When Is Speech Technology too Creepy?

Industry, Privacy Considerations

Personalization usage is limited in some industries. The debt collection industry, for example, will use outbound IVR systems to identify call recipients. For debtor John Doe, the system will ask if a person answering the call is John Doe.

Privacy rules prohibit the message from actually mentioning a debt, so if the debtor is not available, the system will ask for a return call "due to a matter that needs to be discussed further." The company must identify itself (but is not required to mention it is a debt collector), and the name of the collection agency need only be revealed when asked, during a subsequent live call.

Though not spelled out by law, to avoid creepiness (and ethical concerns), similar privacy considerations should be taken into account for outbound IVR systems for healthcare, financial services, schools, and other institutions in which the messages involved could divulge personal information to the wrong party.

Wilpon adds that companies requesting or responding with personal information should first ensure that the customer isn’t taking the call through a TV or computer or using a speakerphone so that privacy can be assured.

Sentiment Matching: Cool or Creepy?

Speech technology has advanced to the degree to which it can now detect sentiment, determining whether a person is angry, calm, frustrated, etc. With such technology, an angry caller can be quickly routed to an agent rather than being forced through an IVR menu, Nowlin-Green says.

But sentiment matching can be tricky, says Tara Kelly, president and CEO of Splice Software. The idea is to read the consumer's emotion and tone. But the angry caller might simply want to vent, and could be further angered by any response other than being connected to an agent. A "cheery" response to a "cheery-sounding" person may sound too cheery.

Consumers will consider any misunderstanding of tone and sentiment as creepy, Kelly says.

Advanced Natural Speech

Companies can also personalize speech technologies to detect and respond to different speech patterns and use "local language," but the effectiveness of—and wisdom behind—such personalization is a matter of debate.

"If I'm speaking with a thick southern drawl and the voice response system responds in a thick southern drawl, I might think it's mocking me," Witcher says.

Similarly, just because someone lives in a certain part of the country doesn't mean he or she is a native. So someone living in Georgia or Tennessee might not use "y'all" and other examples of regional dialect; many others will. Those who do might prefer systems that "speak the language" the customer is comfortable with, so companies need to choose how colloquial they want their system to be.

Nowlin-Green says Nuance lets its business customers determine how localized they want to be with their voice response units.

The Future of Speech Personalization

Personalization needs to be truly personal to be a customer-pleasing experience, Nowlin-Green says. "Personalization is a big job. It takes as many people to own personalization as it does to own the technology. You have to ask yourself if it truly increases customer satisfaction because it is an effort. It's not just one guy in the office. It is not a small endeavor."

Some of the promise of personalized speech technology will become more evident as the connected devices in the home become more commonplace, Kelly says.

She foresees insurance companies using speech systems connected to smart devices in the home to make outbound calls to customers informing them when a pipe has burst, a furnace or air conditioner isn't working, or similar problems. Such a call would enable the customer to quickly contact someone to attend to the problem, potentially limiting damage and insurance company payouts.

Ten years ago, a similar call to someone's home would certainly have felt invasive. But as consumers become more comfortable living in a digitally connected world, what was once considered creepy can today seem cool. Proper execution—with transparency and integrity—can help swing personalized speech technology more toward the latter and away from the former. "Personalization should never be creepy if you do it right," Witcher says. 

Phil Britt is a freelance writer who focuses on high-tech, financial services, and other industries. He can be reached at spenterprises@wowway.com.

6 COOL PRACTICES

1. Request permission before using personal information.

2. Enable customers to opt out, even if they have already opted in.

3. Carefully consider customer response to sentimentality and tonality indicators before using them.

4. Monitor customer reaction to the personalization technology to determine acceptance.

5. Don't collect unneeded personal information.

6. Use personalization only if it adds value.