Supply Chain and Logistics

Warehousing, logistics, supply chain, and aftermarket offerings are fertile ground for AI. When we asked manufacturers to rank which areas of their operations would see the most transformation by AI over the next three to five years, supply chain and logistics ranked number one. This part of the value chain contains large amounts of freely flowing information already, and heightened geopolitical volatility makes manual analysis of this data based on traditional parameters insufficient. AI’s ability to broaden the aperture of analysis and rapidly move decision-making from reactive to predictive delivers high rewards. 

MSA Safety is using AI to improve demand forecasting. “We were not happy with our previous automated demand planning tool. It lacked sophistication and accuracy on top of being really expensive. So, we wanted to reduce cost while improving forecast accuracy, and we created an AI assistant to help us improve forecasting for a very narrow area and a specific range of products,” Chief Information Officer Heiko Will explained. “We’re seeing pretty good results so far. The accuracy went up by about 25%. As we add more parts of our portfolio to the tool, we expect the cost to go down significantly, probably by more than 50%,” Will said. Scaling the solution is the next step, Will noted, adding, “It's relatively easy to expand this to a variety of other forecasting tasks, such as financial forecasting, parts forecasting, general forecasting.” The ROI is straightforward. “There were clear costs associated with our previous forecasting tool, so the ROI on that is relatively easy to calculate. In addition, when the AI tool is applied to an area such as parts forecasting, it may lead to less stock or better precision with sourcing, so it translates relatively easily into business value,” Will said. 

Lincoln Electric is using AI for dynamic inventory management because, “With our original system, we were not able to run the right level of scenarios as external conditions changed,” according to Naïty Jacel, Vice President of Digital Supply Chain Strategy. It was also difficult for the company to see what was happening with suppliers and its locations around the world because of the company’s varied base of customers and suppliers. 

Working with Palantir, Lincoln Electric was able to create one connected model of its operational world including its plants, SKUs, orders, costs, vendors, and forecasts. This allows every application to read and write from the same source of the truth. With this AI engine, Lincoln Electric can perform deep dive analysis on materials, flag potential shortages and excesses proactively, and make proposals for corrective actions. While the system has the ability to make decisions autonomously, Lincoln Electric has opted to keep humans in the loop for now. “We are just not at that level of maturity and comfort yet. This was a conscious decision on our part, and it’s not necessarily the decision that everybody will make or that we’ll make in the future. We are just introducing the technology, and we thought it was a step too far to completely automate everything,” Jacel said.

Material harmonization is an important use case helping manufacturers engage more efficiently with their suppliers. “We're working with companies that purchase thousands and thousands of parts and raw materials and helping them leverage AI to simplify and reduce costs in the purchasing process. For example, AI systems tell them where a part could be substituted for a better price based on specific boundaries set by engineering, compliance, and other members of management,” said Sam Batey, Head of AI at Foxtrot. AI agents can be added as a second layer of intelligence that proactively contact suppliers based on design specifications and purchasing needs. “It’s like giving procurement Google Maps but constantly rerouting to the most efficient option across the entire fleet,” Batey continued. As a result, the “manufacturer has the transparency and ability to make a better decision because the AI agent can go through thousands of complex pages of design documents and discover, for example, that the screw from a different supplier costs 10% of the one you’ve been using for the last 10 years,” Batey said. 

GALLO has deployed AI for dynamic order fulfillment. “We deployed machine learning models to dynamically select the right fulfillment distribution center (DC) based on cost, lead times, and other parameters,” Nitin Murali, Vice President of Supply Chain Excellence explained. Then GALLO added AI on top of the machine learning models to enable prediction. “The AI can predict that a particular distributor is going to order certain materials, and those materials can be shipped most efficiently from certain DCs. We put some anticipation logic against it as well,” Murali said. Shifting from manual processes to AI dramatically cut decision time and costs. “We saw close to a $1,000,000 in savings very quickly,” Murali said.

GALLO has moved largely to agentic AI for dynamic safety stock decisions. “AI is able to look at the forecast and make a prediction and a recommendation,” Murali explained. “For example, it may say, ‘Hey, you probably need to increase your safety stock to this level by April. If you don't, you will have a problem in June because of the volatility in demand that we're seeing from these specific distributors.’” GALLO has gradually increased the proportion of decisions that it allows to happen autonomously. “We didn’t trust the AI until we validated it, so we started with approximately 10% of the decisions being made by AI. Now the system has proven itself, and more than 90% of the decisions are made entirely by AI,” Erich Kaepp, Chief Operations and Supply Chain Officer at GALLO said.

Other companies are using agentic AI for raw material procurement based on real-time supply levels. Luis Lopez Garay, Partner at Crowe, shared the story of a client in the cement industry that “set up a camera right in front of their raw meal silos containing materials such as limestone, clay, etc. When the quantity of the silo gets below a certain level, the AI agent automatically places a purchase order to buy additional raw materials.” Cement producers can also use AI agents to communicate with a sales database to prioritize which silo to fill based on upcoming orders. 

Agentic AI for tail spend management has been successfully deployed by several manufacturers we interviewed. One manufacturer set up an AI agent to negotiate with lower tier suppliers whose sales fall under a certain dollar amount. A senior executive told us: “We have over 10,000 suppliers, so previously the company was not negotiating with the smaller ones, as is typical for most manufacturers. After deploying the autonomous negotiation tool, we saw savings within the first couple of weeks. In addition, we’re hearing from these businesses that they enjoy the chance to be more competitive. We've actually been able to bring some up into our higher supplier tiers where we have deeper relationships, so it’s a win-win.” 

In contrast to the small business approach, one manufacturing partner is helping companies find new ways to interact with their largest vendors. A senior executive related an example of autonomous negotiations with large portfolio companies selling everything from multimillion dollar solutions to spare parts. “When it comes to very low dollar, high-volume items, these aren't really worth a professional buyer's time.” The purchasing company can set up agentic negotiations allowing its buyers to respond back to either reject or accept the offer. “They’re driving savings through this new process,” he added. 

Aftermarket services for assets in the field are a promising growth area for AI. “Manufacturers have assets with their customers and their channel, so there is an opportunity to look across the value chain to find new sources of revenue from both service and parts,” Farooque Munshi of EY told us. In addition, “manufacturers are trying to leverage their inherent strategic advantage to unlock differentiated services with their customers,” Munschi added.