fb-pixelMA companies that aren't replacing workers with AI and ChatGPT Skip to main content
Top Places to Work

These companies are using AI to help their employees, not replace them

Tools like ChatGPT still make mistakes — sometimes big ones. Many companies on this year’s Top Places to Work list are approaching with care.

Phil Wheeler for The Boston Globe

Since ChatGPT was released a year ago, businesses have been rushing headlong to find ways to use the artificially intelligent app that can seem uncannily human. But ChatGPT still makes mistakes, lots of mistakes — the kinds of mistakes that can get freelance writers fired, lawyers sanctioned by courts, and university deans in hot water.

So among the winning companies of The Boston Globe’s Top Places to Work, AI adoption to assist workers is progressing with care. That’s in sharp contrast with what’s happening in the larger economy. Three out of four C-suite executives said their companies were using some form of AI, with the tech directly in use by more than half of all workers, according to a survey by UKG, a Massachusetts- and Florida-based HR software company. And almost 30 percent of professionals said they’ve already used ChatGPT or a similar program at work, a survey by workplace social network Fishbowl found.


Top Places to Work companies, however, appear to be taking it much slower. At software developer mabl in Boston, generative AI is in limited use within the company, helping suggest first-draft copy for marketing events and other noncritical written communication. The company is also trying out the technology in products that help customers better test their own software, mabl’s main business.

“We look to it more for inspiration, certainly not operationalizing it as one of our writers,” says cofounder Dan Belcher, who has worked at Microsoft and Google. “We’re focused on fundamental innovation that’s reliable.”

Companies that have rushed into AI, on the other hand, have run into problems. Nearly 85 percent of AI innovations at companies fail before being deployed, says Carnegie Mellon professor Jodi Forlizzi. And of those projects that make it into use, another 40 percent ultimately fail, too. “We have a real problem in the way we’re developing these things, with a hasty orientation around saving revenue,” Forlizzi says. “It’s not that simple.”


By contrast, the way to succeed with AI is to go slowly, testing and looking for unforeseen problems, Forlizzi says. “When there is a robust process, people are thinking about questions like, How is it designed? How is it procured? How are we deploying it?

Sun Life insurance company in Wellesley is definitely taking it slow. The company operates multiple call centers and theoretically could try to replace operators with AI-enabled assistants, a growing trend around the world. But customers calling Sun Life’s center to talk about life insurance policies frequently have sensitive or complicated questions. Sometimes, a person is calling after the policyholder has died, notes Lisa Miolo, senior vice president and head of in-force management.

“I think [AI] will never replace the human element, that human empathy, the human judgment,” Miolo says. “You need that human overlay to say, ‘That doesn’t look right,’ or ‘Yes, that’s right but I need to convey this message in a different way.’”

Instead of rushing in, Sun Life is testing using generative AI as a tool to improve efficiency. In an initial trial, the company fed transcripts of three months of customer calls into a generative AI program. The software summarized the topic of every call in a few words, enabling Sun Life to quickly tabulate customers’ top concerns. That will allow the company to make changes to its website and other customer-facing resources to clear up some questions before they result in a call to customer service.


At Celtic Angels Home Health Care in Weymouth, AI combined with virtual reality could be helpful for training nurses, says company owner and registered nurse Maria Burke. The company employs about 250 people who provide in-home care to patients with ailments ranging from cancer to Parkinson’s to Alzheimer’s.

Training caregivers to understand the experience of patients with dementia is challenging, and Burke has turned to technology to help.

Using a virtual reality app, the agency is giving its nurses more of a taste of what the patient is experiencing. In one scenario, the VR simulates the experience with the distorted visualizations and audio of the internal narrative a dementia sufferer might be experiencing. “There’s a big difference between [caregivers] actually experiencing what it feels like to have dementia versus just talking about and showing videos of patients,” Burke says.

Adding generative AI could make the simulations more interactive and improve training, Burke says. The tech could be added to a robot for an even more realistic training scenario.

“What we may be doing for our training is taking care of a robot,” she says. “I know that sounds crazy, but things are changing so much in the health care field.”

Explore the 2023 Top Places to Work (by company size) and more:

TO PARTICIPATE IN THE 2024 TOP PLACES TO WORK SURVEY: Visit bostonglobe.com/nominate

Aaron Pressman can be reached at aaron.pressman@globe.com. Follow him @ampressman.