The Meandering Path from IVRs to IVAs
Still, despite their many flaws, because IVRs have been so integral to customer support, companies continue to rely on them heavily to this day. But for how long? The long-term prognosis is that these systems will be phased out as companies loosen the purse strings a bit for more modern voice interfaces. “With the emergence of new consumer voice systems, corporations see voice as a very good customer interface,” says Adam Rubin, senior director of product strategy at 7.ai.
Look for IVR systems to be supplemented by emerging technologies, like artificial intelligence and machine learning (see sidebar).
Chatbots represent the first wave of new solutions. The difference between an IVR system and previous options is the user interface. While early systems relied on touchtone phones; the new systems recognize users’ speech. As processing power has grown, speech recognition accuracy has improved, and companies have been adding such features to their contact center applications.
Yet like IVRs, chatbots have underwhelmed customers: 71 percent of Americans would rather interact with a human than a chatbot or some other automated process, according to a PwC survey. Why? “In many cases, chatbots are not intelligent; in fact, many are quite stupid,” Fluss contends.
The growing popularity of consumer-based voice solutions created a gold rush, resulting in a bevy of new voice application suppliers. “Everyone and their brother, mother, father, stepchildren, aunts, and uncles seem to be developing chatbots,” Fluss adds.
The market boomed from a small number of highly focused IVR companies to a glut of chatbot suppliers. In the consumer market, chatbots have been used for simple functions: Fans calling a special number might get to “talk” with their favorite movie star, for example. Such use cases have bled into the business world, where expectations are much higher.
Compounding the problem, system capabilities were oversold. “Some companies believe that AI is magical: all they have to do is throw a chatbot at a problem and AI will quickly solve it,” says Arakelian.
The levels of sophistication among such solutions varies widely. “A lot of chatbots seem like they were built in 30 minutes,” says Rubin, noting that consumer-based systems have eased speech application development considerably, allowing virtually anyone to get a system with rudimentary capabilities up and running quickly.
Also in certain cases, companies did not learn from their mistakes. Like IVRs, some chatbots present users with a list of narrowly defined options. If they need something else, the system becomes confused and goes in circles. The consumer, not surprisingly, ends the interaction quite aggravated because expectations are a bit higher, since the system is supposed to be intelligent.
THE NEXT WAVE
IVAs represent the next milestone on the journey. Recently, companies’ have changed their views of contact center technology. Rather than a cost deflector, customer service is seen by many as a revenue generator, or a revenue inhibitor: After one frustrating interaction, 77 percent of customers are ready to switch companies, according to a Frost & Sullivan report. Companies, in some cases, now focus on the total cost of poor customer service rather than a short-term reduction in personnel expenses.
Also, speech recognition has been improving, with many systems offering conversational AI capabilities, which are supposed to be more flexible and open-ended than traditional voice responses. “The first question that an IVA asks is ‘How may I help you?” explains Interactions’ Freeze.
Roadblocks still exist to keep companies from reaching such lofty goals. Though conversational Al applications are limitless and computers have become better in many areas, they are not human. “Humans can give different answers to a question and mean the same thing,” Burnett says, pointing out that sometimes, IVAs will not make such connections, or conversations will veer off in directions that they do not understand.
Companies do have a few options to address these shortcomings. They can limit the type of inquiries that systems handle. If a user inputs a seemingly straightforward question, then the IVA handles it. If the user has a complicated issue, that interaction can be routed to an agent.
Interactions, for one, has taken a novel approach to addressing the problem: Human agents stand by at the ready, and, if an IVA conversation veers into unknown territory, the agent steps in, responds, and the end user never knows she interacted with a human and not a machine.