THE FUTURE OF THE AUTOMATED SUPPLY CHAIN
says. The temptation is to address every challenge with AI. But for some simple decisions, heuristics like “If A, then B” will do the job and be quicker to implement and easier to understand. If a warehouse has limited space, making it impossible to fill all demand all the time, a replenishment rule like “Reorder once inventory drops to 10 units” may suffice, Thiede says. AI solutions tend to be more complex to set up and may require more maintenance. As a result, they often make more sense when there’s enough margin to cover the additional costs. Some organizations are getting pushed to implement AI so quickly that they risk overlooking the need for guardrails, security, and governance. “Data science and AI is the wild, wild west,” Harvey says. Outliers or biases in a dataset or model could lead the algorithm to generate insights that don’t reflect reality. Without solid governance and human oversight, any decisions made based on the model could be sub-optimal. A starting point for addressing these risks is a thoughtful business case. “Don’t just sprinkle AI across everything,” Davis says. Prioritize and develop a roadmap to move toward the organization’s goals.
The AI compliance and governance model should provide alerts when the results go beyond an acceptable realm, Davis says. Humans should be kept in the loop, as they can decide how to handle a particular AI insight when their expertise indicates the solution is incorrect, Harvey says. For instance, if a solution is evaluating transportation routes and some routes were used in the past only because others weren’t available, that will influence the results. An experienced employee can pick up on this. SMALL OR BIG? Unlike many advances in technology that start with established enterprises and trickle down to smaller firms, some say smaller businesses, and particularly newer ones, may have an advantage when it comes to artificial intelligence. Newer companies can create native environments in which AI can thrive, says Talal Abu-Issa, chief executive and founder of Beebolt, a supply chain technology company. In contrast, bigger companies often are structured around certain processes and ways of doing things. “They’re not necessarily optimized for these models to really shine,” Issa says. DISMANTLING SILOES One overarching benefit of deploying artificial intelligence in supply chain and logistics is its ability to break down divisions between strategic, tactical, and operational decision-making, Winkenbach says. Traditionally, companies have tackled these separately because the issues are so complex that it hasn’t been feasible to assess them at the same time. Artificial intelligence and machine learning, however, can address multiple big problems at the same time, such as helping companies determine where to build their distribution centers, place inventory and decide on transportation modes, Winkenbach says. Ultimately, AI’s biggest bang for the buck may come from making complex decisions simultaneously.
MIT, MECALUX ACCELERATE WAREHOUSE AUTOMATION A new five-year collaboration between MIT’s Center for Transportation & Logistics and intralogistics leader Mecalux aims to drive groundbreaking advancements in warehouse automation. Through MIT’s Intelligent Logistics Systems Lab, researchers are focusing on two goals: enhancing the productivity of autonomous warehouse robots and optimizing order distribution systems.
Thieves often identify themselves as legitimate truckers then enter warehouses to pick up loads, which they transport to other locations and likely resell. Artificial intelligence could be used to track vessels and send alerts when a truck detours from the expected route. It also could identify shipments at greater risk of theft, whether due to their location, type of cargo, or other factors. Security efforts could then focus on these shipments. POTENTIAL RISKS Along with its promise, AI carries risks. One is, somewhat ironically, the lure of its promise. “When you have a new hammer, everything is a nail,” Thiede Robotic collaboration. The first area of research will develop a “swarm intelligence” system, enabling autonomous warehouse robots to work collectively, making smarter, coordinated decisions. The aim? To create robots that learn from human behavior for better efficiency and collaboration in dynamic warehouse environments. Predictive distribution. The second research focus is on training AI models to anticipate customer demand patterns. This approach will help companies operating extensive warehouse and distribution networks determine the most efficient order fulfillment strategies in real time.
Although it may seem counter- intuitive, most organizations will
benefit by starting with a process that’s understood and yet will improve with automation. If the organization has never solved a particular problem before, it probably doesn’t have the data needed to train AI on it, and the organization would lack the ability to assess whether the AI solution is doing a good job. “Don’t start with the big hairy thing in the distance,” Winkenbach says. Given that AI solutions interpret data, the quality of that data is key, Theide says. Say a retailer is trying to analyze consumers’ reactions to price changes and promotion information in order to optimize its demand forecast. If the company hasn’t been tracking the impact of price changes or promotional information, it can’t expect the model to provide solid predictions of consumers’ future actions.
30 Inbound Logistics • February 2025
Powered by FlippingBook