Survey Shows Frustration with Traditional Approach to AI

Speech tech providers should realize that people expect any voice communications on a smartphone to be very short and will shy away from voice recognition “help” that is often more frustrating than helpful because the actual speech recognition is poor, says Abinash Tripathy, founder and CSO at Helpshift. 

Tripathy based his comments on the findings of a recent customer survey conducted for his company that found that 66% of Americans feel that voice-based digital assistants are more unsettling than text-based ones, meaning that Americans find chatbots and digital assistants likable while more advanced AI technologies are far behind their promised service, so they are more frustrating than helpful.

He adds that speech technology providers have to move away from traditional technologies, which tend to have a 75% speech recognition rate, and move to technologies employing deep learning and neural networks, which boost the recognition rate to 90% to 95%. Customer support must also change in a big way, according to Tripathy. The most successful brands, like Amazon, Apple, and Google have removed automated speech response now moved into a digital experience and away from a phone experience. Those companies seek to provide help through their websites or mobile apps, rather than IVR or speech recognition. Then if voice communications are needed, these companies rely on an outbound calling approach rather than an inbound approach.

The Helpshift survey also found that a majority of Americans (57%) feel that it has not improved in the past few years, and 74% agree that contacting customer support is a frustrating experience. The frustration tends to come from:

  • Long wait and hold times — 60%
  • Being transferred and having to repeat the problem over and over — 59%
  • Never-ending automated “self-service” menus — 50%
  • Difficulties communicating with agents in offshore contact centers — 49%
  • Poor levels of service — 41%

It’s so frustrating, in fact, that many (44%) would rather spend 30 minutes cleaning a bathroom than waiting on hold for customer service; 26% would rather spend 30 minutes at the dentist; 25% would rather sit in traffic or visit their in-laws; and 24% would rather do their taxes.

The survey also revealed the perceived pros and cons of chatbots. Americans’ top pet peeve about them is that they feel chatbots prevent them from reaching an actual human (51%), and a majority (77%) get annoyed if they have to type more than four responses to a chatbot.

On the other hand:

  • If forced to leave a brand’s website or app to reach customer support, almost half of Americans (47%) would take their business elsewhere.
  • Three-quarters (76%) say chat-based messaging would be their preferred method of contacting customer support if they knew they would get an immediate response.
  • Ninety-two percent feel it is important to be able to return to a conversation with a customer service agent at a later point and pick up where they left off.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

3 Ways AI is Making Your Office a Better Place to Work

Many workers are worried about losing their jobs to AI, but others are grateful for the many ways new, AI-based technologies are making their offices a better place to work.

Speech Tech Systems Need Expanded Capabilities, Training for Improved Customer Service

Seventy-four percent admit they are more loyal to a company if they can talk to a person—versus a machine—when they need assistance, according to a new report from Calabrio.

Engaging Younger Audiences with Voice: New Opportunities to Connect with a Younger Demographic

Younger consumers—those 18 to 24 years old—are helping to drive adoption of voice technology, yet a study by PwC shows their use of it remains low. What does this mean for companies in the speech sector, and where can inroads be made?

GDPR Implications for Speech Technology

With regard to AI-enabled devices like Siri and Alexa that listen to surrounding conversations, vendors will likely have to gain permission from European consumers to record and store voices, and ensure that downstream applications are GDPR-compliant as well.