It’s Time for an End to ‘Speech-Disabled’ IVRs

The turn of the calendar always brings the delight and joy of open enrollment for health plans and other corporate benefits. So critiquing interactive voice response (IVR) systems becomes an unavoidable task for millions of Americans. 

This year it occurred to me that we are experiencing a dichotomy of flawed speech-enabled IVRs. In nearly all cases, the applications are either nothing more than old-fashioned dual-tone multifrequency (DTMF) offerings with a layer of “Hey look at us, we have speech recognition!” on top, or they’re overly ambitious “speech-disabled” (not enabled) products. 

In the first case, these faux speech applications waste callers’ time with long explanations of what they can do in the IVR: “If you are calling with a financial question, say ‘Billing’ or ‘Invoice’; if you are calling about a claim, say ‘Claims’ or ‘Case.’” I asked several companies why they took this route, wondering if it was part of a crawl-walk-run strategy to move into more complex speech applications. The answers fell into two buckets. The first was “Our Siri-equipped customers expect speech given their familiarity with it” (see my Speech Technology column “As Consumers Embrace Speech Technology, Contact Centers Must Adapt,” in the Spring 2017 issue). The second: “Because most of our callers are calling from a mobile phone, we don’t feel they should be forced to use keyboards/buttons to interact with our IVR.”

Unfortunately, the first reason turns out to be a case of contact center technology being dumber than smartphones. Expectations of Alexa, Google Assistant, Siri, or Cortana functionality are dashed by these speech-enabled DTMF applications, and these companies are fooling nobody. By increasing the time spent listening to overly long prompts, these applications needlessly raise costs while lowering functionality.

I tested the second reason by calling the IVRs while on a mobile device, in a mobile setting. Every single application failed due to background noise. Clearly, the companies providing this excuse have not tuned their speech recognition for true mobile use. Therefore, they are, again, spending more money to not serve their mobile customers.

The complex speech-disabled offerings, meanwhile, shared a common theme of attempting to facilitate overly complex routing among too many agent skills (see my “Easing Customers’ IVR Journeys,” in the Spring 2015 issue). Most of these applications are also attempting to offer open dialogue/natural speech recognition yet are miserably failing in their operation.

The first part of this problem is business-related and can be addressed by using different toll-free numbers for each of the top reasons customers call. The idea of “One phone number to reach us” rarely has worked for industries that have complex business processes. Furthermore, since many customers attempt to self-serve digitally, offering a link to the correct phone number is any easy solution. Healthcare or insurance providers can list those numbers on the member card—e.g., for financial/billing, claims, benefits, all other. Each of these numbers can present a short routing selection when dialed to refine routing if required. Otherwise, all you are offering is a 1-800-IHate-You vanity number that none of your customers will remember anyway.

The second part of the problem is the age-old problem of deploy and forget. Open dialogue/natural speech recognition is not easy, and it requires a significant investment for full-time application tuning. Alternatively, outsourcing these applications to providers of human-assisted speech recognition is likely to be more cost-effective and customer-friendly. 

Quick, name two speech-enabled IVR applications that you have recently used that you can point to as being very good to excellent. I hope that you had a strong cup of coffee before you had to think about this. It is very difficult to name one, even for those of us connected with speech technology.

Though speech recognition has creeped into nearly every American’s daily routine, we still suffer from the pain of speech-disabled IVRs in contact centers—now perhaps more than ever. As someone for whom speech recognition has been a large part of my career while also being an inevitable part of my life as a consumer, I think it’s time to call for the end of speech recognition in contact centers’ IVRs—unless the technology is delivered in a nearly flawless fashion.

Perhaps your mother told you, “If you don’t have anything nice to say, don’t say anything at all.” I find this appropriate customer experience advice for today’s contact center technology owners.