to using chatbots that leverage GenAI's ability to process and automate language and media.

Rather than having to run reports or even rely on complex dashboards for information on the status of supplies, for example, planners can use natural language prompts to ask the system for typical lead times on key components.

In the longer term, GenAI will serve as a translation layer between existing AI and machine learning (ML) techniques—which can be quite complex and opaque—and human workers. Ultimately, the combination of GenAI with traditional AI could bring together multiple types of data—from within the company (inventory levels, sales data), throughout its supply network (contract manufacturers, distribution centers, third-party logistics providers) and conditions around the world (weather patterns, terror attacks)—and recommend how to make supply chains more resilient, even in real time.

“Right now, there’s so much data that it’s hard to synthesize it into something actionable,” says Subit Mathew, principal in Deloitte’s enterprise performance—SAP consulting practice. But when companies learn to use GenAI appropriately, they can free employees from tedium to use their brains in higher-value work, while applying technology to synthesize more data more effectively. The combination could be an elixir that brings increasing value. “In the future, supply chains will be steered by human creativity and powered by intelligent technology like AI,” says Mathew.

To get there, however, companies need to improve their data collection and data quality. They also have to figure out how to smoothly integrate GenAI into the heart of supply chain planning, rather than just around the edges. Meanwhile, software vendors are working to support customers by incorporating GenAI capabilities within their current systems.

More on supply chains

Key Supply Chain Issues, Explained

Uncertainties wrought by a global pandemic and climate change are prompting fresh thinking about planning, sourcing, and sustainability.

The New Era of Demand Planning

Sensing and responding to fast-changing demand signals requires analyzing more data—and more recent data.

How Manufacturers Meet the Scalability Moment

To pivot faster, product makers are investing in workers, deploying smart factory tech, and rethinking processes.

Guide to End-to-End Supply Chain Sustainability
Learn more about sustainable processes.

GenAI enthusiasm in supply chain

Executives are excited about the potential for GenAI to improve supply chain processes. In a November 2023 Gartner survey, 80% of supply chain leaders said they would begin or continue to implement GenAI in the next 12 months.

Initial implementations are chatbots and similar applications, which can increase efficiency, lower costs, and enhance productivity. More substantial benefits of GenAI will come through a gradual process of combining GenAI with current technologies, particularly traditional AI and ML. In fact, experts predict that GenAI could fade into the scenery, becoming just another feature in an application.

Supply chains have used certain types of AI and ML for more than a decade, notes Randal Kenworthy, a senior partner at West Monroe, a management consultancy. AI and ML techniques are incorporated into many supply chain tools and platforms. (SAP’s software has used ML algorithms for data mining and forecasting for years, for example.) GenAI will be a further evolution, becoming “an extension of existing platforms, to help companies do what they’ve been doing in supply chain planning and analytics for the past 10 to 15 years,” Kenworthy says. “There's not going to be a generative AI project. There's going to be a tool they've already deployed that will incorporate generative AI capabilities.”

Other analysts agree. “Supply chain leaders must recognize that, while [GenAI is] incredibly powerful at what it does well, it’s not suited to every task,” says a 2024 Accenture publication. The publication recommends viewing it “as part of a continuum of automation capabilities that include traditional process automation and classical machine learning models, as well as large language models.”

Use cases in three waves

GenAI will likely be applied in supply chain management in a series of three waves.

The first wave is simple applications for automation and augmentation.

In the Gartner survey, executives said they were already seeing reduced costs and increased productivity from such uses. For example, planners can ask a chatbot in natural language (and in any language) to summarize reports or search log files for errors. GenAI can scan and compare numbers and terms in bids or contracts. Chatbots can also answer specific questions, if built on the right architecture, with guardrails for data security and responsible AI. Accenture has built a “supply chain navigator” that can answer questions to help identify possible vulnerabilities in a supplier network, says the firm.

GenAI can run “what if” scenarios, says Michel Roger, managing director of supply chain and operations in the SAP business group at Accenture. “For example, what if my supplier provides only 70% of the raw material originally committed? Generative AI can help planners prepare for various scenarios.”

In the second wave, GenAI will help improve user experience—and sharpen decision-making—by explaining existing technologies.

Recommendations produced by AI and ML in current supply chain platforms are sometimes disregarded because people don’t understand how the technology arrived at those recommendations. And if they don’t understand the logic, they will distrust the recommendation, especially if it goes against their instincts.

“With the complexity of today's supply chains, it's difficult even for an expert to understand why they should shift certain things,” says Roger. “Sometimes when you don't have the big picture, it doesn't make sense.”

At SAP, we have noticed, for example, that sometimes users ignore recommendations by our software’s supply optimizer, which applies various constraints and then presents supply plans optimized to maximize delivery times or profits. GenAI could “show the work” and explain the optimizer’s logic to make it clear to the planner. Similarly, GenAI could review system alerts—which can be so numerous they are overwhelming—sorting them by priority so planners only need to attend to those most important. Taken together, these changes have the potential to improve not only the user experience but also the practical results users gain from the systems.

Gartner thinks the use of GenAI with traditional AI and ML “is actually going to be a virtuous cycle,” says Noha Tohami, distinguished vice president and analyst at Gartner, who spoke at a recent Webinar about supply chain leaders’ 2024 plans. “GenAI is going to result in more usage of these technologies and more return on investment.”

A third wave’s potential: innovation through deep analysis

In the third wave, GenAI could help planners analyze the deluge of information and indicators that no human mind can capture and interpret. To name just a few: Shifting weather patterns, new trade rules and tariffs, changing locations of key sources, wars and terror attacks, and climate change. The hope is that the combination of GenAI with traditional AI could analyze these factors and recommend actions to make the supply chain more resilient, even reacting in real time.

