The 2017 State of the Speech Technology Industry: Interactive Voice Response
Another plus is that consumers gain more input options. Rather than pressing a button or saying “Yes” or “No,” users are becoming more able to deliver free-form responses. One example comes from Nuance, which worked with the city of Dallas to implement the Dallas311 customer information system. Dallas311 is based on Nuance’s Conversational IVR with Natural Language Understanding (NLU). Callers in need of information about city services are greeted with a prompt asking them: “How can I help you today?” They can then respond in their own words, asking about trash pickup, animal services, code compliance, or dozens of other issues.
Proactive, Not Reactive
In addition, IVR systems are now being built to do more than just collect information; they also try to determine the caller’s intent. Here, another shift in system design has been taking place. Traditionally when systems encountered words that they did not understand, they tried to translate the sounds into words and then linked the words to a rigid, predefined set of options. Now, the goal is to put the words into context and automate more of the customer service process.
Vendors are pushing these systems further through artificial intelligence. Newer systems are starting to take historical information and deduce why a call was initiated and what the customer wants. For instance, adaptive IVR personalizes the customer experience based on each customer’s historical behavior rather than merely presenting generic sets of prompts. If a consumer always buys a certain type of product—say, red lipstick—then information about that item greets her when she arrives so that she no longer has to sift through a couple of layers of voice prompts to reach that point.
Another seemingly never-ending quest is integrating the IVR into a more cohesive customer experience. Customers now connect with businesses through many channels. Younger customers especially start their transactions online and only pick up the phone when they run into roadblocks.
Consumers do not see a difference between a business’s web page and its call center, but the enterprise usually does. “If customers start online and then call, they want to pick up where they left off rather than start all over again,” Verint’s Thurlow says. Often, though, the process does not unfold in that manner. Barriers arise among applications and force the consumer to paint the picture all over again.
To avoid such inefficiencies, vendors have been pushing for tighter integration among IVR, mobile apps, chat, and SMS. In effect, IVRs are transitioning from interactive voice response systems to interactive multimedia response solutions. Suppliers are pulling different pieces together to meet such desires. In November, Verint extended its customer engagement portfolio through the acquisition of OpinionLab, which developed products to optimize web and mobile customer experiences.
What’s on Tap for 2017?
The ultimate goal is to integrate and automate more call processing and reduce staffing costs. Ideally, the systems handle more, or perhaps all, of the interaction. For this change to occur, machine learning needs to take on a bigger role. “Right now, machine learning is in the early stages of development,” says Nuance’s Livingstone.
This technology enables computers to learn without being explicitly programmed to perform certain functions. Machine learning features algorithms that learn from past behaviors and make predictions about future actions. Such programs overcome the strictly static programming instructions used by traditional programs.
Machine learning has been employed in a range of computing tasks—such as spam filtering, detection of network intruders, or search engines—where designing and programming explicit algorithms is unfeasible.
The technology could help IVR systems respond to customer needs more proactively. The goal is to make the customer service process more cohesive and effective. Recently, a number of pieces have fallen into place, but more work will be undertaken in 2017.
Paul Korzeniowski is a freelance writer who specializes in technology issues. He has been covering speech recognition issues for more than a decade and is based in Sudbury, Mass. He can be reached at paulkorzen@aol.com or on Twitter @PaulKorzeniowski.