Everyone's talking about "autonomous" agents. In warehouses today, what's real and what's hype? Answers from Vishal Minocha, vice president of product management with Infor.

Agentic artificial intelligence “is the talk of the town,” says Minocha, but it’s only now maturing to the point where agents can take over from humans the responsibility for making certain decisions in the warehouse.

AI is shifting the burden of work so that supervisors can transition from “reactive” duties to handling exception management — that 5% to 10% that still needs a human to make the decision.

A typical warehouse without modern AI will have configured a pick sequence according to static rules, Minocha says. AI, by contrast, can look at real-time data about such aspects as traffic flow and equipment type, and alter the operator’s task load accordingly.

Condition analysis is another feature of agentic AI, avoiding the prospect of sending too many human pickers to a crowded aisle all at once. The AI-driven system might direct some pickers to a less congested area, then have them return to the previous site when there’s more room to maneuver. “It is not static data,” says Minocha. “It works on real-time information.”

To reach this point of effectiveness, warehouse AI had to gain the trust of human management. Now, depending on the complexity of work involved, it’s acting autonomously in situations that entail relatively less risk in case of an error. Where there’s more at stake, “agents can propose, and the human can still make that decision.”

Warehouses that run on highly customized on-premise systems can’t take advantage of the integration offered by AI today, Minocha says. A facility needs to be able to learn from others in the network, and “that’s only possible if it’s built on modern architecture.”