Google Embraces the Contact Center

  • The last contact center slideshow that appeared here on No Jitter, Top Contact Center Stories of 2017, had the subtitle: "Artificial Intelligence Groundswell." If AI's importance to the future of customer experience software could be compared to a rolling of the sea last December, by this July it's an outright tidal wave.

    Nowhere was that more evident than at last week's Google Cloud Next event. The company reported that 25,000 people attended -- either in person or via the hours of streaming keynotes and breakout sessions on its in-house network, YouTube. One of the key announcements relevant to No Jitter readers was Contact Center AI. You can read the highlights of the announcement in my post, "Google Enters the Contact Center AI Fray," and in one from my colleague Dave Michels, "Google Clouds Enterprise Communications."

  • Amazon Connect vs. Google Cloud Contact Center AI
    The biggest piece of news at Enterprise Connect 2017 was the Amazon Connect announcement by Amazon Web Services. So how might you compare Amazon Connect with Google Cloud's Contact Center AI? Perhaps the biggest difference is that Google isn't coming to market with something to compete with the existing contact center solution providers, as AWS has done. Quite to the contrary -- according to its contact center partners, Google began working with them months ago to help shape their solutions.

    In my opinion, a more apt comparison is to Nuance with speech recognition. Years ago, as speech recognition was being brought into interactive voice response (IVR) systems to supplement or replace touch-tone signals, Nuance was often a key partner for the speech recognition technology. Even as recently as this May at its annual partner summit, Nuance recognized Avaya, Cisco, and Genesys (among others) with partner awards.

    The Google model appears similar to the initial Nuance approach (note Nuance went on to create an IVR to compete with its partners). Google has AI expertise, including machine learning (ML) and natural language processing, but not contact center domain expertise. Instead of creating something that would compete with the established contact center players, it's hoping to build something that all of them would use, with the ultimate goal of bringing more workloads to the Google Cloud Platform.

  • Fei-Fei Li, AI Rock Star
    Fei-Fei Li, Google Cloud's chief scientist of AI/ML, introduced Contact Center AI during a keynote on the first day of the Next event. Li, shown above, joined Google in January 2017 from Stanford University, where she had been a professor since 2009, the last three years as director of the Stanford Artificial Intelligence Lab.

    Partner executives who had the opportunity to meet Li over the course of the year Contact Center AI was in development or at Google Next spoke of her in reverential tones -- the way a Springsteen fan might sound after a backstage meeting.

  • Google Cloud AI
    In a session at Google Next, "How AI is Transforming Customer Care," Google Product Manager Daryush Laqab put Contact Center AI in context. He described it as part of a larger Google initiative called Cloud AI. Cloud AI efforts fall into three categories:

    • The AI Platform – for ML engineers and data scientists who need to run their ML and analytics workloads in the cloud, hybrid, or on-premises environments
    • Pre-trained models and APIs -- Google has models for speech transcription, text-to-speech, and natural language, easily consumable by other applications via API. A company interested in customizing those models with its own data can use Google's AutoML tool to create customized pre-trained models for a specific company's workloads.
    • Laqab described the AI solutions category as having the existing pre-trained models and APIs plus the ability to create new pre-trained models and APIs and package them in a contained solution to address a specific set of use cases. Contact Center AI is one of the solution use cases that Google built, he said.

     

  • Contact Center Pain Points
    In his session, product manager Laqab further described how Google Cloud worked with enterprise customers and contact center solution providers to understand pain points in today's contact centers. Google selected the three listed in this slide as problems it thought Cloud AI could help solve.

    Worth noting is that Laqab went on to describe contact center pain points not just in terms of the customer who is calling in to a contact center, but also for the agent struggling to provide answers.

  • Contact Center AI
    The first component of Contact Center AI is Virtual Agent/Dialogflow, based on technology that Google acquired when it bought four-year-old AI start-up API.ai in September 2016. Dialogflow can allow a business to replace its IVR completely, instead using AI to discern how to route a customer based on analyzing call intent, Laqab said.

    Added to what has been previously available in Dialogflow is the ability to connect the customer to a live agent, with the context of the automated conversation.

    Agent Assist is a new product that monitors the conversation between the customer and agent, determining what the question is, searching the available knowledge base, and presenting the answer or an article that contains the answer to the agent -- in real time.

    Conversational Topic Modeling, which Google essentially described as a work in progress, will be the first offering in a contact center analytics suite. Think of it as a black box that would be fed call and chat logs. Using AI, key topics would be identified as well as the keywords and top sentences customers used to discuss those topics.

  • Contact Center AI Diagram
    This graphic portrays where Cloud Contact Center AI would fit in a contact center conversation flow. The green boxes in the diagram are elements outside of Google's scope:

    • Contact Center Provider refers to a partner contact center solution (see next slide)
    • Agent Desktop could be provided by the contact center vendor or likely a CRM vendor, such as Salesforce
    • Backend fulfillment is is defined as the steps taken for receiving, processing, and delivering orders to customers. In this case, it refers to an order being sent directly for processing, without the need for a live agent.

     

  • Contact Center AI Partners
    Contact Center AI is available as an alpha release exclusively through partners. The vendors listed here are the ones described on Google's Contact Center AI website as those with which it has "strong partnerships." Each of the five contact center vendors listed -- Cisco, Genesys, Mitel, Twilio, and Vonage -- participated in Google Next, and I share more information on how they're using Contact Center AI in the slides that follow.

    Vendors listed on stage, and which may have had booths in the expo area of Google Next, included Five9 and RingCentral. Also, Google did mention that it has a limited number of additional partners and customers it can support in the alpha phase of product development, so this list might expand over time.

  • Genesys: First Among Equals
    If any contact center vendor could be described as first among equals, it's Genesys. It alone joined Google chief scientist Li on the day-one keynote stage. Genesys CMO Merijn Te Booj (center) and customer Dan Leiva, vice president of customer service technology, eBay, joined Li to talk about how they're working together on AI in the contact center.

    Te Booj outlined three guiding Genesys AI tenets that map to Contact Center AI capabilities:

    • Genesys thinks customers should be greeted by natural language, that a normal dialogue is the ultimate answer. With Genesys and Contact Center AI, that's now possible, he said.
    • Blended AI is the path to the best customer experience.
    • Bringing AI to the desktop is important; this can be in the form of articles and knowledge -- even microbots for making certain tasks easier to do.

     

  • Genesys and eBay Contact Center AI Prototype
    The demo showed a customer, Mala, calling to return a pair of shoes. First, the virtual agent answers and asks if the call is about the recently delivered shoes. Mala responds that yes, it is, and she wants to return them. The virtual agent sets up the return and tells Mala she'll get an email with details... in about 10 seconds.

    Proactively, the virtual agent then asks Mala if she'd like to speak to a live agent for help selecting the right pair of shoes. Contact Center AI works with Genesys predictive routing to find the right agent to help. Genesys predicts that Josh, of all the available agents, is the best one to help Mala. Josh receives the call as well as the context of the initial virtual agent interaction, plus articles that will coach him on assisting Mala.

    Genesys also presented the solution in a breakout session, "AI Powered Contact Center Analytics."

Exploring how the cloud platform giant is teaming with major players to bring its AI expertise into the contact center