For many business leaders, success in introducing artificial intelligence is measured by how many jobs they can cut. In the last few weeks alone, companies have announced tens of thousands of layoffs that they blamed on AI, a wave that one global bank chief undiplomatically described as the replacement of “low-value human capital” with technology.
However, such views reflect “a very narrow understanding” of AI’s potential, said Erik Brynjolfsson, who leads the Digital Economy Lab at Stanford University. “Many people mistakenly assume that the only way to increase productivity through AI is to reduce labor costs.”
Mr. Brynjolfsson is among a group of economists who argue that companies can reap greater profits by using artificial intelligence to make workers more productive rather than replacing them.
Schneider Electric, a global energy technology company based in France, has taken this message to heart. Schneider, which employs nearly 160,000 people worldwide, relies on artificial intelligence throughout the company.
First, it identified “where our employees are either wasting time on repetitive tasks, doing tedious tasks, or doing things that are fundamentally not right,” said Philippe Rambach, the company’s chief artificial intelligence officer.
In other words, the work that gets in the way of the work.
Use of AI in call centers
For Schneider, call centers were a clear case. Anyone who has ever navigated the labyrinth of automated telephone assistance will groan at the prospect of further technical adjustments. But Mr. Rambach said the goal is to use technology to get answers to customers faster.
Before the company began using AI, customer service agents received thousands of questions from callers and conducted a large search through millions of pages of information to find the answer, Mr. Rambach said. “Guess what?” he said. “Our customers were not particularly satisfied with the quality of the response or the speed of the response.”
Now the AI takes over the search and details how the information was selected and where it came from. The agent then reviews the answer with the caller and modifies and refines it if necessary.
In the last three months of 2025, call centers answered 150,000 questions. In three-quarters of cases, AI was able to provide the correct answer to simple questions such as: “Why does a newly connected energy monitor not show consumption?” In these cases, the agents used the response generated by the AI. The rest of the time, agents worked with callers on more complex issues, including helping building managers determine the root cause of energy alerts.
Response times were quicker and employees were much happier, Mr. Rambach said, because the time saved searching through databases to answer common questions gave them more time to work with customers.
Other companies saw productivity increases for their employees. A study conducted by Mr. Brynjolfsson along with two other researchers, Lindsey Raymond and Danielle Li, A study of more than 5,000 customer service agents at a Fortune 500 company found that AI assistance enabled agents to resolve 15 percent more problems on average, with less experienced and less skilled agents achieving the most improvements in speed and quality.
At the same time, they found that callers were more polite and less likely to utter the phrase that every customer service representative now dreads: “I want to speak to a manager.”
AI on the factory floor
At a modernized factory in Le Vaudreuil, about 60 miles north of Paris in Normandy, Schneider is using artificial intelligence to manage complex industrial processes at a decades-old site that has already been modernized with robots and digital tools – some with French accents.
The automated guided vehicles, or AGVs, that glide around the factory floor delivering parts, for example, are named after great French writers like Zola and Hugo Émile and Victor.
Artificial intelligence is not necessary everywhere, said Virginie Rigaudeau, project manager at Schneider. “We only use AI if we know that it delivers added value.”
As with the production of the 74 million silver tips the factory produces each year to make electrical contactors – the switches used to turn circuits on and off in elevators, motors, electric vehicles, heating systems, lighting systems and more.
The recipe for cooking the tops contains silver nitrate and sodium. The mixture is whirled in a centrifuge and the resulting silver paste is then washed repeatedly in large steel-gray tanks to flush out excess sodium.
But knowing how many wash cycles were enough was always something of a guessing game, Ms. Rigaudeau said.
With AI, operators can see a visual representation and learn the exact amount of sodium remaining after each flush cycle.
“Now the system tells us when to stop washing and we know immediately whether the powder meets quality standards,” said an employee.
The savings have been enormous, Ms. Rigaudeau said. Within a year, the company reduced waste from the process by 73 percent and water consumption was drastically reduced.
Samples from each batch no longer need to be sent to an outside laboratory for testing, a process that can take between 24 and 48 hours. This has saved thousands of euros in laboratory costs while reducing gasoline consumption – from trucks that used to transport samples to and from the laboratory – by 22 percent.
Cameras enriched with artificial intelligence are also used to assess the quality of finished shooters within seconds.
Make employees better, not lay them off
In some European countries, the use of artificial intelligence to increase – rather than replace – labor productivity is encouraged by strict labor laws that can make firing employees difficult and expensive.
In the United States, Mr. Brynolfsson said at Stanford, government policy often encourages companies to invest in capital and cut labor. He referred to the tax code.
“If you start a new company and have a lot of labor, you have to pay more taxes,” Brynolfsson said. “Those who only invest in capital pay less taxes.”
Of course, predictions about the impact of artificial intelligence on the labor market encompass a multiverse of possibilities. And while many economists agree that policymakers and companies have a choice in how AI is used, some question whether those options are becoming increasingly narrow.
It’s “very unpredictable,” said Anton Korinek, who helped lead the Economics of Transformative AI project at the University of Virginia. Artificial intelligence “will create and destroy jobs, and it is not clear which will predominate,” he said.
Mr. Korinek said he began looking into developing AI as a tool to support the workforce more than 15 years ago, but the spectacular advances have made him even more doubtful about society’s ability to manage the development and use of that AI. “You can no longer easily choose the direction in which things go,” he said.
At some point, he said, AI will be “much more productive and cheaper than humans.” (Mr. Korinek, who recently joined the economic research team at the Anthropic Institute, the AI company’s research arm, made his comments before taking up his new position.)
The built-in dilemma is evident even at Schneider, a company that has found ways to use AI to supplement the work of its employees.
Sandra Ferraguti, general manager of the Le Vaudreuil factory, unveiled a new plug-and-play contactor developed by Schneider’s AI-powered workforce that no longer requires an electrician to wire.
“Now a robot can install it,” Mr. Ferraguti said. “You don’t need a human.”
-
Patricia Cohen
Business reporter
The news about artificial intelligence-related layoffs has been pretty discouraging lately. However, several economists I spoke to emphasized that we have a choice in how AI is used and developed; that it does not need to be configured to replace workers. As Erik Brynjolfsson, the Stanford economist, put it: People are wondering what AI will do to the workforce. “I think it’s kind of backwards,” he said. “What do we want to do with AI?” I asked for some examples and Schneider Electric was mentioned as a company that took an approach aimed at empowering its employees with AI rather than replacing them. An interesting question is whether companies and policymakers are doing enough to manage AI development in a way that can empower workers. What do you think?