The objects of our fascination with AI often come in the forms of driverless cars, robots, and drones–but AI in customer service is making a huge impact on consumers and companies across the globe. Whether they’re answering phones, emails, or web chats, “millions of people worldwide work in customer service,” said Andy Mauro the AI expert for Nuance Communications, one of the companies leading speech recognition. “People don’t think about the call center. But we think it’s sexy because we’re going to take something that people do today and train machines to do it in the future.”

“When we talk about the self-driving car disrupting the transportation industry,” said Mauro, who worked on the world’s first customer service-focused virtual assistant, “everyone gets it. They forget about the call centers. But it’s really fertile ground for AI research.”

Although digital channels to customer service have expanded, Mauro noted that phone calls to a call center have not gone down. “In many instances,” he said, “they’ve gone up. If you get a text about data usage, that information, which you didn’t used to have such easy access to, is now causing you to say ‘wait a second, I need to call the call center.'”

Mauro believes that AI in customer service is going to change the narrative around AI. “We’re going to say, ‘remember when we used to let people do menial, repetitive, horrible jobs?’ AI will be able to pull out the menial jobs and allow people to focus on the parts of the job that are meaningful and human.” He acknowledges that this may result in the need for fewer employees. “But the jobs that are there will be much more interesting.”

With the recent development of Facebook’s virtual assistant, M, attention has been refocused on AI in customer service. But Facebook’s effort, Mauro believes, is operating on a much smaller scale than the way AI has been applied in customer service. “Facebook is making a big hullabaloo about putting human agents behind M, and how they’re planning on learning from those interactions,” said Mauro. “That’s something we’ve been working on for quite a while now. We’re so used to thinking about humans in the loop because we have call center agents in our business. It’s almost like Facebook M is just starting to catch up to AI for customer service.”

Whereas FB hired 30 people, said Mauro, a large call center can have 5,000 employees. This makes a big difference when you’re talking about machines observing humans. “Machines learn by watching what the person’s doing. The agent does their work on a computer screen and a machine can observe that work, and map the intent, inquiry, transaction,” in order to figure out how to automate.

He predicts that in 2016, we’re going to see customer service enter the conversation around AI. “There will be legitimate excitement about how many human agents there are in the world, and what an opportunity that represents to do Facebook M-type things, but at a huge scale–well beyond the 30 agents Facebook hired.”

Nov. 25, 2015, 12:38pm ET: A previous version of this article misstated the number of agents at a call center as 50,000–it has been corrected to 5,000.