Many companies are failing to analyse their incoming data; they either don’t have it annotated, or they don’t have the right tools or skill sets to make sense of all the traffic that comes in. So instead, they end up being very reactive. For example, they get an influx of calls and divert people to answer the calls. But they fail to take a step back to understand why people are calling in and what they are calling about. They don’t look at different potential resolution options for different calls or chats, or how those conversations could be used to drive automation. Unless you understand what your customers are calling you about and have concrete data points that support it, you’re not going to be able to automate. It’s putting the cart before the horse.
Scale can also be an issue. Imagine you’re at an airport, you can visualise the people waiting in line. If you’re at the customer service desk, you will know what type of questions are going to be asked again and again. But imagine it’s 10,000 people a month in that line. How do you identify and filter all the information that’s coming in? You don’t see the people angry on the phone, or at the chatbot. How do you tag that data? How do you determine the patterns in these conversations? What are the indicators for an escalation and how can this create give a great user experience?
We are already seeing the rise of self-service activities – tag your own baggage, for example – so you can automate a line of 50 people waiting to get their bags checked. We need to be optimising these systems and creating better efficiencies.
There are already numerous examples of self-service activities that are being effectively delivered through voice AI technology using Smart IVR and multi-modal formats from I’m trying to replace my card to I’m trying to book a room or I’m having trouble accessing my account. However, there remain some calls that create frustration among callers and which require escalation to a human for resolution and an automated pathway for this escalation to occur is best for both parties. We have tried to make our technology as comprehensive as possible for use cases across the board: from AI chatbots to smart home assistants, and from smart IVR applications to multi-modal experiences, we are aiming to both deliver and optimize almost any conversational experience. As such, there are various solutions in our voice technology that allows us to annotate and analyze calls as they go to a live agent. We can then extract as much context from that voice interaction or the multimodal interaction, including the event leading to a prospective escalation to a human agent. Prospective is a key word here because there are lots of situations where the technology can effectively automate end to end. The customer gets resolution without ever coming into contact with a human. That’s the ideal scenario.
But in the case of calls where you do have to take someone to a human agent, it’s important to do so with as much context as possible, because that then makes the call between the end user and the human agent more efficient. It just gives companies better insight into what people are calling about and, after a friendly greeting, the agent can immediately get into call resolution at a very granular level – using the structured annotated data that flows directly into the contact center. So, the more we can capture that context and encode it in the support tickets that are escalated to the human agent, the more we are providing valuable information for companies as they look to make their contact centers more efficient.
Unquestionably, there will be tasks out there that will continue to require human assistance. But we can use Voice AI technology and analytics and all the other capabilities that we have in our ecosystem to add value to that human agent.