Four Challenges to Overcome for AI-Driven Customer Experience

Are you ready to turn your company’s customer service over to AI-powered virtual agents? Whether your goals include creating a fully digital business, improving customer experience, or cutting costs, virtual agents and automation offer many benefits. But, as with any technology, making the switch isn’t trivial.

Consider the experience of Mark Baylis, vice president of customer service and digital customer engagement at Optus, Australia’s second-largest telecom operator. He believed in the potential advantages of using virtual agents for customer service, but he needed to convince the company’s leadership team that the investment would make sense.

Baylis planned to have virtual chat agents handle half the incoming queries, which would cut costs by 50%. But the cost savings alone weren’t enough to overcome corporate resistance. Baylis recognized that this shift was not just about the customer service division but about the company’s larger digital transformation. By positioning the switch as part of the larger transformation effort, the investment in virtual agents became more persuasive to management. It also helped that virtual agents could handle more customer requests at peak hours and improve the quality and consistency of Optus’ customer experience.

By focusing on strategy, not just cost savings, Baylis was able to win over Optus’ CEO and C-level decision makers at a meeting in 2017. One person at that meeting described it this way: “The general reaction was excitement. It was in line with trends in the market.” Leaders at the meeting recognized the need to adopt the new technology and master it to prepare for the future.

In many ways, what happened at Optus is typical for companies that are considering AI-enabled technologies for their customer-experience strategy. Proponents must overcome four types of hurdles: economic, technical, political, and cultural.

The economic hurdle: demonstrating improvements to the bottom line. If your company currently has more than 1,000 phone or text chat agents answering customer questions, then it’s a candidate to benefit economically from a virtual agent system. Because they scale without the need to hire more people, virtual agents can pay for themselves — provided there are enough customer service intents that are easy to automate.

For example, at Dish Network, where customers initiate 6 million chats per year for sales and support, virtual agent automation generated millions of dollars in annual savings. And at cable operator Charter Communications, AI-based customer service enabled the company to automate 50% of customer service chats.

You can also make the economic case by accelerating sales as you reduce costs. At Dish Network, a surge in demand for pay-per-view fights was challenging sales staff; switching to virtual agents enabled the company to recognize more than $6 million in pay-per-view fight revenue.

Consider economic benefits in the broadest possible perspective. We’ve seen companies justify the decision to implement AI by citing not just cost cutting but sales increases, customer service speed improvements that improve retention, and even delivering on promises to shareholders.

The technical hurdle: ensuring that systems are ready. Virtual agents need to answer questions like “What’s my balance?” or “What’s the status of my order?” That requires access to corporate systems, which only happens when IT agrees on priorities with the lines of the business directly implementing virtual agents.

Sometimes, technical obstacles are insurmountable. At one large online retailer, a “decisioning engine” system was responsible for resolving customers’ queries. When the company tried to hook up a virtual agent system to it, over 95% of the queries returned a response of “transfer to agent.” Because no single department would take responsibility for the decisioning system, the problem was unfixable. The virtual agent system never launched because the rest of the corporate technology wasn’t equipped to support it.

For a customer service group to avoid such problems, it must first review the breadth of queries coming in, classifying them by intent — the types of problems that the customer is seeking to fix. Success is most likely if a majority of the queries come from a common set of intents, such as “check balance,” “return product,” “reset password,” or “increase credit limit.” If the queries are all over the map, automating them is less likely to succeed.

A company is a good candidate for service automation if it has existing APIs (application programming interfaces) or can build new ones that return necessary data from corporate systems with subsecond response times. Those systems need to be extremely reliable and capable of scaling up in times of peak demand. IT staff members often already share these goals for corporate system improvements. Virtual agents need such access because when it comes to delays in returning information, customers are typically less patient with automated agents than they are with human agents.

The political hurdle: winning over management. Virtual agents can affect multiple parts of the company, including its brand. As a result, the decision to commit to virtual agents typically involves not just the head of customer service, but an array of C-level decision makers, including the chief operating officer, the CIO, some of the CIO’s direct reports, the head of sales, and often the chief experience officer (assuming the company has one). Because the virtual agent can become a highly visible part of the brand, chief marking officers and heads of product management may also become involved. At Bank of America, for example, the company’s virtual agent Erica appears in ads and marketing emails because it reflects the company’s responsive and technologically advanced culture.

The best way to corral a group of decision makers who are this interconnected and influential is to frame the project within an existing strategic initiative at the company. At Optus, as at many companies, this was the corporate push for digital transformation. Digitizing and automating customer service creates a desirable result: It brings a highly human-centric and sometimes inconsistent part of the business under the digital umbrella.

Digital transformation isn’t the only justification. At one travel company, virtual agents got the green light because the company had promised cost savings in the wake of a merger. At another, virtual agents allowed an IT department to handle more queries without expanding beyond its physical space constraints.

The cultural hurdle: getting past human-centric prejudices. In the end, much of the resistance to virtual agents comes down to a simple question: Is the company willing to let machines talk to customers, or is it a job that only people can handle? That’s a cultural and emotional challenge, not a technical one.

The surprising reality is that many customers would prefer to talk to a computer. Just as consumers became comfortable with websites and apps, they are likely to become happier with Alexa-like responses from machines — so long as those machines can understand the problems and supply the right answers.

As we wrote in our last article, managers can become more comfortable when virtual agents and human agents work side by side to solve customers’ problems. This can help enable policy shifts that lead to cultural shifts. For example, at one credit card company, policy demanded that customers had to talk to a person to file a claim for alleging an unauthorized purchase. At SiriusXM Satellite Radio, rules specified that customers couldn’t cancel service without talking to a person. In both cases, the companies revised their rules once they realized that virtual agents could be as efficient and secure as human agents.

IT’s not-invented-here attitudes can be another cultural obstacle. Building virtual agents from scratch requires not just AI mastery but knowledge of the most common queries in the company’s vertical (for example, travel, banking, or retail). The large vendors in this space already have that knowledge and have implemented similar systems. Quick and successful pilot projects built in collaboration with knowledgeable vendors can help reduce this cultural hurdle.

How to succeed with virtual agents. Examining these four types of hurdles reveals what it will take to make virtual agents successful. Virtual agent proponents must demonstrate both strategic fit with the company and the readiness of technical systems. They must tie virtual agents to corporate priorities like digital transformation. And they must start by rolling out pilot systems that demonstrate both customer acceptance and efficiency advantages.

In the end, companies will figure out that they need to get started in implementing automated conversational interfaces to their customer service. Because, like the web and mobile apps before them, conversational interfaces like Amazon’s Alexa and Apple’s Siri will eventually raise customers’ expectations of every service interaction with every company.