AI agents are now running supply chains without approval

Large manufacturers are handing daily supply chain decisions to AI agents (software that acts on its own within set limits) that execute orders, reroute freight and rebalance stock without waiting for human approval.

What happened

According to artificialintelligence-news.com, multi-agent systems (multiple AI agents working together, each handling a different task) are replacing the approval stage in logistics operations.

Instead of predictive models that show recommendations and wait for a person to act, independent software agents read live data from carrier arrival times, warehouse cameras and stock management systems. They then execute freight rerouting, stock rebalancing and dock allocations directly inside the company's main business software.

Lenovo reported this shift across its global supply network, which covers 180 markets, more than 30 factories and 100 logistics centres. The hardware manufacturer connected an Order Fulfilment Agent and a Risk Management Agent directly to its transaction systems. Lenovo said fulfilment decisions ran three times faster, disruption response four times faster, risk assessment operated at 85 percent accuracy, and delivery accuracy increased 30 percent.

A mid-size automotive parts manufacturer, documented by Simor Consulting, deployed five specialised agents across 15 countries and 200 suppliers over 18 months. On-time delivery rose from 82 percent to 94 percent. The disruption agent detected supply threats 48 hours ahead of manual monitoring teams.

Fujitsu and Rohto Pharmaceutical ran an initial trial on a virtual network that produced transport cost reductions of up to 30 percent. Both companies scheduled an expanded trial on Rohto's live physical supply chain between January 2026 and March 2027.

Companies are also building supervisor agents that coordinate multiple functions at once. Kohler deployed a supervisor agent coordinating demand, inventory and planning. Belden built a multi-tier supplier network with task agents responding to transport incidents.

The article notes that supervised autonomy requires strict boundaries. Transport rerouting scripts hold authority only within cost ceilings and service level agreement limits. Inventory adjustments that exceed predefined financial values or volume percentages pause automatically for manual approval. Supplier-facing communication agents remain restricted to draft modes on unvetted supplier accounts until interaction accuracy passes set benchmarks.

What it means for a small business

This technology is not just for global manufacturers. The same thinking applies at any scale. Say you run a small online shop. Right now, when stock runs low, someone checks the system, emails the supplier, waits for a reply, then places the order. An agent could monitor stock levels, send the order request directly to your supplier's system, and confirm delivery, all without you opening an email.

A retailer I work with automated alerts that tell them when stock is running low, using a workflow tool. The next step is letting the system place the reorder itself, within a spending limit they set.

Three things you could try this month

First, map one repeating decision in your business where the same rule applies every time. That is where an agent could act instead of waiting for you.

Second, if you already use predictive tools or dashboards, ask whether the software can execute the action it recommends, not just display it. Some tools already offer this.

Third, set a hard limit before you let any system act on its own. A spending cap, a volume ceiling, or a rule that says certain suppliers need manual approval. The limit protects you while the agent learns.

What could your business automate?