AI, ML and Voice Come to the Contact Center

    Artificial intelligence (AI) and voice are affecting contact centers, and that will grow as the technologies continue to evolve and increasingly support each other.

    Powerful technologies are coalescing in the contact center. The trio of artificial intelligence, machine learning (ML) and voice recognition is increasingly the focus of research, investment and new services. While it’s certain that the impact will be great and varied, the pieces are just starting to link together. Precisely how the trio will change the contact center landscape is not yet clear. One thing is certain, however: There will be fundamental changes.

    Vonage Chief Product Officer Omar Javaid told IT Business Edge that the entire category is developing rapidly. “The thing is that while it’s early, it’s moving fast,” he said. “The popular image of AI is somebody chatting with a voice assistant that is indistinguishable from a human. That exists…Also there is a lot of AI and machine learning on the back end in terms of data analysis. In contact centers, there is typically a lot of post-call analysis.” Nexmo, Vonage’s API platform, announced partnerships, expanded target businesses and introduced features in June.

    AI and voice are about more than a nice human-to-machine chat. Javaid said that AI can analyze what leads to caller anger, frustration, satisfaction and other reactions. Human and automated vocal responses can be tailored accordingly.

    People are learning how to integrate the technologies more effectively and efficiently. “I would say from a technical perspective, we have reached kind of a unique tipping point,” said Cory Treffiletti, the chief marketing officer of Voicera, which just joined the Nexmo network. “Until now, you had to learn the language of a computer to communicate with it. Now we have software with natural language processing and machines understand what we say. They are understanding our language. It makes it so you can talk to a machine.”

    Powerful Use Cases

    How far voice recognition has come is evident in Google Duplex and Vonage’s Nexmo.

    Duplex was introduced in May. The best illustration is a video of a voice assistant making a hair cutting appointment for her “boss.” The startling thing is how natural the voice assistant in the call (which starts at about the 1:10 mark of this video) sounds. For instance, at two points she makes the sort of “filler” noise humans make during conversations (ie, “Ummm” and “Mm-hmm”). The inflections are almost perfect.

    The lag between when the human at the hair salon speaks and Google Duplex answers is a tiny bit longer than in a human-to-human interaction. That’s unavoidable and not disruptive. The other flaw is that the machine repeats AM and PM in each mention of a time for the appointment. That flaw certainly can be easily rectified. It’s unlikely that many people would guess that they are talking to a machine.

    In a blog at Nexmo, Thomas Soulez describes a scenario in which a woman calls for roadside assistance from a cold climate locale late on a Saturday night. The AI platform quickly calls up the information associated with her account.

    That investigation may find that the subscriber has never called roadside assistance in 10 years (indicating that the situation likely is serious). The system also will know that exposure to cold can quickly lead to hypothermia. Thus, the call would be given priority. The AI system also gives the contact center help in finding the woman’s precise location, such as sending her pictures of structures that may be nearby in hopes that she sees them.

    Wrote Soulez:

    While it’s not a leap to assume someone calling late at night from an icy city might need urgent help, non-obvious patterns will be revealed in both public and private sources of data. Machine-learning tools will then anticipate how best to respond when it sees those patterns unfolding. Everything from staffing levels, through the best promotions to run, to the type of interaction a customer prefers will be set by software programmed through machine learning.

    There is a tremendous amount of creativity at play. Javaid offered an example. Large companies (and some smaller ones) often have strict rules about travel. In this example, an executive’s travel plans to return home must be changed due to a storm. Variables to consider include what options are available from the airlines and which are permissible under corporate rules. The AI platform would find something that passes muster with the company and the employee. The backend system will make all the arrangements, such as putting the information in the traveler’s schedule.

    Enter Voice

    Voice eventually will be part of this example, Javaid said. At some point, the employee will be able to discuss the travel situation with a chatbot. There may be things that the platform doesn’t know about that would influence the rerouting, the person may have questions, or other more subtle changes may need to be made (such as a change in how the traveler gets home from the airport).

    AI and voice in the contact center have not fully arrived, however. Orbifold Consulting owner Francois Vanderseypen, Ph.D., says that it arrives in stages. The first step generally is use of purpose-built chat boxes. The next step is to embed intelligence in the system, such as software that determines the best person to whom to route a call, keyword detection, or voice analysis to detect fraudulent insurance claims, Vanderseypen wrote.

    There is still a long way to go. Vanderseypen suggests that there are a lot of issues left to work through before this sophisticated advanced communications truly takes hold. He wrote that these systems generally only work well for English and a few other languages and that they even struggle with accented English and difficult to pronounce names.

    Even Google, which has the largest data repositories, has trouble. “No company can compete with Google when it comes to data,” Vanderseypen wrote. “Because of this, they have the very best language-understanding right now. This understanding is, however, generic. If your business is health care or something very specific, the Google services will fail [in] understanding the domain-specific language.”

    Venderseypen added that some oppose Microsoft or Google storing data in the cloud or servers that may be in other countries. There also may be legal issues. Finally, bringing these systems up to speed – “training them” — is difficult and expensive.

    There is a lot of research and innovative thinking at the nexus of AI, machine learning, voice interfaces and the contact center. The next few years clearly will be an interesting ride.

    Carl Weinschenk covers telecom for IT Business Edge. He writes about wireless technology, disaster recovery/business continuity, cellular services, the Internet of Things, machine-to-machine communications and other emerging technologies and platforms. He also covers net neutrality and related regulatory issues. Weinschenk has written about the phone companies, cable operators and related companies for decades and is senior editor of Broadband Technology Report. He can be reached at [email protected] and via twitter at @DailyMusicBrk.

    Carl Weinschenk
    Carl Weinschenk Carl Weinschenk Carl Weinschenk is a long-time IT and telecom journalist. His coverage areas include the IoT, artificial intelligence, artificial intelligence, drones, 3D printing LTE and 5G, SDN, NFV, net neutrality, municipal broadband, unified communications and business continuity/disaster recovery. Weinschenk has written about wireless and phone companies, cable operators and their vendor ecosystems. He also has written about alternative energy and runs a website, The Daily Music Break, as a hobby.

    Latest Articles