The age of self-driving cars and trucks is nearly upon us. What about self-driving supply chains?
In recent decades, technology has insinuated itself into every aspect of supply-chain management. Up to now, however, the presence of a human to make key decisions on production and distribution has been required.
That’s especially the case when it comes to risk management, a discipline that has become essential to the survival of modern-day global organizations. Supply-chain managers must weigh the probability of any number of potential disruptions occurring, whether natural disasters like floods, tsunamis and volcanoes, or human-caused events like labor actions and terrorist attacks, and take appropriate action. Such a determination has been deemed to be beyond the capability of machines.
Until now. Artificial intelligence, with its ability to sift through massive amounts of data and detect patterns that are invisible to the human eye, is playing an ever-increasing role in forecasting and risk management. The day might soon come when machines have better insight into the future than the humans who built them.
The most promising innovations in A.I. today are taking place in pharmaceuticals and consumer packaged goods, according to Fred Laluyaux, chief executive officer of Aera Technology, Inc. They’re under the most pressure from Amazon to radically transform the way in which they serve customers.
“A client described it to me in an eloquent way,” recalls Lalulaux. “They said, ‘We are the next ones to go if we don’t change.’”
For traditional merchandisers, the challenge lies in protecting their brands at a time when consumers are more interested in price, convenience and speed. Thanks to Amazon, Laluyaux says, going to a store is becoming “irrelevant.” To match the capabilities of the e-commerce behemoth, sellers must boost efficiencies at every stage of their supply chains. And that includes doing a better job of managing risk.
Much of the advances in A.I. are occurring in the area of predictive analytics. Manufacturers are forced to speed up the process of getting product to buyers. At the same time, the availability of “big data,” while presenting a far broader picture of the market, is inundating planners with information. Without the help of automated systems to interpret that input, they can’t distinguish “signal” from “noise.”
Companies can no longer get by on yearly promotion plans and six-month sales and operations planning (S&OP) cycles. Conditions in the marketplace — in particular, the tastes of fickle consumers — change too quickly for that. “You have to get to the next level of performance,” says Laluyaux. “That [degree of] automation gets you to where humans can’t follow.”
The ultimate if elusive goal is a system that responds to market conditions in real time. Demand sensing is hardly a new discipline, but it has long been hampered by legacy tools and processes that delay necessary action. In its present form, the capability extends well beyond accessing sales figures to include such elements as “smart” packaging.
Inputs consist of everything from point-of-sale (POS) data to weather forecasts, Nielsen ratings, social media posts and competitive intelligence. In the pharma world, add in updates on government approval of new drugs.
“The net is getting wider and wider,” says Laluyaux, “and the grain more and more fine.”
As a standalone system, A.I. is of little value. To be effective, it has to combine disparate streams of data and spread itself across multiple functions of the supply chain. Say a retailer is experiencing a 2-percent bump in sales for a particular item. To meet the unexpected demand, it needs to be able to identify the optimal source for ramping up production, assess manufacturing capabilities, revise the bill of materials and adjust inventory levels accordingly. If transaction systems aren’t “speaking” to each other, that process can take days — too late for the merchandiser to take advantage of a fleeting trend.
What’s more, the A.I. system needs to blend its trove of data with human experts, who are still responsible for making final decisions about when and where product should be shipped.
Lalulaux describes the concept of the “cognitive workbench,” whereby A.I. interprets the data and makes recommendations, which are then (for the most part) executed by human experts. But that’s just a transitional phase in the advancement of A.I. The nature of machine learning is that the system gets better with experience at coming up with the appropriate actions to be taken. Ultimately, it should be able to make many of those key decisions without the need for human intervention. At that point, “predictive” analytics becomes “prescriptive.”
We’re far from the point where automation takes over entirely from human managers. If the progression runs from predictive to prescriptive to fully autonomous, then many companies remain stuck between stages one and two, Laluyaux says. The prospects for a “self-driving” supply chain are reasonably good; it’s just the timeline that’s in question. For now, it’s too early for humans to think about taking their hands off the wheel.