Practical AI For Speech-Based Collaboration

In my role at 8x8, I’m working with our teams in building the next generation of 8x8 products that make our customers obtain insights and capabilities needed to make better business decisions. Having been in this industry for a while, I’m constantly evaluating the maturity of AI & ML technologies and the challenges faced in implementing them for enhanced customer experiences. The July 2018 Gartner AI technology hype cycle tracks 35 technologies of which 15 are in the Peak of Inflated Expectations category. In other words, they are mostly hype. There are another 7 in the Trough of Disillusionment, which means only 4 have progressed to some degree of maturity with Speech Recognition the most evolved. Gartner research also points out that only 4 percent of CIOs worldwide have an AI project in production. Yet, everyone is talking about AI.

Why are they doing that, if it’s mostly hype? We did our own research to better understand what companies are thinking. We found that 78 percent of Contact Center Professionals surveyed believe AI will have a positive impact on Contact Center applications in the next 2 years. There was pretty strong consistency on this point. Yet, 64 percent said that when it comes to the overall customer experience, AI is not ready. Also, pretty strongly consistent, aren’t they?

So what is everyone doing? Investing in AI-based projects right now - at least 77% of them anyway. Why? They know they have to get good at AI now, even if the tech isn’t quite ready. They see areas with tangible benefits and real ROI possibilities.

The key is, where do you start? This blog is the first in a series where we will walk through what we’ve learned from experience along with research, regarding how AI, when done appropriately, can actually enhance speech-based collaboration in your contact center today.

Working with our customers, we’ve identified a number of practical uses for AI in the contact center. Three of these involve how incoming calls are handled:

  • Virtual Agent
  • Agent Assist
  • Expert Finder

Let’s take a look at each of these in more detail.

Virtual Agent

virtual_agent.png

Let’s say a call comes into your contact center. The contact is identified and the context of the caller is generated from the CRM records. The context provided includes past history, caller propensities, previous satisfaction levels and caller importance. This context is passed on to the virtual agent. The virtual agent can now predict what the caller is most likely calling about, so it can provide that option at the very outset. For example, “Are you calling about the ticket ABC123?”

Once it has provided that option, and based on the voice response from the caller and the subsequent topic modeling of intent, the topic of the call, etc., the virtual agent will now query backend systems and retrieve the information that answers the question from curated knowledgebases. This information is then presented by the virtual agent to the caller.

The ability to quickly answer the caller’s question and resolve the issue in the first call increases customer satisfaction and NPI scores, and improves the efficiency of the contact center. It also prevents the call from being routed to queues and agents, thereby improving agent utilization by freeing them up for calls that require their skills, and optimizing contact center staffing utilization for issues that require their expertise. The 8x8 Virtual Agent is powered in part by Google’s Contact Center AI suite.

Agent Assist

agent_assist.png

Agent attrition in contact centers continues to be a major issue. Due to a lack of adequate help and knowledge readily available to agents, many of them struggle to be highly efficient and successful at their jobs. This is a problem that AI/ML can readily solve.

Continuing the above example, if the Virtual Agent is unable to answer the question, then the question and the topics are passed along with the call and the call is placed in the queue. When the agent accepts the call, the question asked is pre-populated in the agent’s GUI along with relevant articles, documents, manuals, etc. in a dynamically updating Agent Assist section of the UI. Articles picked by the agent as useful are marked, and this data is part of the learning feedback that improves the content presented and published for the agents. This directly helps with reducing call handling time and improves first call resolution and efficiency of the contact center.

Expert Finder

expert_finder.png

Enterprises often have forums or team rooms that cover one or more topics and have subject matter experts (SMEs) contributing to a discussion, answering questions or adding context. Despite having the expertise to answer caller questions, Agents in a contact center need the ability to dip into this pool of collective knowledge of the organization and find the groups, subject matter experts and the knowledge they create, to answer questions from callers.

Expert Finder is an AI capability that helps agents find the team rooms and subject matter experts in a company that are most relevant to the areas the caller’s issue lies in. It traverses both the internal Unified Communications space in a company and the Contact Center space to find the necessary subject matter experts who can quickly help the caller and provide for a wonderful customer experience. 

Everyone knows that, when it comes to AI, more data is better as long as it’s accurate, complete and timely. These three examples have focused on the contact center, but what if we could add insights from meetings and business communications? It would be amazing. That additional insight would include more interactions with each customer to provide an enriched set of data that would include better initial context, more data about touchpoints along the customer journey, better sentiment analysis prior to initial call handling, etc. This would directly impact customer experience and boost NPI stats. An AI Meeting Recorder captures and stores critical discussions, ensuring nothing important is missed. The possibilities are very exciting! 8x8 is in a unique position of having all this data and in large volumes allowing it to continuously enhance the insights needed for its AI solutions.

Google Contact Center AI (CCAI)

Many of you may be asking, isn’t this what Google CCAI does? Good question.

Google’s Contact Center AI (CCAI) offers an integrated platform of services to do the following:

  1. Receive audio or text
  2. Extract topics to query for knowledge from knowledge bases (KB) using knowledge connectors, thereby providing assistance to agents with relevant articles, FAQs and manuals during a live call
  3. Text to speech (TTS) and speech to text (STT) capabilities
  4. The ability to enhance the accuracy of answers by identifying intents and entities from the transcribed speech

We learned, from deploying Virtual Agent, Agent Assist and Expert Finder, that there are a number of “last mile” topics that were needed to actually make things work. In a future blog, we’ll share what they are and why they are required.

In the meantime, find out how to get your organization ready for it’s first AI project, so you can build your AI muscle and stay competitive in your industry.