“Please stay on the line for the next representative…” is a phrase we are all too familiar with that engenders mild feelings of anxiety and dread. However, calling customer service is an unavoidable part of our lives. Whether it’s paying the utilities bill, troubleshooting Wi-Fi, or rescheduling a flight, anytime we come across a problem, we usually need a human on the phone to help us resolve it. Any way you look at it, customer service is a mission critical problem that will only get harder to manage. Many systems are interconnected now, so customers are more likely to experience similar problems at the same time and overwhelm support agents. This explains why over $200 billion is spent each year on customer support, with the amount expected to double by 2024¹.
As early-stage AI investors at Point72 Ventures, we have been looking for technologies that could make customer service better, faster, and cheaper. In particular, we were interested in AI-powered dialogue systems so robust that customers would rather speak to them than human agents. But despite the promise of AI transforming call centers, we are still far away from the “magical” AI agent that can automate all customer interactions. Even Google Duplex, considered the most advanced AI agent in this space, is still limited to simpler interactions like booking restaurant reservations and salon appointments. As advanced as it seems, we’re still pretty far away from AI agents walking us through each step of installing a modem and router (i.e. “What black cable? In what port? …nope, it’s still not blinking green”).
Our frustrations with customer service can be categorized into three buckets — 1) long wait times, 2) dumb interactive voice response (IVR) systems, and 3) inexperienced call center agents. Long wait times are terrible and are one of the top 5 gripes of customer support² (chart shown below). How all too real is the story of a cancelled 11pm flight that leaves hundreds of people meandering around the airport terminal, waiting on 2+ hour holds to rebook their flights? Also, at some point, we’ll come across the interactive voice response (IVR) system. This system was designed to make calls better by efficiently routing us to the agent most likely to answer our questions. Instead, these IVR systems are mostly exasperating, causing us to mash 0 immediately upon hearing the cold robotic voice and demanding to get connected to a human ASAP. Lastly, once we’re finally connected to a human agent, he or she might be unable to answer our question, completing the perfect storm that is known as customer support.
There are structural reasons for these issues. The IVR systems we are greeted with are often rigid, low-level systems, which quickly frustrate us, driving us to demand human agents and drag out wait times. These systems also further compound the problem because they are unable to understand the varied syntax of human speech, forcing us to repeat the same information multiple times. Additionally, call centers can be tough places to work and face massive turnover with agents staying only an average of 12 months. To keep their lines staffed, call centers are forced to rapidly hire and train agents by the hundreds each week; this turnover means that agents are always inexperienced, which further exacerbates long call times and low customer satisfaction.
We believe that PolyAI can transform this industry from the bottom up. When we first met Nikola, Shawn, and Eddy, we were impressed by their humility about the problem and ability to empathize with the call center agent. Although Nikola, Shawn, and Eddy all received their PhDs in AI-powered dialogue systems from the University of Cambridge, they were the first to admit that AI would not be the “magic wand” in this industry. Instead of replacing human agents, the PolyAI team is focused on deeply understanding and learning from them — everything from call routing to how they logged calls once completed — and build technology that can make these agents smarter and faster. To do this, PolyAI is not planning on selling software to call centers but to become the call center itself. Through deep integration with the day-to-day operations, PolyAI aims to become the AI-driven call center that redefines what it means to call customer service, and we are thrilled to be PolyAI’s Series A lead investor for this journey.
¹ “Market Size: Just How Big is the Call Center Industry?”, CustomerServ, 25-Oct-2017.
² “Top 4 Contact Center Trends for 2015”, Fonolo, 28-Jan-2015.