Credible Chatbots Outperform Humans

First impressions make a big difference in business. It turns out that’s not only the case for human-to-human encounters, but also for interactions with chatbots.

Artificial intelligence (AI) chatbots are four times more effective at selling products than inexperienced workers, according to a new study. However, if customers know the conversational partner is not a human, they are curt and purchase less. And sales rates decline by more than 79.7%.

See also: With DIY and Conversation-as-a-Service, Chatbots Are Here to Stay

The study, conducted by researchers from Temple University Sichuan University, and Fudan University, targeted 6,000 customers from a financial services company. They were randomly assigned to either humans or chatbots, and disclosure of the bots varied from not telling the consumer at all, to telling them at the beginning of the conversation, after the conversation, or telling them after they’d purchased something.

The main reason for the decline in sales when customers know they are dealing with a chatbot is that they perceive the bot to be less knowledgeable and less empathetic, according to Xueming Luo, one of the researchers in this study and a professor at Temple University

What Makes a Chatbot Successful?

So how does a chatbot come off as knowledgeable and empathetic to put customers at ease? Simpler chatbots will simply not do the trick.

For example, many voice and text chatbots listen for or look for key words to guide the conversation. The customer mentions “account balance” or “pay bill,” and the bot knows which road to go down with its replies. The problem with this approach is that we’ve all had trouble with chatbots that do not understand or miss the meaning of what we are saying. And most certainly do not pick up on our sentiments (hint: a building frustration during an exchange).

Continuous intelligence can help alleviate these problems. Chatbots that use natural language processing, speech recognition, and artificial intelligence can be designed to get deep insights into customer content in real-time. In that way, an intelligent bot could interact with the customer just as a live agent would. An even better solution is one that leverages sentiment analysis to determine whether a customer is happy with a product or if a web site visitor is struggling to find information about their account.

An intelligent chatbot operating in an insurance company might engage a customer calling into a call center or one surfing the company’s web site. Acting as a virtual agent, the bot could ask the customer — via synthesized voice on a call or text on a Web site screen — what they need and either answer the question or guide the person to an appropriate resource.

For example, an intelligent chatbot might ask what type of insurance a customer is looking for, help the customer learn about and select options (premiums, coverage amounts, deductibles, etc.), and offer add-ons such as bundling home-owners insurance with their auto insurance.

Such capabilities will help businesses offer better services to their customers. The customers will accept intelligent chatbots only if they are indistinguishable from humans. As the use of continuous intelligence becomes more pervasive, businesses will have a way to make their chatbots more human-like than ever before.