Who’s really at the wheel for Uber and Lyft? In many ways, AI.

Those drivers considered high value may earn additional incentives to drive more, while others may get fewer opportunities. But neither the customers nor the drivers fully realize how or why they’re treated differently.

Weil’s findings offer a glimpse into the power of AI and its potential to reshape the relations between companies, workers, and consumers. AI’s ability to almost instantly sift through mountains of data collected from apps, online activities, and point-of-sale systems could give companies great advantages. It would allow them to customize prices and pay according to individual circumstances and would mean that each decision consumers and workers make could eventually factor into how much they pay or get paid.

“One of the things that is troubling about the greater and greater use of very, very nuanced data harnessed to AI and machine learning is that you can figure out ways, essentially, to coax out people with exactly the amount you need to pay them to get them to work,” said Weil, the former head of the Department of Labor’s wage and hour division. “And not a penny more.”

The insights into how Uber and Lyft operate grew from materials turned over to state during the discovery phase of a lawsuit against the companies for misclassifying drivers as independent contractors rather than employees. The case was settled the day before closing arguments, awarding drivers new benefits, wages starting at $32.50 an hour, and back pay, but didn’t resolve drivers’ employment status (employee or contractor?) or address the transparency of the companies’ algorithms.

Weil, as the state’s lead expert, conducted 170 hours of research into how the companies and their algorithms operate. What he found was astonishingly powerful and efficient technology that seamlessly matches workers to work, sets routes, and determines prices and pay according to real-time circumstances and data from millions of trips.

Uber and Lyft rejected Weil’s assertions about rider fares and driver earnings. At Uber, the calculations are based on routes: origin, destination, day and time, and supply and demand, a spokeswoman said. The characteristics or behaviors of riders and drivers aren’t factored in.

Fares and wages will still be calculated the same way under the new $32.50 hourly pay rate (which accounts for drivers’ expenses such as gas and car maintenance), but earnings will be analyzed every two weeks and adjusted if needed to meet the new minimum.

Promotions, such as extra pay for trips that begin in certain times and areas, are offered to certain drivers, and past performance can determine eligibility, according to Uber.

Certainly, these algorithms have created social benefits. Uber and Lyft are providing transportation to riders who might not otherwise have it. Research has found that these types of platforms increase employment for low-skilled workers, and result in fewer crimes and fewer drunk-driving fatalities.

But such algorithms give companies even greater ability to squeeze more out of workers and passengers, at the lowest possible cost — and increase profits, labor experts say. The efficiency with which these platforms operate could encourage companies to hire more independent contractors, who aren’t protected by minimum wage laws, unemployment insurance, workers’ compensation, or other safety nets. Research shows that avoiding these responsibilities can cut labor costs by roughly 30 percent.

The increased use of contractors could also put downward pressure on the wages of regular employees, Weil said. And because drivers’ pay can be individualized — referred to as “algorithmic wage discrimination” — it’s also more difficult for them to understand how they are paid and whether they need to push back against what they feel are inequities.

Leonel De Andrade, 61, of Brockton, drives full time for Uber and Lyft, sometimes working 12 hours a day, seven days a week to get by. He’s well aware that the platform controls every aspect of his job in ways that he doesn’t fully understand, and it’s not just about trips and wages. If he gets into an accident or has other issues, it can be difficult to find a human to talk to, he said. “The algorithm is everything.”

Low-wage workers in these situations have the most to lose, said Daniel Susskind, a London economics professor and senior research associate at the Institute for Ethics in AI at Oxford University, because they are more likely in need and willing to accept less than others.

“There’s something particularly unjust about these systems being used to do this sort of wage discrimination in those already insecure parts of the labor market,” he said.

In any industry where contract work is possible — from health care to retail to law — algorithms can infiltrate. AI could determine the lowest wage contract nurses might accept, for example, based on the willingness of similarly situated nurses to take shifts.

In food service, AI platforms could bring in contract workers just for the dinner rush, saving the cost of scheduling more regular employees for a full shift. The hiring algorithm might start by listing a low wage and gradually raise it by $1 an hour until hitting the magic number workers will accept.

“It’s wage limbo,” Alexandrea Ravenelle, a sociology professor at the University of North Carolina who studies the gig economy. “How low can you go and still get workers?”

Technology has improved jobs in many ways, of course — allowing remote work, increasing flexibility, and automating mundane tasks — but Susskind and others warn there’s also a danger it could make them worse.

Part of the problem is labor laws haven’t caught up to the 21st century and the technology that allows companies to monitor almost every aspect of work, analyze vast troves of data, and use that information to assign work and determine pay. It puts workers, who don’t have access to this information, at great disadvantage, labor experts said.

Economists call this imbalance “information asymmetry.”

“Information asymmetry creates power for the party that has the information,” said Marshall Van Alstyne, an information systems professor at Boston University who recently cochaired a conference on platform strategy. “And the algorithms are doing exactly that for Uber and Lyft.”

One possible way to help reset the balance would be to share data with drivers, Van Alstyne said, allowing them to make better decisions about when and where to work.

This kind of flexibility provided by app-based jobs is a huge selling point for workers. But Uber and Lyft drivers who rely on the platforms to make a living are likely compelled by algorithmic nudges offering more money to work at high-demand times and locations, said Lindsey Cameron, a University of Pennsylvania management professor who testified for the state.

Many drivers may feel they have no choice but to take the jobs, even if they’re not convenient, she said.

“This autonomy feels real to people,” said Cameron, who worked part time as an Uber driver for three years. But it’s more like “your mother saying you can have broccoli or carrots, but either way you’re eating vegetables.”

Diti Kohli of the Globe staff contributed to this report.

Katie Johnston can be reached at [email protected]. Follow her @ktkjohnston.