Admit it, you dread calling a customer service rep to help you with your malfunctioning internet service. Tensions rise, the call itself can feel like an eternity and sometimes your issue isn’t resolved at all. But that could change with help from a familiar, if not mundane, phrase:

“This call may be monitored or recorded for quality assurance.”

Startups like Voci Technologies in Squirrel Hill are reliant on that mantra because it means a growing pool of valuable voice data is being mined for analysis. Call centers in North America produce over 6 billion hours of such content annually, according to the company.

Voci, a Carnegie Mellon University spinoff specializing in voice transcription and speech analytics, is using artificial intelligence and machine learning to make meaning of those frustrating calls — and to figure out how agents can better handle their customers.

Advertise with us

That’s welcome news for call centers, which typically audit employees by monitoring a few random calls per month, with a high probability of missing that one problematic interaction, that needle in the haystack.

Mike Smocer, chief revenue officer at Voci, noted that one large bank  the company is beginning to work with monitors only one in 100 calls for each agent.

“That’s why teaching computers to hear is so valuable,” he said.

New age call centers

When customers say they want to cancel their subscription, a red flag is raised, said Paul Stockford, research director at the National Association of Call Centers, a nonprofit based in Hattiesburg, Miss.

Advertise with us

Words like “hate” and “cancel” trigger a type of software system called speech analytics, which monitors call center agents for quality assurance.

The technology has advanced to where a supervisor can listen to a call in real time and assist the agent by literally whispering into their ear, Mr. Stockford said.

“... You’re complaining because you hate your TV service and I’m hearing you in my right ear,” he said, “and in my left ear, my supervisor is whispering, ‘Offer them HBO free for six months.’”

Voci takes speech analytics a step further than existing technologies by integrating machine learning and AI into the software package.

The startup can not only extract the content of a call in real time, but also detect which language you’re speaking, your gender and even your emotion. There are measurable, physiological changes in the vocal tract when people become emotional, said Anthony Gadient, president and CEO of Voci.

Advertise with us

When Voci generates call center transcripts, they even come with an emotional intelligence score so that companies can seek out the points in a conversation when an agent or customer is flustered; that’s where intervention is probably needed.

It’s an opportunity to call back the customer and fix their issue before they switch to a competitor, as well as a chance to coach the employee on what went wrong.

Data here, data there

A crowd of competitors work in the speech analytics space, like Toneboard, based in Latvia, and San Francisco-based Gridspace. Mr. Smocer said Voci differentiates itself with artificial intelligence.

By 2020, the market will be worth $1.6 billion, according to the Indian research firm MarketsAndMarkets.

Advertise with us

When a client works with Voci, it most frequently purchases the software as part of an annual subscription. Once it opts in, the company must decide how frequently it wants reports on the audio being captured and analyzed because it factors into the cost.

Some clients ask for real-time analytics and others prefer reports at the end of a day or week. The cost to opt into a Voci subscription ranges, but it’s usually hundreds of thousands of dollars per year, Mr. Smocer said.

When it comes to handling the data, Voci is flexible depending on the type of business.

Banking and finance clients, for example, are uncomfortable outsourcing call center data due to their sensitive nature. Those clients pay for hardware that houses Voci’s tech on-site in a data center. Others, though, can have their data sent to Voci through the cloud.

Voci Technologies

Teaching computers to hear

Advertise with us

Voice data is becoming increasingly prevalent, Mr. Gadient said. Though Voci and Amazon’s voice assistant, Alexa, both analyze it, they take different approaches.

“Just like you have different-sized wrenches in your toolbox for different-sized screws, you’re going to have to have different speech recognition systems for different applications,” he explained.

When chatting with Alexa, the conversation is task specific. Conversations with a call center agent are more complex.

Plus, when you talk to the agent, telephone providers are actually squeezing your voice data into half its size to accommodate more voice traffic over the same number of lines.

When you’re trying to teach a computer to hear, Mr. Gadient said, it receives about 50 percent of the info it would otherwise capture. In other words, the compression algorithms that make the calls compact are limiting — it’s like having a photo reduced by half of its pixels.

Advertise with us

To teach its algorithms, the company needed access to a huge amount of voice data.

InMoment, a Salt Lake City-based customer experience behemoth, became Voci’s first customer in 2012 after supplying the company with call center audio. InMoment provides services like call center management to a significant portion of retailers in the U.S.

Now, Voci has more than 100 enterprise customers.

To date, the company has attracted at least $18 million in funding, according to Crunchbase, a portal where firms self-report capital-raising rounds. In March, Voci secured $8 million in Series B venture capital backing.

Some of that will be used to expand its team, which should include 50 employees by the end of 2018. Voci declined to number its current staff.

Advertise with us

The “profit center”

Right now, around 31 percent of North American call centers are using some sort of speech analytics software to analyze calls, Mr. Stockford said, citing a 2018 survey of members within the National Association of Call Centers.

Adoption of the technology is not yet widespread, he noted, because most companies still view call centers as “cost centers,” and want to avoid dumping additional capital into that portion of the business.

But call centers are rife with data not only on what problems customers are having, but on the problems themselves. “They don’t look at it as a potential profit center, which is a huge error,” Mr. Stockford said.

Voci’s customers use its technology to not only monitor calls for quality assurance, but also in marketing. Customers tend to reveal interesting information about competitors, like new features, pricing or promotions.

Advertise with us

Call center agents often know more about customers than C-suite executives, Mr. Stockford said.

“Why not share that throughout the enterprise?”

Courtney Linder: clinder@post-gazette.com or 412-263-1707. Twitter: @LinderPG.

Join the Tech.pgh Community!

Sign up for Courtney Linder's free newsletter.