If you’ve ever used an online patient portal to send messages to your doctor in the middle of the night, you won’t be surprised to learn that responding to those messages is an increasingly large part of doctors’ workdays.
That’s why in recent years, hospitals have begun adopting an artificial intelligence tool that can compose responses for them. The tool was supposed to make a time-consuming task done more quickly and smoothly, said Philip Barrison, a doctoral student at the University of Michigan Medical School who studies AI in healthcare.
Instead, the tool has provided doctors and nurses with a new to-do list. They first have to read the AI-generated response and decide if “it’s really something they think they would say,” Barrison said. Humans are suggestible, and looking at something and deciding whether you would have thought of it on your own is a cognitively complex task.
Even if the message seems correct, the doctor still needs to “edit it to the point where they think it’s acceptable” to send it to a patient, Barrison said. The AI tool introduces a whole new set of complicated decisions into what used to be a relatively simple process. As a result, many doctors have chosen not to use it at all.
They are lucky to have the choice. Driven by expectations of cost savings and skyrocketing productivity, companies are increasingly asking (and sometimes requiring) employees to use AI to make their work more efficient. Meta, for example, last year instructed some workers to use AI to “go 5 times faster by eliminating the frictions that hold us back.” Shopify’s CEO told employees they would have to prove they “can’t do what they want using AI” before the company would approve new hires. Some companies are even evaluating or classifying employees based on how much they use artificial intelligence tools.
Workers in some sectors have achieved significant time savings thanks to AI. But for others, the tools simply change the job instead of doing it faster. Workers may spend less time writing patient portal messages, for example, but more time editing communications that the AI tool writes.
At best, this mismatch between employer expectations and employee reality can be annoying. In other cases, however, it can result in workers being fired for failing to meet unrealistic efficiency demands. Some critics say that over-adoption of AI in high-risk environments, such as healthcare, even puts people’s lives at risk. Now, workers, unions, and experts are increasingly calling for guardrails to be put in place to protect employees from inflated expectations around AI, and customers, students, patients, and the public from the mistakes that can occur when managers put AI adoption above all else.
The hidden costs of using AI
Corporations are increasingly presenting their employees with a choice: Use AI to be more productive or “they will be automatically out of a job,” said Aiha Nguyen, director of the futures of work program at research organization Data & Society.
But the effects of AI on productivity are not as simple as some CEOs claim. In a 2025 study, software developers believed AI made them faster, but they actually took 19 percent longer to complete tasks. (The researchers tried to repeat the experiment this year, but had trouble recruiting developers who would agree to work without AI.) And in a recent survey of 5,000 white-collar workers, 40 percent of rank-and-file employees said AI wasn’t saving them time at all.
Workers in fields highly exposed to AI point to hidden losses of time that come with the use of this technology. Julie, an art teacher, wrote in response to a Vox reader survey that her school administrators routinely suggest using AI for lesson planning, emails, and progress report comments. You’ve tried AI-generated lesson plans, but they don’t take into account the fact that children can perform an activity at different speeds.
“First, I check what the AI suggests, then I edit them. Why add a step that I can do on my own?”
— Julie, art teacher who wrote in response to a Vox reader survey.
“First, I check what the AI suggests and then I edit it,” he said. “Why add a step that I can accomplish on my own?”
For one employee at an East Coast communications agency, an internal AI tool was supposed to speed up the process of writing press releases and other documents about the pharmaceutical industry.
“I think the goal is to be able to plug and play this machine and be able to process a lot of materials much faster than we already do,” said the employee, who asked to remain anonymous for fear of career repercussions.
But when the employee tried to use it for basic research, he made too many mistakes. Double-checking your work erased any time savings. When the employee tried to use it to communicate with customers, his people-pleasing tendencies became a problem, as the tool put a “strange, happy spin” on even messages warning of bad news.
“Part of the reason we use human speed to change things is because there are so many nuances behind everything we do,” the employee told me. “AI just won’t be able to pick it up.”
It’s not just about AI making mistakes. With the advent of agent AI, workers are increasingly asked to edit and monitor the output of multiple AI tools, a new type of work that can have unexpected costs.
A recent study of 1,488 workers across industries, for example, found that excessive supervision of AI agents could lead to what the researchers called “AI brains,” a type of cognitive fatigue. “Participants described a ‘buzzing’ feeling or brain fog with difficulty concentrating, slower decision making, and headaches,” the researchers wrote in Harvard Business Review. Brain fry was also associated with a higher number of errors and a greater desire to leave work.
The researchers also found that while using one or two AI tools increased productivity, adding additional tools produced diminishing returns, and after four tools, productivity actually decreased.
What workers really want from AI
Despite these findings, companies continue to pressure employees to use AI and cite investment in AI as justification for layoffs, even as companies that try to link workforce reductions to AI adoption tend to struggle in the stock market.
Some workers and organizations, however, are beginning to fight back. National Nurses United, the nation’s largest nurses union, has criticized hospitals’ use of artificial intelligence tools to estimate staffing needs or recommend treatment protocols for patients.
There is no guarantee that these tools will take into account a patient’s individual profile, including underlying medical conditions, as human doctors do, Cathy Kennedy, the union’s president, told me. AI is “supposed to help us do our jobs more efficiently, but at the end of the day, it makes it even more cumbersome,” he said.
Hospitals should evaluate, with nurses at the table, whether AI tools really work as advertised, Kennedy said. “We have to stop, we have to go back and really see if this is really doing what it needs to do,” he said.
The same is true across industries, Barrison, the health care researcher, told me. “Organizations need to be prepared to say when, if you were looking for a return on investment, if you were looking for value in a technology, how do you define what that value is? And if there’s no longer value there, how do you turn it off?”
Some workers have found ways that AI actually helps them do their jobs, but not the ones management expected. Julie, the art teacher, likes to use Claude to learn more about subjects she is less familiar with, such as kiln-fired ceramics.
Meanwhile, researchers have found that AI can actually reduce employee burnout, if it is used to complete tasks that employees consider burdensome. “Everyone in every job has a list of things they put off,” said Julie Bedard, managing director and partner at Boston Consulting Group who led the artificial intelligence study. “Those are the places where, unsurprisingly, I get a lot of enthusiasm for trying out AI.”
But employers won’t find out what those onerous tasks are unless they listen to rank-and-file employees. “Standards and workers’ rights should remain at the center of all of this,” Nguyen said, “rather than focusing too much on AI.”