MHI Solutions Improving Supply Chain Performance
Sci-fi films may portray artificial intelligence (AI) as a threat to humanity, but in the real world of supply chain, AI provides many benefits to the humans who interact with it. As a powerful decision-making tool, AI helps workers be more effective and more productive.
The term AI encompasses many related technologies, including machine learning and generative AI (natural language processing), that simulate human thinking and problem-solving. It can be seen as the next step in the evolution of predictive analytics, which enable companies to look at data to understand future demands, and prescriptive analytics, which help them determine how to use their assets optimally.
“The power of AI is in the machines being able to learn by working with massive amounts of data, seeing patterns, learning based on those patterns and then applying what they’ve learned to put forward a recommendation on some business situation,” said Jagan Reddy, managing partner, Netlogistik US.
In the supply chain today, AI is used primarily for augmented decision-making. It sifts through data from a wide variety of sources to determine possible solutions, then presents the best options to humans for further evaluation and final selection. But AI also has the potential for autonomous decision-making, where it is authorized to make decisions on its own within certain parameters and then implement them. The transition to autonomous decision-making for routine tasks is probably inevitable, but will take some time.
“As humans, we want to be very certain before we let a machine take over. It’s a change we all have to become very comfortable with, because in the end, businesses are still led by humans,” said Reddy.
As AI handles the more mundane aspects of decision-making, people will be free to spend their time developing solutions for the more challenging supply chain problems that require real human creativity.
AI in Action
The supply chain decisions that people are making today involve more and more real-time data. “They’re dealing with all sorts of disruptions and events that are happening around the world,” said John Lash, group vice president, product strategy at e2open. “It’s getting harder to predict what demand is going to be. Suppliers are constrained by materials availability and transportation capacity; some days it is great, other days it is not.” Planning is no longer done on a weekly or a monthly basis, but on a daily or even real-time basis as information comes in.
People simply can’t process this flood of information, but there’s never too much data for an algorithm, he added. “Once you start connecting all these tiers, this flood of information is going to require AI to process it. And that’s going to open up the capability of making more connected decisions than is possible today.”
Without AI, companies can input variables such as projected sales orders and inventory availability to generate a consensus forecast. “But what AI and machine learning can do is see through all the other surrounding factors that might indirectly affect a forecast,” said Ram Krishnan, global head, customer success, Aera Technology. “You can throw in weather data from Google, or if you have access to Nielsen for competitive dynamics, you can add that.”
With this additional data, and with its huge computing power, AI may detect sales patterns that people can’t. Humans who look at reports generated by advanced analytics may be able to determine there is a 21-day cyclical pattern for their products, but the AI may find a more subtle 17-day pattern leading up to holidays like Easter. Awareness of that second pattern could impact a company’s buying, warehousing and transportation activities.
In a warehouse, the introduction of AI can free coders or technicians from the never-ending chore of updating static, rule-based warehouse management systems (WMS) so that they reflect current conditions.
“We’re seeing businesses where people have to constantly go in and change rules in the WMS. With the speed of change today, that’s not dynamic enough,” said Reddy. Now imagine AI taking in the data about these changing conditions, learning from it and incorporating that new information into the decision-making process automatically, without human intervention. The information in the WMS is constantly updated so that operational decisions are better.
Suppose a warehouse customer that had been ordering a product in cases and pallets for years is now ordering those same products as “eaches.” The AI can use that data to recommend a different slotting in the warehouse and/or a different picking method for that product.
Preserving Knowledge
AI is constantly learning from the data that is input into the system, from previous decisions that people and that the AI itself have made, and from the outcome of those decisions. This enables AI to preserve institutional knowledge. That’s of particular concern these days, when so many older workers are retiring and companies are losing decades of their experience-based decision-making know-how, according to Krishnan.
Imagine a scenario where an experienced buyer usually bought 1,000 units of a product each month from a supplier. Every so often, because the supplier had excess inventory, it would offer an additional discount if the buyer would take 1,500 units instead. Since the buyer had been on the job long enough to understand the company’s business cycles, she would accept that offer, knowing that the company could handle the extra inventory and save some money in the process.
Then that senior buyer retires, and a new, inexperienced person takes over. He is reluctant to take the supplier’s offer because he doesn’t know what to do with that extra inventory. If he asked the AI for assistance, the AI would be able to analyze past data, determine that taking the additional units had worked out well previously, and recommend that the new buyer accept the additional inventory and the additional discount.
The senior buyer would no longer be there, but her experience would still be able to inform the decision-making process. The new buyer would make the right decision because he had AI support.
“AI-driven decision automation enables companies to create a digital memory of every decision made, including the context and rationale with the ability to build a continual base of institutional knowledge for workers today and in the future. They would essentially be digitizing their secret sauce,” said Krishnan.
Conversing with AI
Until recently, one big problem with using AI for supply-chain decision-making has been the technology’s interface with workers. People require training and a willingness to learn programs that can extract the information they want from AI. With generative AI and large language models (LLMs) like ChatGPT, it’s much easier for people to work with AI in a familiar, non-threatening way.
Generative AI provides an intuitive way for humans to interface with AI. “We now use ChatGPT in our personal lives, where I can ask a question and get an answer about a dinner recipe or a vacation itinerary,” said Noha Tohamy, distinguished VP analyst, CSCO enablement team at Gartner. “What I see organizations focusing on right now with generative AI is giving their staff a better, easier way to get information about supply chain performance.”
Generative AI can also incorporate into its analytics text data from an organization such as policies, standard operating procedures, chat logs of a team’s Slack conversations or email. Feeding that information into the LLM tool provides it with more context for decision-making; if a question arises about inventory, the AI can answer using not only inventory numbers but also conversations about what happened on the floor that affected that inventory. “There’s a lot more rich context that’s captured from these conversations that can now be part of the analysis,” said Christopher Andrassy, CEO, Astral Insights.
Click here to read the full feature.