If we want a world without waste, we might need something other than the human brain to achieve it.

The answer, as with so many other aspects of business today, lies in artificial intelligence — in this case, its ability to eliminate industrial waste in manufacturing.

Stephen Pratt is chief executive officer of Noodle.ai, a company employing AI to streamline industrial flow operations. He views the fledgling technology as a tool for “getting back to the way factories and supply chains are supposed to operate.” And that means achieving “a seamless flow from raw materials to the shelf.”

Of course, the elimination of waste in the factory and beyond has been an obsession of supply-chain executives for decades. The famed Toyota Production System, which identifies seven forms of muda, the Japanese word for waste, traces its origins to the late 1940s. More recently, Lean and just-in-time (JIT) theories for suppling parts to factories and finished goods to warehouses stripped away what managers considered to be excess inventory. (So much so, unfortunately, that many retailers were left without enough product to satisfy surging demand for consumer goods during the COVID-19 pandemic. One person’s “lean” supply chain is another person’s stockout.)

One has to ask whether previous theories for washing waste out of the system have yielded more books and academic papers than real results in the factory. According to Pratt, the World Bank estimates that global industrial waste today is 18 times larger than municipal solid waste — “the things we call trash.”

To be fair, Lean, JIT, Six Sigma methodology the Theory of Constraints and other advanced-planning tools have all made their mark on management practices in the factory and beyond. Production waste has been drastically reduced, but remains a $2-trillion problem today. (Pratt breaks that number down below.) “There’s nowhere near a perfect state of flow,” he says. “A lot of supply chains have fixed business rules, but ask any inventory planner, and they’ll tell you that the only thing they know is that those rules are wrong.”

Along comes AI with a fresh approach to waste control. The difference, says Pratt, is the use of complex algorithms to predict when excess parts, products and practices are threatening to clog up the works. Such alerts allow humans to take action to head off the problem before it affects the flow of product.

The use of AI specifically to attack production waste is less than a decade old, Pratt says, but has already proved to be “incredibly effective.” Previous manufacturing applications, including material requirements planning (MRP) and enterprise resource planning (ERP), were hampered by slow computers and expensive data.

“The bane of existence of those technologies is that they assumed averages for production yield and delivery times,” Pratt explains. “That’s like talking about the average shoe size of a person on the planet. The difference between average and reality is about half the waste in all of the world.”

Toyota’s seven forms of muda, as identified by the Japanese industrial engineer Taiichi Ohno, are transport, inventory, motion, waiting, overproduction, overprocessing and defects. With AI in the picture, Pratt extends the concept beyond the factory to delineate what he views as the four biggest areas of waste in the supply chain today:

• Defective products. A faulty item that makes it all the way through the production process has to be thrown out. The cost to manufacturers, Pratt says, is around $861 billion a year.
• Factory breakdowns. Unplanned downtime is “a giant problem,” costing $689 billion a year.
• Errors in distribution. The factory turns out a flawless product, which stalls at the distribution stage because current I.T. systems can’t meet delivery targets for time, place and quantity. The cost of excess inventory: around $446 billion a year.
• Empty shelves. The absence of something might not seem like a form of waste. But retailers’ fear of stockouts when customers want to buy — which happens about 9% of the time — causes them to flood the supply chain with inventory. As a result, Pratt says, “they produce more than they have to.”

For the first time, ever-growing computing capability makes it possible to attack all four of those areas effectively. The average computer used for such calculations today is 2,000 times faster than the world’s fastest supercomputer in the year 2000, Pratt notes.

The role of AI is moving rapidly from descriptive to prescriptive analytics. An AI-driven system today can scan for abnormalities, alert managers of their imminence, predict the consequences of such events, and recommend corrective action. In addition, says Pratt, “It tells you how to adjust your factory so you don’t produce defective products.”

None of this means that people are out of the picture entirely. Pratt views the ideal system as one that combines the brute-force computing power of AI with the creativity and insights of the human brain. And by any measure, AI still has a long ways to go before fulfilling its promise in manufacturing environments, let alone the larger world.

As with any cutting-edge technology, acceptance of AI will be gradual. But Pratt believes its maturity is being accelerated by demands for big retailers like Walmart for flawless performance by suppliers. Their ability to meet such criteria “could be the difference between existence and perishing,” he says.