Supply chains detect fast, act slow: How AI agents fix it
ainews ·
Supply chain disruption cost businesses roughly $184 billion in 2025 according to the J.S. Held Global Risk Report, yet most of that spending still buys faster detection rather than faster action, writes Ryan Daws. The past decade of supply chain AI—visibility platforms, control towers, risk scores, digital twins and exception dashboards—has succeeded at shrinking the time between an event and awareness of it, but not the gap between awareness and a commercial decision. Detection is described as a "solved-enough" problem: demand sensing, ETA prediction, supplier risk scoring, inventory optimisation and lane analytics work, and a vessel delay or second-tier fab outage can be flagged early. The remaining cost sits in the interval after the flag, where bounded, repeatable choices—expedite or wait, split an order, retender a lane, consolidate movements, swap ocean for air—still queue behind a human inbox. A 2026 Knosc survey of mid-market manufacturers and distributors found supply chain teams spend 28 percent of working time responding to disruptions, mostly investigating what happened, while Gartner found in 2025 that only 23 percent of supply chain organisations have a formal AI strategy, and FourKites with ABI Research reported only 27 percent allow AI autonomous action. The article argues the next model is "bounded action": pre-authorising a narrow class of moves, such as retendering a lane within an approved rate band or reallocating safety stock, that agents execute within policy, contract and spend limits. It lists three conditions for change—decisions written as policies rather than tribal knowledge, execution systems that accept machine-initiated transactions, and accountability that moves with the action—and concludes the competitive split will show up in cycle time from detection to commercial act.
ainews