It’s Time to Cross Speech Tech’s Annoying Valley

Article Featured Image

Leonard Klie, our fearless editor, has an article in this issue about speech technology and “the uncanny valley.”

The uncanny valley describes the stage of a technology in which it has become humanlike but at the same time not human enough, making people uneasy. Given that speech technology continues to improve, have we moved into the realm of acceptance and out of the uncanny valley?

I’ve got a different question: Will we move out of “the annoying valley?” Having coined the term I get to write the definition: “The annoying valley is the gap between when a new technology is introduced and when someone using it actually finds it useful instead of annoying.”

Case in point: Most interactive voice response (IVR) systems, even speech-enabled ones, replace human interaction with a series of fixed menus. And with all due respect, they are almost all annoying.

Of course I have examples, and not only from hastily-built IVR systems and web pages but from carefully designed systems in use for years. My internet service provider uses a national call center and makes no effort to recognize my phone number when I dial in. Even after I enter my account information and authenticate myself, I get a menu that asks me what service I want to talk about—even though I’m only subscribed to a single service. And you’d think that somewhere, someplace in their database there’d be a notation that said, “This person is an expert. He’s already rebooted his modem and logged in to his personally owned cable modem and checked to see if the upstream and downstream channels are working. If he’s calling, he needs serious help.” Instead, their IVR system insists on telling me every time that “most problems can be handled on the [X company] website.” Uhhhh … if I could reach the website, I wouldn’t be calling. That company is stuck in the annoying valley.

Inside call centers, advances in AI and speech technology give management new tools to implement oversight over their personnel—and create new annoying valleys, this time for the employees. One article I read gave details about a company’s new AI-powered, speech-enabled “assistant” that monitors phone calls and prompts agents to use certain key phrases during calls that, in management’s opinion and assessment, provide the best practice.

Some agents, formerly top-ranked according to the article, eventually quit in frustration. The rigid strictures of the AI- and speech-driven system narrowed their personal and effective styles; in the end, work became too annoying and they left. Of course they’re probably a minority; but if the creative agents leave, how will the company improve service? (My criticism may be misplaced; in other contexts I’d prefer that management retain strict control. There’s nothing like, e.g., a lab accident with chemicals [colleagues] or radiation [me] to make someone a little more careful with technology.)

So what is the role of AI and better speech technology? Let’s start with a useful, helpful, and smart IVR system to provide a baseline of expectations. My local electric company has a clever IVR system. If I call to report an outage—and before I even interact with the IVR—the system uses my caller ID, checks my account, and lets me know immediately if an outage is known to the company and (if I recall correctly) the estimated time to repair. All this is automated: The IVR volunteers the information, so there’s no need to talk to an operator or even select an option. This is one of the few companies I know of that has escaped the annoying valley, while at the same time cutting costs by getting me off the line ASAP and keeping me away from live operators.

And that example provides the key. That electric company IVR system doesn’t use AI or sophisticated speech technology, but it looks at all the information it has about me as a customer, realizes why I’m calling, and acts accordingly.

AI and speech technology, taken together, provide a potentially terrific way to collate information about callers. Did the caller just receive a package? Maybe that’s the problem. Did their bill just go out in the mail? Maybe that’s it.

I might be asking too much. With current tools, if I proposed a system like this to a client, the cost would be astronomical. But perhaps there’s some hope for very smart systems with an affordable price tag, and IVR will finally leave the annoying valley. x

Moshe Yudkowsky, Ph.D., is president of Disaggregate Consulting and author of The Pebble and the Avalanche: How Taking Things Apart Creates Revolutions. He can be reached at speech@pobox.com.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues