Customer Interactions Need to Be Conversations



In an age of smart speakers, intelligent appliances, and message-based marketing, it is increasingly clear that the optimal customer engagement model is a conversation. Ideally, those conversations are persistent, asynchronous, and span every device, communications modality, and the passage of time. I'd go even further and say that all investment in marketing, advertising, and customer service technology must coalesce around the ability to give customers control over their conversations. They should be effortless, friction-free, and quickly resolved.

This suggestion used to be a pipe dream, but in this age,  Amazon has made it possible to reorder toilet paper or trash bags by barking at an intelligent speaker, and Comcast's technical support can be invoked through a voice remote. I think you get the idea. Customers are taking control, but, depending on the industry and context, it can still get a little messy.

All of your company's interactions with customers must be conversational. Conversations enable customers to communicate their intents or objectives using their own words. Hence, there your company needs to invest in some flavor of natural language understanding (NLU). Both parties to a conversation follow a set of unspoken rules that the best contact center agents have long understood, and the bulk of automated chatbots, virtual agents, and intelligent assistants are assimilating, thanks to judicious implementation of machine learning (ML).

A Single Source of Truth Rendered Over Multiple New Channels

As displayed in the graphic below, companies have two major investments to make in the age of conversational commerce. The first fall into the category of single source of truth and describe information, images, records, and other structured and unstructured data in the form of product descriptions, collateral, transaction records, inventory databases, and the like. The other objects of investment have long been called channels, but more recently can be rendered as the latest shiny object among conversational user interfaces. Here’s where speech-enabled interactive voice response (IVR) systems and mobile apps have been joined by chatbots, virtual agents, and Messenger bots.

Though the rules for rendering correct answers and proper presentation vary across modes and devices, the content offered must be accurate and consistent.

What Makes a Good Conversation?

The most obvious feature of an efficient conversation is turn-taking. Neither live agents nor automated counterparts do very well when they interrupt or talk at the same time as customers. That's easy. But the definition of efficient conversations, especially when thinking of people talking to computers, has been very dynamic and now encompasses context and state awareness. At this point in time, systems that prompt live agents or their automated counterparts to provide relevant responses to a customer's concern, query, or instructions have access to a broad set of data and metadata about the individual customer and take information, such as location, age, gender, transaction history, recent activity on a web site or search engine, and current sentiment indicators to create the appropriate responses or reactions.

As conversations take multiple turns, there is a high probability that multiple channels have been used and that conversations become asynchronous. It used to feel like a gimmick when a contact center solution provider showed the ability for a busy parent to start a conversation by initiating a query though a company's app on her smartphone, only to have a chatbot suggest a transfer to a live agent to answer a question or fulfill a transaction. Now it's fairly routine, and it has exposed some new gray areas in the CX domain where brands seek best practices.

Here are some other things to consider:

Should a company make it evident when a customer is interacting with a bot versus a live agent? Answer: Yes. In almost every case a customer should be aware that he or she is engaged with an automated agent.

How can a transfer from chatbot to live agent be handled conversationally? Answer: As much context as possible should, including chat transcripts, should transfer with the conversation.

What if a live agent is engaged during the conversation to better understand meaning? Answer: This will become more common and make the distinction between live and automated service less relevant.

We've seen many trademarked initiatives surrounding a conversational engagement model, such as 1:1 marketing, personalization, mass customization, and customer-centric communications come and go yet always stay present and relevant. The move toward conversational commerce is coming to fruition because of the popularity of the Facebook Messenger app. It has enabled companies to be "friended" by each of its 1.6 billion users.

Now voice-first devices like Amazon Echo and Google Home are about to be joined by Apple's Homepod and a variety of other intelligent speakers that spark conversational commerce. Automobiles are next, making it an imperative for every company across fast food, finance, consumer electronics, healthcare, and even government services to offer a conversational connection to prospects and customers.


Dan Miller is founder and lead analyst at Opus Research.