When you start thinking of voice as information, anything said within the course of a phone conversation has the potential to be used. Real-time processing of voice by call centers and enterprises is poised to introduce new tools to streamline business processes and possibly cut out a lot of middleware in the world of "unified" communications.
Tapping into voice as information isn't new. Voice analytics has been in use for over a decade at large call centers, aggregating thousands of phone calls a day and then applying big data-style tools to extract best practices, market intelligence, and competitive threats. Since the value of the insights gathered comes from calls made on a daily and weekly basis, analysis typically takes place in batch mode. There's typically no need or value to doing real-time analysis, except possibly in a public safety access point (PSAP) – an E-911 center.
Processing voice in real time requires processing time and the tools to understand voice as it is naturally spoken in a conversation. Moore's Law (yes, shoot me for the cliché) is providing enough CPU power to enable real-time voice processing while natural language processing such as the ColdSmoke platform can parse through the words to find meanings.
What is natural language processing? On “Star Trek: The Next Generation,” Captain Jean Luc Picard had to tell the computer to make "Tea, Earl Grey, Hot." That's not natural, at least not for 21st century English.
Capt. Jean Luc Picard with his Earl Grey tea. Image via HealthSnap
Instead, I'd say "I'd like a cup of Earl Grey tea." If the computer had any brains at all, it would ask me "Hot or cold?" Alternatively, based upon the time of day and my previous actions, the computer would automatically figure out if I liked my tea hot or cold, as well as what type. It’s a lot faster to say "Tea" or "Tea, please" than providing a sequence of words in an awkward format.
One idea for applying actionable voice in today's world is the "supervisor" intervention, briefly discussed at the first Smart Voice event in Miami earlier this year. A call center agent gets into a discussion with a customer. Sometime within the conversation, either the customer expresses frustration or the call center agent says "I don't know" and "I'm not sure" once too often; the actual trigger words and conditions would be set through some initial trials. Once the trigger threshold is hit, a supervisor would automatically be notified and brought into the call in real time, perhaps with a 10 to 20 second delay so he could read a real-time transcript of the call before being injected into the situation.
Human intervention doesn't have to be the only action triggered by voice. A conversation between a sales representative and a customer could generate a draft contract or fulfillment order on the fly. At the end of the conversation, the work piece could be quickly reviewed by both parties via email or a mutually accessible Web portal and then executed with a click from the customer.
Some will argue that you don't need a discussion between a customer and a sales representative, with orders more efficiently handled strictly through the Web. But people will have questions and seek reassurance through a conversation. Providing automation through voice processing speeds up the process of work order generation without a customer agent having to search through a webpage and click away to generate an order.
Other options include trigging a follow-up action of some sort, be it a quick follow-up call from a supervisor or subject matter expert, an email message providing more details on a product or serve, or even the postal mailing of a more substantial package of information.
The key point with actionable voice is being able to trigger actions based upon what the customer says during a conversation. Businesses can customize both triggers and actions as needed and make them as simple or complicated as they need.
