Is AI taking supply chain planning to the next level?

From the January-February 2025 edition of Supply Chain Management Review.

January-February 2025

As much discussion and deployment of artificial intelligence took place in 2024, 2025 is shaping up to be an even bigger year. This year will likely see the acceleration of AI, and specifically Generative AI, into…

Browse this issue archive.

Access your online digital edition.

Download a PDF file of the January-February 2025 issue.

The supply chain planning process hasn’t changed much over the last 10 years to 20 years. It’s still about orchestrating the many interconnected activities and steps that go into optimizing the flow of goods and services in the supply chain. What has changed are the external factors influencing the planning process: the “black swan” events that interrupt global networks, customers who expect ultra-fast deliveries and geopolitical factors are just some of the forces that are exerting new pressures on supply chains and, by default, the planning processes that orchestrate those critical networks.

The good news is that technology has evolved to the point where it can now play a leading role in helping organizations address these issues—sometimes before issues spiral into major problems—and keep their global networks running. Artificial intelligence (AI), machine learning and generative AI (GenAI), are some of the game-changers that can transform supply chain planning into a more proactive, data-driven, and resilient process.

AI comes of age in SCP

Artificial intelligence made its debut in the supply chain planning space about 10 years ago, when machine learning began enabling “touchless predictions” based on data that went beyond just shipment history. “AI continues to create new waves now, with GenAI being the fastest-moving hype on the Gartner SCP technology hype cycle,” says Jan Snoeckx, director analyst at Gartner, Inc. “We haven’t seen an innovation travel this fast in the past decade.”

Responding to the trend, vendors began expanding their solutions and system integrators started leveraging GenAI in their service offerings. For example, a common use case is an AI-based assistant that can navigate the plan and provide fast and efficient insights on it. Snoeckx says the advanced technology also supports higher-quality decision-making across the end-to-end supply chain. Most organizations strive for this as they find ways to drive the uncertainty and variability out of their global supply networks.

“AI holds the promise to not only automate, but do it in an intelligent way while improving core planning activities, data and reporting,” says Snoeckx, who adds that the “explainability” of specific planning decisions has always been a challenge for companies. For example, there may be several reasons why one order is favored over another (e.g., customer prioritization, available capacity or current inventory levels may play a role in the decision). Snoeckx says AI and GenAI can help answer some of the “why?” behind those decisions which, in turn, helps streamline the planning process.

Snoeckx sees more AI, GenAI and machine learning pilot projects advancing to the next level in the supply chain planning space in 2025 but cautions that there is still some work to be done before mainstream adoption can happen. “We do expect quite a few GenAI pilots being translated into real-life use cases, but there’s still a need for further technology investment in this area,” says Snoeckx, who does see a “healthy appetite for planning technology right now” from individual companies.

Whether those organizations make the right choices when evaluating than acquiring the AI- or GenAI-infused planning applications remains to be seen. “The main challenge will be making the right choice at a time when there are so many different solutions available on the market,” Snoeckx cautions. “While the core planning technology remains fairly constant, there are so many data and analytics platforms that have AI coverage; navigating the market isn’t always easy.”

Proactive thinking & accurate historical data

 Formerly the consumer healthcare division of Johnson & Johnson, Kenvue, Inc., ships its well-known brands like Aveeno, Band-Aid, Benadryl, Listerine, Tylenol and Neutrogena to 165 global markets. Meri Stevens, COO, says the company places a “huge” focus on supply chain planning as it works to discover the next marketplace need, develop consumer health products, find the ingredients to make those products and then deliver the final product to the consumers that need it.

“New ideas are coming to light all the time here,” says Stevens, who has been with the company since it was a J&J division. “We want to be able to understand those ingredients and how they’ll impact our formulation. That way, we can continuously provide the best science to our consumers.”

Pulling all these elements together requires quick, proactive thinking and accurate historical data, both of which drive the supply chain planning process. By infusing AI and GenAI into that process, Kenvue can more accurately pinpoint its end customers’ needs and then develop short- and long-term plans that ensure a steady source of supply. For example, GenAI helps the company “drive better decision-making on how we place bets,” says Stevens. “We’ve invested quite a bit in demand sensing to understand what indications will help us position products in order to be able to meet and anticipate demand.”

“These AI and large language models (LLMs) are data-hungry. The more data you can give them, the better off the outcome will be, so the biggest challenge is ensuring that we not only have the transactional data that drives the AI models in order to be able to deliver great outcomes and decisions, but that the data quality is excellent.”

The technology also helps Kenvue improve its inventory management processes, schedule transportation and create accurate forecasts. This leads to better outcomes, less system waste, cost improvements and optimized inventory levels. But before any of those benefits can be reaped, companies need to have their data houses in order. This is one area where Stevens says Kenvue is continuously improving and honing.

“These AI and large language models (LLMs) are data-hungry,” she explains. “The more data you can give them, the better off the outcome will be, so the biggest challenge is ensuring that we not only have the transactional data that drives the AI models in order to be able to deliver great outcomes and decisions, but that the data quality is excellent.”

You wouldn’t drive a race car 5 mph

The core sales and operations planning (S&OP) process hasn’t changed much over the last 25 years. Terms like “integrated business planning” may be thrown around a bit more, but the foundational elements are still conducting a demand review (week one), a supply review (week two), an inventory supply balance review (week three) and then replenish/manage as needed (week four).

“This is the same process that’s been in place for 25-plus years,” says Ashutosh Dekhne, supply chain & operations practice leader at EY Americas. “Sure, we’ve moved from pen and paper to whiteboards, Excel and advanced planning systems, but the fundamentals are still the same.” All this to say, there’s definitely room for improvement on the supply chain planning front, and technologies like AI, machine learning and GenAI may just be the best candidates to jolt planning out of its comfort zone.

Unfortunately, many companies view those technologies as “add-ons” that will boost planning efficiency in certain areas, all while sticking to the same planning cycles that they’ve always relied on. “There’s no point in buying a race car and driving it at five miles per hour. It doesn’t help,” Dekhne points out. Instead, companies should flip the equation and look at how they can reduce end-to-end planning cycle times in a way that enables more frequent planning.

“If it takes you four weeks to complete the cycle, you’ll be discussing what happened during week one around two weeks to three weeks later,” says Dekhne. “That defeats the purpose of asking for end-to-end visibility and real-time information because you’ll never be able to act on that information fast enough.”

Vera Trautwein, senior supply chain expert at McKinsey & Co., says companies that want to leverage more AI in their planning processes should also know that it’s not enough to just “plug and play” GenAI into an existing system and then sit back and watch the magic happen. Some core fundamentals have to be put in place first, she says, even if that new advanced planning system comes with built-in AI and machine learning capabilities.

“Particularly on the supply side, we’re not at the point where one AI- and machine learning-heavy APS can solve it all,” Trautwein says. “Add GenAI on top of that and it won’t get you to the anticipated step change. You really have to put the foundations in place first, and that requires end-to-end data connectivity that can feed the AI and machine learning [algorithms].”

How to avoid “pilot hell”

To organizations that are rolling out their first AI or GenAI supply chain planning initiatives, Stevens suggests starting with the actual problem that needs to be fixed, versus evaluating and piloting myriad different solutions. Doing so will reduce the time, money and hassle involved with shopping for, selecting and trying different applications.

“It’s easy to get caught up spending a lot of money chasing shiny objects in this space,” she cautions. For example, look at the constraints or “flow blockers” in your ecosystem and then use them to focus on first steps. “This will help you avoid ‘pilot hell,’” Stevens adds, “where you’re just testing out multiple systems without ever creating something that can scale to deliver value.”

SC
MR