This capability could also help companies overcome departmental silos that prevent important data from being used in supply chain planning. It’s difficult, for example, for businesses to factor in the full range of sustainability considerations when supply chain managers are picking sources. When engineers design a product, they often lack full visibility into the availability of all the components needed to build the product. If companies could bring such information together, they could make more progress in transforming their entire business. They could operationalize sustainability.

In doing so, GenAI and ML can give supply chain managers a better understanding of demand patterns and disruptions that can affect supply chain flows, with more precise insights about production and inventory to strengthen resilience (and diminish the bullwhip effect).

With the right data and communications infrastructure, you can increase productivity and accuracy.

Challenges

Before they can reap these substantial benefits, however, businesses need to overcome two challenges in their supply chains: lack of good quality digital data and the difficulty of integration.

The lack of useful data, for example, leads to an incomplete picture. A Gartner September 2023 survey found that a majority of the supply chain environment is not captured by current digital systems, which sometimes leads to bad decisions. “More than half of the supply chain leaders reliant on digital technology to make a strategic decision told us that they felt they would have landed on better decision outcomes without the use of their models, and our analysis suggests that they are correct,” according to Susie Petrusic, senior director analyst in Gartner’s supply chain practice.

The lack of data also limits the ability to act. In a May 2023 survey by Oxford and SAP, only 37% of logistics executives said their organizations could capture data and act on it in real time, and only 14% actively use AI and predictive analytics to capture real-time insights. Moreover, when it came to implementing intelligent technologies, increasing visibility in the supply chain took a distant fourth place, with only 10% of executives choosing it.

Integrating GenAI into the heart of supply chains is a different category of challenge. Companies are already adding GenAI features to existing systems, like a chatbot to allow interrogation or a GenAI engine to summarize documents. However, integrating GenAI more deeply into the supply chain planning process, to help systems draw on all available data to make better decisions will be a much heftier lift.

Gartner analyst Tohamy says during her Webinar that integration with legacy systems could be “a massive hurdle. … We are very early in the process, and few organizations at this point are even attempting to integrate GenAI insights back into their planning and execution systems.”

Recommended next steps

Nevertheless, companies can take steps now to implement first-wave applications as well as prepare for the subsequent waves.

• Start with what GenAI does well. Right out of the gate, its ability to help employees find answers to questions, scan and search documents, and summarize information can make them, and your company, more productive. Just make sure the data and documents on which you're training the models are correct, consistent, and unbiased.

• Focus on solving a business problem. Avoid the temptation of doing “science projects” without clear strategic intent. “Don't do generative AI for the sake of generative AI,” says Krueger. “Look at it from a business process value perspective.” Mathew suggests evaluating potential projects through two major lenses. First, what is your goal in terms of business value? Second, what capabilities are in place to carry out the plan and govern its operation?

• Look for opportunities to bring other data into supply chain planning. Take care, though, to evaluate the usefulness of such data. Some companies spread their net far afield, collecting information from social media, for instance, without thinking about its accuracy, validity, and utility. Other departments in your own company may be sitting on good data that is relevant but not incorporated into the planning process. In addition, evaluate your infrastructure connectivity. Networking challenges frequently hold companies back from using data at scale.
  
  With the right data and communications infrastructure, you can increase productivity and accuracy. Yogurt company Danone, for example, won an SAP innovation award in 2023 for a project that used 30 types of internal data, including marketing and sales data, to transform demand planning. The project dramatically reduced the time spent in generating forecasts and improved forecast accuracy, which in turn reduced waste in finished goods and thus had a positive effect on sustainability.

• Consider how GenAI might help improve your data. A persistent challenge for supply chain managers is the quality and quantity of data. An original-equipment manufacturer’s contract factories, for example, may serve as key partners tasked with providing components, subsystems and even finished products—but the parties often lacked the trust to open their systems to each other. (For more, see “Contract Manufacturing: Supply Chains’ Blind Spot.”)
  
  To build better relationships, OEMs can share data about customers’ needs and process expertise to benefit their contract manufacturers—and in return—contractors can share their own data about supply flows. And GenAI can support the process of improving data collected from suppliers. For instance, optical character recognition systems have been used for years to scan documents into digital form, but they can be prone to error. Using GenAI to scan those documents could increase rates of accuracy.

• Keep context in mind. If you're trying to use GenAI to present data to users in a more natural way, make sure you consider the context in which questions will be asked. While a certain amount of training people on how to use the right prompts is necessary, the more you can train models with not only correct data but also the right context, the less detailed those prompts need to be.

In fact, the best path forward for many companies could be to rely on existing supply chain planning systems. Software companies, including SAP, are already considering things like how to build context into systems to lessen the need for prompt engineering. If done well, vendors could pave a smooth path for businesses to make the most of GenAI.

Kenworthy says that in West Monroe’s GenAI pilots, employees tend to shy away from stand-alone GenAI tools. “If it's embedded in existing software, we find we get a lot more traction,” he notes. In addition, while Fortune 1000 companies may have the time and budget to experiment with GenAI applications, most mid-market companies do not. They stand to benefit most when it becomes a feature of existing tools.

Ultimately, GenAI could be a key weapon in the battle to gain more visibility into and control over supply chains. From early chatbot applications to integrations that wring tremendous business value out of the supply chain, companies that can successfully apply a combination of AI, ML, and GenAI can make their supply chains ultimately more resilient and their businesses more profitable.