Skip to Content
Uncategorized

An Escape from Automated Customer Support Hell

Software that understands natural speech—using a combination of software and humans—could save businesses from losing customers.
October 11, 2012

Mike Iacobucci wants to help you stop screaming at customer service agents—specifically, the automated ones that misunderstand what you say or require an endless series of button pushes to complete a simple task.

Iacobucci is president and CEO of Interactions, a company that sells a new kind of interactive voice-response software for customer-care phone systems. With its software, instead of having to go through an interminable series of push-button choices or stick to overly simple verbal commands, you can talk just as you would to a human representative—and, surprisingly, it actually works.

Interactions’ software is, hopefully, more than a solution to impossibly annoying automated support systems. It’s also an example of software and human intelligence working together. Rather than relying entirely on software to handle calls, Interactions automatically hands speech that its software can’t cope with over to human agents, who select an appropriate response.

The interactive voice-response systems that you typically encounter when calling a company’s support line send chatter through speech-to-text recognition software, which submits a text result to an application that decides what step to take next.

Until recently, computers have had nowhere near a  level of understanding or accuracy to match that of a person, so systems have required users to restrict possible replies.

Now, thanks to gradual but substantial improvements in natural language processing, along with Interaction’s hybrid machine-human approach, the company’s software can do something like setting up a hotel reservation—which could take six steps with some programs—in just three steps, says Phil Gray, Interactions’ head of marketing and business development.

Cutting down the steps requires a lot of work behind the scenes. If you call the customer service number of a business that uses an Interactions application, you’ll be greeted by an artificial assistant. Your responses are passed through speech recognition software, and answers to potentially complex questions like “How may I help you?” and “What’s your e-mail address?” are routed to humans where they can be “translated” into a straightforward typed phrase and sent on to the application. Answers to simpler questions like “What’s your ZIP code?” are sent straight to Interactions’ software. The application brings together the two piles of data in real time to help you complete your call.

Interactions, based outside Boston in Franklin, Massachusetts, offers 15 different types of applications tailored for specific markets, including health care, financial services, retail, and hospitality. Its customers include hotel operators Hyatt and Best Western. Health insurer Humana uses it for Medicare enrollment calls, which Gray says would normally require a 20-minute conversation with an agent.

Rob Miller, an associate professor of computer science at MIT who studies human-computer interaction and crowd computing, thinks Interactions’ idea is a good one, because the quality of current automated systems drives people away. “It seems like a great way of using machines for what they’re good at and using human operators for what they’re good at,” he says.

I called Hyatt’s customer service line to give Interactions a try. A pleasant, mildly robotic-sounding man’s voice answered the call, asking me how he could help.

I said I wanted a room in a Hyatt in Manhattan. It found four Hyatts, listed them aloud, and then repeated the entire list when I asked, “Which was the third one you said?”

Admittedly, I tried to trip the software up—stumbling over my words, fumbling the date of my arrival (“September—I mean December 3”) and responding “Just me” when it asked how many people would be in my party. It paused several times, perhaps to let a live operator parse a bit of our conversation, and each time the silence was punctuated by a recorded typing sound.

It managed to get things right, though. And after gathering basic information about where I’d like to stay and when, it punted me over to a live customer service representative to finish booking my room (Hyatt, apparently, does this for all reservation calls).

I spoke with one of these reps, Brandon, who said the automated system gathers most of the information he needs and cuts down on call times by a minute or two—a large improvement in Brandon’s line of work. Also, he told me most customers say they like the system.

The time savings provided by Interactions’ software can add up to big savings for businesses. And customer satisfaction is financially important to Interactions too, as it gets paid based on the number of successful transactions it completes, like making a reservation or authenticating a customer’s identity.

Interactions won’t go into details but says it hopes eventually to leverage customer interactions with its human operators to improve the automated part of its service. Miller sees big potential here, saying that groups at MIT that have spent years building speech dialogue systems have found it enormously useful to have people actually use their systems. “If people are providing training data and at the same time getting value out of the service so they’re actually coming back, that’s the best of both worlds,” he says.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.