Fans of “The X-Files” might remember an episode during the sci-fi show’s first season in 1993 that involved artificial intelligence taking over operation of an office building. Astonished that AI could do this, FBI agent Fox Mulder exclaimed, “I thought it was only theoretical.”
While the AI in the TV episode 33 years ago didn’t end well for the scientist who created the system, AI has been far more than theoretical for building management for several years. That’s particularly true for a number of South Carolina manufacturers leading the way in implementation of this technology for supply chain management. ZF Transmissions in Gray Court and Hartsville-based Sonoco are among manufacturers in South Carolina reaping the benefits.
ZF uses AI in freight optimization and inventory management. Early benefits include cost savings, increased truckload density, and a reduced carbon footprint.
Barath Indrakumar, ZF Group’s vice president of digital business development and solutions, says the company started focusing on deploying “impactful” AI solutions in 2024. The impetus was the launch of ChatGPT in 2023. Indrakumar leads the implementation of AI in supply chain and logistics.
“With the launch of ChatGPT, and the availability of mature models and being able to access more data from our business systems, we kickstarted our journey focusing on AI solutions,” he says.
ZF uses AI for product innovation and development, and to boost productivity.
“We’ve experimented and implemented a number of ChatGPT and co-pilot solutions for requirements in engineering, software development, testing, and calibration in our product portfolio,” Indrakumar says. “We also use AI to optimize productivity in our business processes. AI has helped us streamline workflows, reduce manual tasks, and enhance decision making. We've been able to generate quite a bit of efficiency with automation of workflows and repeatable tasks.”
ZF operates about 200 plants worldwide. As of late February, about a third of them are using AI in some way. The intelligent inventory management solution, which launched in 2024, has been implemented in about 15 plants.
“We’re at the beginning stages of this journey,” Indrakumar says. “Typically, the way this works is, we come up with a solution. For example, a solution for inventory management, where we come up with an intelligent inventory management solution, we typically develop a solution with a lead plant and test it out. Once it is mature and working, we roll it out to other plants. It is a journey that takes two to three months.”
ZF identified areas in its freight operations where AI could be used to optimize truck routing to achieve savings.
“We worked closely with the plant to get all the packaging data pulled together,” says Joshua Walasky, head of supply chain management, ZF North America. “We figured out what other systems needed to be connected in order to have the AI start looking at our trucks, at where our material is coming from, and how we can improve our load density. It also looked at how we can update the routing, in order to maintain that increased load density. "
Walasky says the challenge was getting data in the correct format, so it can be executed properly. Once that was done, the speed of opportunity identification increased, so AI started showing where ZF could consolidate loads and increase the overall load density. ZF then adjusted routing, in order to support getting that load density through the systems.
AI not only identified solutions faster, but ZF was also able to execute it quicker, because it doesn’t require employees to manually go through different aspects of shipments.
“AI not only identifies what we can do, it also gives visualizations of what the load would look like. You can visually show the plant and how the load will be constructed. This is how full the truck will be. It made it a lot easier for the teams to get on board and accept these opportunities, because they could physically see what was happening to the load.”
ZF wound up saving about 3 percent on the specific freight it was analyzing, a rate Walasky described as “fantastic.”
“Not only do we get the density improvement, we get cost savings. We get a lower carbon footprint. We have fewer trucks on the road, reducing our overall exposure and risk to accidents, and reducing the amount of trucks moving through the plant.
“Theoretically, you've increased your safety. The plant has been very active in supporting these activities and opportunities that AI is showing us, and we've had overall improvement since we've launched this last year.”
The Gray Court plant, ZF’s largest in the U.S., was selected for the AI pilot program due to its volume of business. The 1.7-million-square-foot facility manufactures transmissions for passenger cars, light trucks, and commercial vehicles. A 45,000-square-foot expansion is due to be completed this year.
“This tool provides us a recommendation,” says Anderson McMaster, senior manager, supply chain and logistics at ZF’s Gray Court plant. “It tells my team, here's potential optimization we see from a freight fill rate or density. Out of the investigations that we saw last year, we had 26 different proposals that the AI spit out for optimization. Of those 26, we implemented 20 of them. That means that not everything that the AI spits out we move forward with, and that's where the human element is important.”
McMaster says his team then decides whether to move forward with the recommendation from AI, or put it on the back burner for further analysis. The tool provides a recommendation based on ZF directives when it comes to inventory management, as well as things like lead time, customer demand, and minimum order quantities. All of those supply chain factors are considered, and the tool produces a recommendation based on the parameters that ZF has in SAP.
“It will tell us we can reduce our safety stock, based on a supplier's performance, or how the demand, depending on how stable it is, has been over the past few weeks or months. We can make adjustments in our inventory. Just like the freight management tool, my team receives those recommendations on a daily basis, based on the system, and then they are making the decision whether or not to implement those changes.”
