Mobile Messaging and Bots: The Next Frontier of Customer Service
By Judith Aquino, 1to1 Media
Chatbots are on the horizon; here’s how to incorporate these technologies into a business’ service arsenal.
Virtual robots or chatbots powered by artificial intelligence (AI) are transforming customer service. Domino’s was one of the first companies to experiment with AI technology by letting customers order pizza via a pizza emoji that’s tweeted to @Dominos.
Pizza Hut has also said it will be launching a social ordering platform in August that allows customers to place orders for pizza using Facebook Messenger’s chatbots. In addition to processing orders, the bot will also answer questions about the food and Pizza Hut deals and even offer menu items that are localized to specific stores.
Additionally, Macy’s is testing “Macy’s On Call,” a mobile Web tool that allows customers to interact with an AI-powered shopping companion on their phones. The tool, which is being piloted at 10 Macy’s stores, taps IBM Watson, via Satisfi, an engagement platform, to answer questions like “Where are the ladies’ shoes?”
These bots offer new ways to ask questions, make purchases, and more within a mobile device. But where do chatbots fit in a customer service strategy?
Donna Peeples, CCO at Pypestream, a mobile messaging platform provider, compares chatbots to the early stages of IVR systems. “In some ways, chatbots take the shape of a digitized IVR,” Peeples says. “You’re given a set of choices and you’re not interacting with a human. At its most basic level, it’s a digital IVR.”
As an example of how businesses could use chatbots, Peeples points to a client, Lynx Services, a subsidiary of Solera Holdings, Inc., which provides software and services to the automobile insurance claims industry. The company is in the process of implementing Pypestream’s mobile messaging solution. As part of the agreement, Pypestream will develop automated chat capabilities to be used by auto glass repair/replacement servicing companies to initiate and complete transactions.
In other words, an auto glass repair shop may send a text message, such as “I’m having trouble with ClaimPoint” to Lynx over Pypestream’s platform. On the backend, the chatbots will use a guided decision tree model with dynamic routing where customers are given options within the message stream.
The goal is to complete transactions in a “more streamlined and efficient manner” and help expedite auto glass repairs, explains John Wysseier, managing director of Solera Insurance Services. “By automating answers to common questions via the Pypestream messaging platform, we’re able to improve the customer experience by providing real-time 24/7 support, while also manage operating costs through efficiency,” Wysseier says.
At the same time, agents will be available to answer questions if the chatbot isn’t sufficient, Wysseier adds. Pypestream has also developed an integration between the chatbots and Lynx’s back-end system that allows agents to review the message stream and the customer’s history to provide knowledgeable service.
A Chance Not to Repeat History
Brian Spencer, general manager of contact centers at Mitel, a telecommunications solutions provider, agrees that there are parallels between the development of chatbots and IVR systems but notes that companies should learn from the mistakes of IVR systems.
When considering the shortfalls of early IVR technology, “we have an opportunity to not repeat history and use chatbots to enhance the customer experience in better ways,” Spencer says. For example, IVR systems were built to focus on efficiency by deflecting or reducing calls to the contact center but the caller experience was “often impersonal and frustrating, especially if you had to listen to a long menu before getting to the right option,” Spencer notes. The result was low customer satisfaction.
Chatbots are different. In addition to providing fast and efficient service, there’s room to add some personality to the conversation. “That’s what a lot of self-service technology is missing: The experience isn’t pleasant,” Spencer notes. “And adding a few ‘fun’ statements while providing efficient customer care can make a difference.”
We’re already seeing companies attempt to give virtual assistants a personality with humorous statements. For instance, users who ask Siri what her favorite movie is will usually get this response: “I’ve heard that Blade Runner [a dystopian film where humans are hunted down by robots] is a very realistic and sensitive depiction of intelligent assistants.”
Of course, as Tay, Microsoft’s artificial intelligence chatbot demonstrated, teaching a robot the art of conversation is complicated. Teaching a chatbot to respond conversationally isn’t superfluous either, Spencer adds. Unless a chatbot understands conversational nuances, it is limited to basic interactions and services.
“You’re not going to see many chatbots be successful at upselling anytime soon because they would need to understand various parts of a conversation and be able to make that transition to discussing other products smoothly,” he notes. But regardless of being able to upsell products and services, the value of chatbots lies in the potential for engaging customers more efficiently and effectively than an impersonal IVR system.
Letting Chatbots Grow Up
Although there’s a lot of excitement about chatbots, companies also need to give the technology time to mature, says Tim Fujita-Yuhas, director of product management and new product strategy at OpenMarket, a mobile engagement platform for the enterprise. “If you think of the voice and IVR space, it took about 15 years to get to where we are now where in addition to pressing keys or saying “1” some systems let you explain in a few words what you’re calling about,” Fujita-Yuhas says. “Chatbots will probably advance faster than that, but it still takes time.”
For example, one area where chatbots fall short is sentiment analysis. Anyone who has misinterpreted a sarcastic text message knows how difficult it can be to understand intent, never mind a robot being able to do the same. A chatbot may struggle to understand language quirks as well. When processing dinner reservations, it might not know that “me and my wife” is the same as “party of two,” Fujita-Yuhas points out. “The bot needs enough ‘smarts’ to realize that you’re referring to the same thing, which is where natural language processing and machine learning come in.”
Of course, chatbots need access to a lot of data to be useful, but that doesn’t mean only data behemoths like Google, Apple, and Facebook will succeed, Fujita-Yuhas maintains. “The market has already demonstrated that you don’t have to be Google or Apple to have a sufficient data set,” he says. “There are also opportunities for vertical markets like financial services, which has plenty of user data.” A bank, for instance, could work with a vendor to craft appropriate responses for a chatbot based on data from an IVR system and other knowledge bases.
And as chatbots advance and become standardized, we may reach a point where “chatbots understand each other,” Fujita-Yuhas says. “In fact, some SaaS providers are already looking to offer a chatbot as an alternative to the API.” Indeed, a future of integrated chatbots is more likely to happen than one company holding the reins, he adds. “I think we’ll reach this future faster than if we were to wait for Google or Facebook to develop a universal assistant for every business and personal use—that’s a much harder nut to crack.”