Why we need speechbots to answer the phone

Industry and media attention have focused on chatbots for digital channels. However, speech bots that answer phone calls could have a greater impact on customer service. First, because phone remains one of the most popular customer service channels, and companies should be doing all they can to meet customers where they prefer to interact. Sixty-eight percent of Americans owned smartphones in 2015, making the phone prime real estate for evolving customer engagement technologies that should not be overshadowed by flashier channels and platforms.

Second, speech bots are a major cost-saving opportunity. Voice agents far outnumber chat or social agents, and companies can substitute live agents for speech bots to greatly reduce expenditures, which means budget can instead go to other business ventures.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":2140927,"post_type":"guest","post_chan":"none","tags":null,"ai":true,"category":"none","all_categories":"ai,bots,","session":"A"}']

And third, the phone already provides a natural interface — spoken language — that is accessible even when a screen is not present. Unfortunately, consumers view traditional phone automation systems (IVRs) as a hindrance rather than a help, caused by poor natural language understanding and limited functional range.

To become intelligent speech bots, companies must enhance IVRs to demonstrate the following capabilities:

• Task resolution. IVRs can no longer act as glorified switchboard operators that simply listen to requests and route callers to agents, as they did years ago. A bot is expected to have knowledge and skills, including deep integration to enterprise systems, to complete transactions and avoid the need for human assistance. In fact, Gartner predicts that 85 percent of customer interactions will be managed without a human by 2020.
• Contextual awareness. A bot must be aware of a customer’s own context, including previous calls, activity on other devices and channels, and relevant profile information. The bot should leverage this information (with the customer’s consent) to resume unfinished conversations, anticipate new requests, and personalize experiences. This is where messaging apps are gaining traction in the customer service space — the customer’s entire conversation with the company is in a single thread without them needing to start over the authentication process. Like a useful assistant, a bot should never ask the customer a question it should already know the answer to.
• Natural language understanding. A speech bot must understand natural language to a greater extent than chatbots to be useful. First, it must be able to accept spoken input (surprisingly, some IVRs are still touch-tone only: e.g., “Press 1 for this”). Second, the speech bot should accurately infer the customer’s intent from an open-ended utterance (e.g., “I was not able to use my card when I tried to use it in the store today” means the customer requires assistance with his or her card). Finally, the system must have the skills (asking, clarifying, resolving, interpreting, etc.) to move conversations to a desired goal.
• Agent fallback. Although a bot will be trained in a broad range of service functions, there will be some requests that it is not equipped or qualified to handle. On those occasions, the bot should promptly transfer the call to a human agent, along with relevant context. If a customer calls in after failing to resolve a problem online, the bot should transfer the call immediately to a live agent rather than attempt additional self-service.
• Brand image. A bot may have a persona that reflects the company’s brand and culture. IVRs are generally soulless systems, designed to dispense with calls rather than engage customers. A bot is also a machine, but a more intelligent one. Humans associate intelligence with personality, so a bot will usually have a style, expressed in tone and language, that conveys the spirit and values of the company.

At their core, bots will have domain knowledge not tied to any specific channel. Rather, bots’ knowledge should encompass all business logic required to handle customer requests (business rules, workflows, backend integrations, etc.). Bots can then vary their interactions and behaviors depending on speech or text input, and further tailor responses based on a channel’s unique constraints and norms.

Viewed this way, just one both can act as both a speech bot and chatbot, rather than using two separate systems.

This dual approach will allow companies to make greater investments in speech bots without sacrificing current chatbot efforts. Attention to speech bots will prove useful as mobile dependency continues. Forty-one percent of American homes used just wireless phones by 2014, and almost half of U.S. smartphone owners say their device is something they couldn’t live without. Telephone devices and usage are certainly changing, and companies can be sure that speech bots will be a key part of that evolution.

Bots can already chat on websites, answer questions on mobile apps, and respond to social messages. Shouldn’t they also pick up the phone?