Global packaging manufacturer Sonoco has been gradually implementing AI across its operations since 2021. Rajeev Ankireddypalli, Sonoco’s chief information officer, says the company has been using AI in its supply chain for a few years, with a strong emphasis on practical, productivity-focused use cases, rather than experimentation for the sake of AI. Many of Sonoco’s early and ongoing deployments are designed to help employees work more efficiently by reducing friction, accelerating decision-making, and improving consistency in everyday work.
“One example is language translation, where we use AI within our own secure environment to support global collaboration,” Ankireddypalli says. “By doing so, we’ve been able to significantly reduce reliance on external translation vendors, resulting in meaningful cost savings, while also improving speed and accessibility for our teams.”
Sonoco also focuses on embedded AI, or capabilities already built into the enterprise applications the company uses. That allows Sonoco to take advantage of AI functionality provided by its technology partners without adding complexity or risk, Ankireddypalli says.
“We are careful and deliberate about where we deploy AI. We refer to this approach as ‘purposeful AI.’ Every use case is evaluated against clear value measurements, and we focus on scaling only solutions that meet our criteria for data readiness, process maturity, and tangible business value.”
Ankireddypalli says Sonoco is upgrading its supply chain and demand planning systems to fully capitalize AI capabilities. Sonoco expects this transformation to significantly improve supply chain efficiency and effectiveness.
“By embedding AI, our Supply Chain Team will move from reactive decision-making to proactive, data-driven strategies,” he says. “Beginning in the second quarter of 2026, our enhanced systems will allow us to use sophisticated machine learning models, and tap into external data sources, including market trends, weather patterns, and supplier risk indicators, to generate more contextual, highly accurate demand forecasts.”
AI will allow Sonoco to automate its entire planning cycle, identifying forecast errors, lead-time deviations and configuration issues in real time. It will then recommend, or automatically implement, corrective actions. They will also support planners by releasing orders at optimal times, minimizing manual intervention, and enhancing consistency across all planning activities.
“Our system transformation will also enable us to leverage generative AI, revolutionizing how we plan, communicate, and collaborate,” Ankireddypalli says. “Generative AI will automate administrative tasks by summarizing order changes, creating planning notes, and drafting customer communications.”
This will reduce administrative work, freeing planners to concentrate on strategic initiatives, rather than routine paperwork. Advanced machine learning for inventory optimization and replenishment will enable Sonoco to detect forecast variability and offer recommendations for demand shaping and restocking. This approach will help maintain ideal inventory levels, lower carrying costs, and improve service reliability.
“Sonoco’s vision for AI in the supply chain is to empower us to anticipate and respond to both internal and external customer needs more quickly and accurately,” he says. “By integrating machine learning, generative AI, intelligent agents, and unified enterprise data, we intend to build a supply chain that is not only more efficient, but also truly adaptive to ever-changing demands.”
What’s next? Sonoco plans to scale use of AI across all business operations. Ankireddypalli says AI is part of the company’s business strategy to make work more intuitive, simplified and automated. In procurement, this involves expanding use of AI-driven sourcing platforms to guide procurement teams throughout the entire sourcing process. In logistics and supply chain operations, Sonoco is advancing toward proactive, AI-enabled planning and execution. By embedding machine learning and intelligent agents into planning and logistics systems, the company seeks to move from reactive responses to near real-time predictive and prescriptive actions.
ZF’s Indrakumar says his company will try to integrate all of its independent or disparate AI tools along the value chain, so they will feed into each other to improve throughput and quality. This includes tools assisting with intelligent inventory management, customer forecasting, and production scheduling.
SIDEBAR
Obstacles Still Exist for Companies Implementing AI
Daniel Kwasnitschka, an assistant professor in management science at USC’s Darla Moore School of Business, works with industry in assessing new digital technologies. He says that while a lot of companies are trying to implement new technologies, very few can capitalize on these investments. One obstacle is achieving connectivity, which is especially prevalent in the U.S.
“A lot of times, data is still trapped within very old machines or legacy systems,” Kwasnitschka says. “It's hard to extract and clean in a way that you can feed into these AI models. Some of the machines, especially in manufacturing, are decades old. AI only works if you have a lot of data. If you don’t, the models will not be very accurate.”
Companies with legacy manufacturing execution systems for production and supply chain planning may need to first implement technology to capture data from their old machines. Kwasnitschka says many companies are doing that now.
Route optimization and improved root cause analysis are among the advantages of using AI, which can reduce the time involved.
“How do I get my products produced, and get them from A to B in the most efficient and cost-effective way? Optimization models we do here at the school, or in research in general, benefit greatly from AI. You gain a lot of speed through AI. You gain improved root cause analysis. AI can figure out a causal relationship in your production process. And what this means is that AI models can figure out if there's an issue in your production process, and you have to scrap some products, or if you detected defects, you can pretty fast figure out the root cause.”
Another challenge is to get people to use a new technology.
“A lot of them don't use it for a lot of different reasons. Some obviously don't trust AI. Others think that they have been working with the company for 20 years and they know it better than AI. The technology eventually can replace them, maybe. So why should they? So far, I can say that with most of the AI tools that are used, the human being is in charge on the shop floor.”
Comments
No comments on this item Please log in to comment by clicking here