the doctor has recently seen a few pieces out there on the forthcoming autonomous supply chain and even a few pieces on the “self-driving” supply chain. Eek! Just like we’re not ready for AI-enabled self-driving cars that will drive us off a cliff as soon as they become self-aware (and that’s why you can have Carl and I’ll stick with Alfred), we’re not ready for self-driving supply chains that “predict” future demand, automatically order large numbers of products for you, and push them to local warehouses and retail stores without any human intervention.

Just because your demand sensing engine, which works well for established products, can use PoS data and other demand signals to auto-reorder staples with 98% accuracy doesn’t mean it can predict the success of an upcoming, or relatively new, product line — especially if it’s new for your business and you are, unbeknownst to your sensing engine, about to be beat to market by your nearest competitor — and it’s in the consumer electronics industry where first to market typically captures 10% to 30% just for being first. The last thing you want is for the platform to increase your initial order by 30%, ship straight to store, and then have it sit there for six months, and depreciate. Not a good use of cash.

Nor do you want your TMS automatically assigning shipments to carriers, intermediate warehouses, and ports without any human intervention. Software with limited data feeds often have no forewarning when a port might shutdown due to a strike, but humans might. Nor will a limited feed software algorithm know when a border might close and also close a supply route. But a human might. And so on.

And, despite what Amazon may think, you definitely don’t want to be thinking about anticipatory shipping. As we noted in our post five years ago, while predictive analytics is getting better by the day, it’s still hit and miss at a granular level. And individual purchases are quite granular. Just because 4 out of every 5 people who buy a Buzz Lightyear Cup also buy a Sheriff Woody Saucer, doesn’t mean the 4 people that your “AI” chooses will. One of them might not like saucers. Or Sheriff Woody. Shipping on an anticipatory nature guarantees that at least one out of every 20 units will be unwanted, and as many as one out of every 4. That’s a lot of returns. And, as we have noted again and again, including our recent article on how there is no free lunch, and there is no free shipping either, that could get quite pricey. How are you supposed to keep costs down if you have to budget for amortized high, wasted, return shipping costs across every unit?

So please, please, don’t try to give your supply chain autonomy. Automate it. Apply the best assisted intelligence solutions on the market to provide one-click recommendations, but always make sure there is a human check before any decisions are made that affect millions of dollars.