They were Lyft drivers. Now they’re cleaning Waymos.

Former Lyft drivers are ceding the roads to robots, and helping the machines get five-star ratings.

In Nashville, Lyft’s Flexdrive subsidiary plans to open its largest vehicle cleaning and maintenance depot yet in October — to service Waymo robotaxis. The 80,000-square-foot facility will employ about 70 full-time workers, about half of whom formerly drove for Lyft. The situation offers an early glimpse at what AI may mean for millions of workers: eliminating some jobs while creating new work around the machines taking over.

Jonathan Baines drove for Lyft for about a decade to help support himself while pursuing a career as a pop and soul singer-songwriter. Now, as a fleet operations lead at Waymo’s existing Nashville depot — which Lyft took over in June — he helps direct robotaxi traffic through the facility and determine whether the cars are ready to return to passengers.

“The general idea is to keep cars rolling through,” Baines says.

For him, the best part of the job is that it gives him experience in an industry he expects to be part of the future. “It’s new. It’s innovative,” Baines says. “This is going to give me opportunities within the AV industry, here within Lyft and Flexdrive, that I wouldn’t otherwise have.”

Similarly, Chauncey Thompson, who played college basketball and later spent about a decade working in the rental-car industry before driving for Lyft, says he strongly prefers being a fleet operations associate to being a driver. He cites the opportunity to work with new technology, consistent hours, and the camaraderie of working on a team. (Lyft arranged the interviews with Thompson and Baines.)

Flexdrive’s internal tech tracks diagnostics and maintenance, alerting the company when cars need service, but workers are still responsible for visually inspecting the vehicles, cleaning and charging them, and flagging anything that needs extra attention. After a storm, Baines might notice mud on a car’s rocker panels or an issue with a window’s calibration and send it for further inspection by a technician.

Thompson inspects each car, charges it, wipes it down, vacuums it, and checks tire pressure. For now, deciding whether a car is clean enough to return to passengers still relies on human judgment.

Thompson’s background also made the transition feel natural. While driving for Lyft, he used his own car, which meant paying close attention to how it looked and whether it was ready for customers. “I wanted my car to look nice on the road and my tires to be in perfect condition before I hit the road,” he says.

For Lyft, which sold its division that built self-driving cars in 2021, this work offers another way to participate in autonomous transportation. Its Flexdrive unit, which cleans and services cars that Lyft drivers rent from the company, is now trying its hand at cleaning and servicing vehicles without drivers. Riders will be able to hail Waymos through Lyft’s app, in addition to Waymo’s, later this year. That puts Lyft on both sides of the operation: providing the platform through which passengers request rides, and the human and physical infrastructure that keeps the vehicles clean and running.

Flexdrive CEO John Parks says autonomous-fleet management is more labor-intensive than the company’s traditional rental business because work once handled by drivers, who kept their cars for 14 to 16 weeks, now must be handled by Lyft daily.

“Autonomous vehicles can’t clean themselves, they can’t charge themselves and they can’t repair themselves,” Parks says. “Now all of those come back to our depot.”

He didn’t share specifics on the relative servicing time, but noted that robotaxis, which aren’t constrained by a human driver’s need to sleep or take time off, could travel as many as 100,000 miles a year, compared with roughly 40,000 miles for a car in Flexdrive’s rental fleet. More mileage means more charging, cleaning, tire replacements, and repairs.

“The big difference is the frequency with which you have to touch the cars,” he says, noting that depot workers are charging and cleaning robotaxis about two to three times per day, rather than every few months, when Lyft drivers return their rental vehicles.

What they lost in passing interactions with passengers, they say, they gained in sustained relationships with coworkers.

Flexdrive manages about 15,000 traditional rental vehicles across 24 markets with only about 100 employees, Parks says. Its Nashville Waymo depot, built to accommodate a growing fleet, will employ 70 people. About 50 Waymos are currently operating in the city, though that number will grow gradually over time.

Some of that work is routine. Other problems require human judgment. A broken charger may need troubleshooting; a badly positioned car can disrupt the depot’s flow. Even when the vehicles move themselves around parts of the facility, Parks says workers still need to notice when something is wrong and figure out what to do next.

The new structure has also changed the social experience for the former drivers. Driving put Baines and Thompson in contact with a changing cast of passengers and required them to read people and situations in real time. Depot work is more hidden and repetitive, but both men find it more compelling.

“I miss driving because I love interacting with people,” Baines says. “But in terms of the actual work itself, I love this more.”

Thompson says he is “definitely less visible” now, though both men still occasionally drive for Lyft. What they lost in passing interactions, they say, they gained in sustained relationships with coworkers, and the potential for more career opportunities.

“These jobs where blue-collar workers get to work on something that is more tied to AI, it’s not a surprise to me at all that those workers feel much more secure about their career prospects going forward,” says Daniel Zhao, the chief economist at Glassdoor. People, he says, “would rather be working on a product that they think is where the future is headed.”

The Nashville depot is one small example of a broader phenomenon economists and labor researchers are trying to measure: AI doesn’t simply eliminate jobs. It can also break them apart, automate some tasks and shift others to new workers.

Alexander Alonso, chief knowledge officer at the Society for Human Resource Management, says the shift from driving to depot work illustrates a more useful way to think about AI and employment. He calls it the “redesign of work.”

“The big metric is not only jobs lost,” Alonso says. “It is: How are jobs changing?”

SHRM’s research suggests that change is already underway: 42% of workers expect AI to significantly change their roles. Meanwhile, 35% of the skills in current job postings did not appear in postings for the same jobs a year and a half earlier.

But Alonso isn’t yet seeing massive growth in AI-related jobs to offset those potentially lost to AI. “We’re not seeing a one-for-one replacement,” he says.

SHRM estimates that 22 million U.S. jobs have the technical potential to be displaced by AI and automation by 2030, but after accounting for consumer preferences, regulations, and jobs that combine automatable and nonautomatable tasks, it considers about 8 million realistically at risk.

“The technology might automate one responsibility, but if the organization still needs the person for the other tasks, it doesn’t actually create savings,” Alonso says.

Waymo, for its part, argues that autonomous vehicles can create new jobs without eliminating driving as a viable source of income. The company would not disclose how many people it takes to clean, charge, and service its national fleet of more than 3,500 vehicles.

Robotaxis may remove the driver from behind the wheel, but they have not removed people from the work of providing a ride. As it stands, some of that work has moved out of the driver’s seat and, increasingly, out of the passenger’s view.

Rani Molla is a technology journalist with nearly 15 years of experience covering Silicon Valley, AI, labor, and the companies reshaping society.