The current state of artificial intelligence in supply chain management is that it’s mostly just a box to check, says Jeff Alpern, vice president of product at Noodle.ai.
“A lot of people put window dressing on what they already have and call it AI or ML [machine learning],” Alpern says. And that’s a shame, because the probabilistic planning and generative techniques on offer are the future of supply chain planning. “It’s simply a matter of time.”
Alpern says supply chains are currently experiencing an “innovators’ dilemma,” where fundamentally new technologies have come on the market but, unlike prior technologies such as demand planning systems, they’re actually very difficult to figure out how to use.
One of the challenges standing in the way of implementation is that AI requires a “new math” approach, and supply chain managers are in the habit of dumping a whole new model on planners without giving them the tools they need to use those models.
Alpern says another problem is a tendency to take a “technology-first” approach, where one encounters a new technology such as generative AI or ChatGPT and tries to figure out how to use it. That’s the wrong way to go — the real challenge is framing a problem, modeling what “good” looks like, then using AI tools to solve the problem.
When they’re doing their job, though, AI’s algorithms can offer useful shortcuts to playing out demand and inventory planning dilemmas, such as whether avoiding a stock-out will cost 10 times more than keeping excess inventory, or vice versa. “Is there a role for ChatGPT?” asks Alpern. “Maybe. But you’re going to have to train it on supply chain information.”
