AI-powered automation is helping retailers keep up with demand, while transforming warehouse work

In an open floor-plan office in Emeryville, California, tucked between Berkeley and Oakland, a mechanical arm in a glass case is hard at work around the clock, transferring packaged T-shirts from one bin to another on a conveyor belt, long after office workers have gone home.

This particular AI-powered robot is not in a warehouse: Robotic induction is one of the many solutions being tackled by Covariant—a robotics company whose founders include scientists from OpenAI—in its Bay Area lab, where the wave of AI-powered automation currently sweeping retail is on full display.

Warehouses are excellent training grounds for AI, because of the sheer number of products that pass through them, said Covariant co-founder and CEO Peter Chen.

“If you want to build a general AI brain for robots, there’s no better place to start,” Chen said. “Everything that we interact with flows through some warehouse somewhere.”

Robot-powered warehouses may sound a bit like the start of a sci-fi film. But mesmerizing rows of robotic arms aside, AI automation in parts of the retail supply chain known for repetitive, manual work comes with new challenges, changing not only the nature of the work itself, but how learning to work happens.

Sprinting toward intelligent automation

In a 2019 report on automation in retail, McKinsey & Co. argued that while around half of activities in retail can be automated, the change will be “less about job loss, and more about the evolution of jobs.”

The report says retailers must be prepared for “skilling and reskilling at scale,” which, McKinsey posits, is 10%–20% cheaper than replacing an employee. And in tackling that challenge, retail may find its role in where “America learns to work” shifting, the report concludes.

Chen likened the shift to the one from PC to mobile in its impact, but said reskilling usually happens smoothly. “It also introduces a new category of work, that is: How do you enable robots?” Chen said.

At a warehouse in Louisville, Kentucky, where e-commerce fulfillment company Radial installed 12 picking robots powered by the Covariant Brain, site director Chris Epley said his employees have seen their jobs become more technical.

“With any new technology, there’s always some reservation,” Epley said. “But what I saw was a quick transformation as the robots came in and people got accustomed to them.”

In Epley’s warehouse, which uses a pick-to-light (PTL) sortation, AI-powered robots replace PTL for certain order segments, like health and beauty, and those human pickers move into roles Epley calls “bot tenders.”

Meanwhile, in a ~1,700-square-foot warehouse in central Oslo, Norway, the founder of DTC beard care business Barbershop said the AI automation installed two years ago has completely changed the business.

Andreas Doppelmayr, who founded Barbershop out of his college dorm room, said the introduction of pay-per-pick, cube-based AI robots from Pio allowed him to maintain a location in the city (which gave them a leg up competing for labor), utilize the same small space even as the company grew, and upskill his warehouse workers.

“Most of the people who used to work in the warehouse now work in the office,” Doppelmayr said, citing one longtime warehouse worker who transitioned into a supplier-facing role focused on product sourcing.

A human-centered patchwork

But the case of a warehouse worker learning on the job to perform a procurement role likely isn’t the norm, said Matt Beane, assistant professor of technology management at the University of California Santa Barbara.

Beane’s research, which has included years of fieldwork inside warehouses, indicates that being upskilled to “bot tender” makes jobs less engaging, not more, he said.

And that could create a new, long-term challenge for retailers, who are already struggling with high turnover and labor shortages.

Retailers are coming to terms with the fact that AI automation won’t solve all warehouse labor problems, said Sankalp Arora, CEO of Gather AI, which makes AI-powered warehouse inventory drone software. “People thought that the magic bullet of AI would come in and solve everything,” he said, “but now we are coming back to reality of a patchwork.”

Beane said firms have an incentive to navigate that patchwork of human and AI and implement automation in a way that both boosts productivity and improves work: stiff competition for workers.

The retail world needs to find “more human-centered ways of handling automation that get good results and enhance the welfare of the humans involved,” Beane said.