The Center for Data Innovation spoke with Bart Lehane, chief technology officer and founder of EdgeTier, a company based in Dublin, Ireland, which uses AI, analytics, automation, and machine learning to generate responses to customer queries. Lehane discussed how AI can enhance human-centric customer service through more personalization and greater accuracy.
Eline Chivot: What has led you to set up EdgeTier, and what challenges did you encounter when creating the platform?
Bart Lehane: Our aim in EdgeTier is really straightforward: We want customer service to work better.
As a company, EdgeTier originally started out doing pure data science consultancy for large customer support organizations. Our heritage is in applying advanced machine learning to simplify complex problems. We found that many customer support organizations were under-utilizing the rich datasets they had available and struggled to get a strong understanding of how they could improve the efficiency of their teams and provide better customer care. We would analyze huge datasets from our customers and, while doing this analysis, we noticed a set of common patterns across virtually all industries.
Firstly—and this is a really under-appreciated fact—customer care is hard. Much harder than most people give it credit for. There is a myth out there that some black box of AI will do away with customer care for good, and there have been some wild predictions around the percentage of customer conversations that will be handled by automated bots. While chatbots are great for handling simple queries, once you look at the data, you’ll see that customers generally don’t contact customer care over simple things like checking opening hours or adding something to an order. Customers generally prefer to self-serve and each company has websites, portals, or apps that allow customers to solve all the trivial problems themselves. Customers who contact customer care are typically doing so because their problem isn’t that simple, or they are distressed, frustrated, or confused about something. So, while bots and self-service absolutely have a role to play in customer care, they are nowhere near being able to reliably handle complex customer queries.
Secondly, customer care agents spend an enormous amount of time doing things that don’t impact or improve the customer experience. If you sit beside a customer care agent with a stopwatch (which we do!) and time the different tasks they perform, you’ll see that they spend the vast majority of their time performing repetitive tasks that aren’t actually beneficial to the customer. Things like finding customer information, copying and pasting text, and drafting responses end up taking most of their time with little time left for the agents to tailor the response to the customer or ensure that everything is correct.
Thirdly, there is an absolute treasure trove of information stored in customer support queries that is usually under-utilized. Most organizations of scale have millions of pieces of customer feedback locked away in their customer service software. The data is often hard to get at, or, when it is accessible, it’s difficult for customer care teams to do any deep analysis on the data.
We built our agent-assist product, Arthur, to address the issues we saw. It uses AI and automation to assist customer service agents in order to dramatically increase the speed and quality of their customer queries. Arthur also provides contact center managers with the insight and controls to manage their organization.
The biggest challenge in building the product is maintaining focus on solving the right problems. When most people think of AI and customer care, they automatically jump to chatbots. While chatbots can be effective in some situations, our analysis shows us that there is a much larger return on investment (ROI) to be obtained by improving the overall contact center performance rather than deflecting contacts. This gives you the added benefit of higher quality customer care also. Maintaining a data-driven approach to adding features to the product allows us to ensure that the problems we solve are real ones.
Chivot: Can you tell me more about how your AI customer support assistant “Arthur” works? What does it do better than humans?
Lehane: Arthur blends AI and automation with intelligent software that is used by real people.
On the surface, Arthur is like any other modern customer support software. Customer service agents log in and use it to respond to customer queries. Team leads and managers use it to track performance and organize their teams. However, under the hood, the Arthur system works incredibly hard to assist agents responding to those queries and to give managers the information they need.
Arthur’s AI engine scans and analyzes all text-based messages to understand what the query relates to. It then searches for and retrieves all relevant information relating to the query and the customer. Once Arthur has that information, it does three main things. Firstly, Arthur will figure out who the best-placed person to handle this specific query is based on the available agents, their skills, and the customer’s information. At this point, Arthur may also decide to automatically respond to the customer without any agent help—although we find that many complex queries will require human interaction. Secondly, Arthur will guide the agent through responding to the customer. Arthur does this by presenting and filtering relevant information to the agent, drafting responses on the agents behalf, warning the agent of suspicious activity, and prompting the agent with suggested text. Thirdly, Arthur will automate any repetitive task that the agent performs such as looking up customer information, prodding inactive customers on a chat, storing a handled query in a customer relationship management system (CRM), etc.
For customer service agents using Arthur, the flow of handling a customer query changes quite a bit and Arthur allows them to focus on tailoring or customizing the message back to the customer rather than performing repetitive and manual tasks.
The Arthur system also has comprehensive real-time and historic reporting and analytics to ensure that team leads and managers can understand exactly what’s going on in their customer care organization.
Chivot: Can you tell me about the concrete benefits and return on investment for your enterprise clients?
Lehane: The most concrete benefit is the reduction in average handling times. Arthur will reduce the average handling time of emails and chats by over 40 percent, while also increasing customer satisfaction scores (i.e., net promoter score (NPS) or agent satisfaction). This allows a customer support center to significantly reduce the costs of handling a customer query, while also getting happier customers. The ROI is significant, as we can almost halve the costs of handling customer queries.
There are also “softer” benefits that are a little harder to quantify. For example, customer support agents love the technology because it takes a lot of the boring tasks away. The best customer service agents are the ones who genuinely want to help customers, nobody goes into work in the morning to copy and paste text a thousand times! Arthur ensures that customer service agents can focus on doing what they do best: Helping customers. So Arthur can improve retention while also reducing training times.
Another concrete benefit from a recent feature addition is around automatically translating customer queries. Many global contact centers have trouble recruiting multilingual staff. We added auto-translation capabilities so that, in cases where there are no native speakers available, Arthur will automatically translate all messages between the customer and the agent, allowing both to speak in their native language. With one customer, this has worked so well that 34 percent of chats are now being translated live, with no drop in customer satisfaction scores.
Chivot: How does “Arthur-ficial intelligence” manage to strike the subtle balance between increasing efficiency, speed, and accuracy by assisting human operations on the one hand, and retaining humans’ personal touch and complexity of communication on the other?
Lehane: Getting the balance right between AI, automation, and humans is crucial. Too much AI and automation, and you run the risk of pushing customers away and losing the personal touch. Too little, and you will fall behind as your costs will be too high. We try to get the balance right by following a simple rule: Let computers do what computers do best, and let people do what people do best. The nice thing about this rule is that the things that computers are good at tend to be things that people are bad at—and vice-versa.
Computers are phenomenally good at handling repetitive tasks consistently well, so it makes sense to get them to automate common repetitive tasks. Advances in AI also mean that it is possible to get an extremely accurate understanding of a customer’s desires from their written text (our AI models typically operate at above 95 percent accuracy).
People, on the other hand, shine when it comes to things like social and emotional intelligence, empathy, sympathy, humour, etc. It’s very difficult for an AI model or piece of software to reliably emulate these kinds of human traits. Dealing with, for example, a distressed or frustrated customer is just not something that you want a bot to be handling.
Most customer-focussed companies realize that retaining this personal element is extremely important, and so the approach of supercharging customer service agents and managers achieves this.
Chivot: What does the future hold for AI technologies in the contact center industry?
Lehane: There is no real ambiguity here. AI will be used in every single contact center in the future. The benefits of using AI and automation are just too big for it to be ignored. Contact centers that don’t embrace AI and automation technology will simply be left behind by their competitors who can deliver better customer service at lower costs.