A new report from the World Economic Forum makes a case that sounds obvious until you sit with it: giving AI more control over supply chains does not automatically make them safer. It can, if companies do the unglamorous work first. If they do not, it can make things worse, faster.
The shift being described is simple to understand. Older AI tools in supply chains were mostly lookouts. They flagged a late shipment or a supplier problem, and a person decided what to do about it. Newer systems, often called agentic AI, are built to act on their own within set limits: finding a backup supplier, rebooking freight, or adjusting a production schedule, often in minutes rather than days.
Money is following this shift quickly. Gartner expects spending on supply chain software with these self-acting features to grow from under 2 billion dollars this year to 53 billion dollars by 2030. Adoption is expected to jump from just 5 percent of companies using such software today to 60 percent in the same period. McKinsey has separately estimated that this kind of automation could cut the cost of goods sold by up to 7 percent and shrink decision times from weeks to hours.
This is not theoretical. Walmart already runs automated systems across inventory management and supplier negotiation, and has managed to grow sales faster than its inventory, a sign the systems are helping rather than just adding cost. The freight company C.H. Robinson runs dozens of these systems that handled millions of shipment tasks last year, with a documented productivity gain.
The catch is that speed without discipline is dangerous. If a system is working off messy or incomplete data, or if several companies use similar automated logic at the same time, they can all rush to book the same backup supplier or the same truck capacity, making a shortage worse instead of better. Security researchers have already documented cases where one faulty automated system fed bad information downstream, and the error compounded before anyone caught it.
The report's real argument is that the winners in this shift will not be the companies with the newest software. They will be the ones who cleaned up their data, connected their separate departments so decisions do not get stuck at handoffs, and built clear rules for when a machine can act alone versus when a person must sign off.
That last part matters more than it sounds. A system that can act in minutes needs humans to have already decided, in advance, what it is and is not allowed to do without asking first. Companies that skip this step are not getting a faster supply chain. They are getting a supply chain that fails faster.
For any business that depends on suppliers, freight, or inventory, ignoring this trend is not really an option anymore. Competitors adopting it well will respond to disruptions in hours instead of weeks. But rushing in without fixing the basics first is how a company turns a minor supply hiccup into a full-blown shortage, just more quickly than before.