download pdf

Chatbots – the new sales agents?   
Advances in artificial intelligence and natural language processing are giving rise to a new kind of service encounter: chatbots. These humanized interfaces allow brands to provide 24/7 support while automating major areas of customer service operations. Recent industry reports indicate that more than 80% of all businesses are considering integrating chatbots within the next five years. Many service providers are already using chatbots, but most applications have focused on cost saving and automation to partially replace human service encounters. However, rather than a mere cost-saving opportunity, chatbots might bring a myriad of advantages in automating customer-company interactions. One critical question for business is:  Can chatbots augment the corporate sales force? More directly, are chatbots able to sell?  
Through a series of field and laboratory studies, we found evidence that chatbots can indeed alter consumer preferences and purchase decisions. What consumers enjoy in particular is the ability to engage in natural dialogue and even to connect to the chatbot’s “personality.” We gained insights on the “optimal mechanics” of such interfaces and found evidence that incorporating chatbots into consumers’ shopping processes promotes more intimate consumer-brand relationships, greater trust, and can be used as a powerful opportunity to upsell.

Sales automation with chatbots: The critical role of turn-taking and personalization
First, a chatbot’s ability to engage in more natural dialogue compared to traditional media is a critical ingredient for fostering a more enjoyable, trusted user experience. Research in human-to-human interaction has shown that turn-taking during conversations makes the conversational partners feel closer and that they like each other more. Thus, turn-taking is key to fostering trust between interaction partners.
The second important driver for building closer consumer-brand relationships is personalization. Both in terms of the chatbot being viewed as more human-like compared to a traditional interface as well as in its ability to adapt to specific consumer characteristics. People see and treat machines like other humans when these machines possess more human-like features. Brands and products with more human characteristics also lead to more positive brand association and more intimate brand relationships, according to prior research.  However, apart from the obvious visual characteristics of a chatbot, it is not very clear what makes an interaction with a machine “feel human” and personal, and how this ultimately affects consumer purchase decisions. A condensed summary of our key insights is presented in Box 1.

The powerful impact of chatbots on upselling potential
We found evidence that a more humanized chatbot is able to build a more intimate and trusting brand relationship, is perceived as more enjoyable, and increases a company’s upselling potential. Specifically, in the context of our car rental bot, we found that consumers were almost twice as willing to trade up to more costly options and add-on services when they were offered by a humanized chatbot (see Figure 1). This effect increased even further with more personalized interfaces that matched fundamental consumer characteristics, according to additional experiments.

The powerful impact of chatbots on trust 
In our work with a financial advisory chatbot, known as a “robo advisor,” we found astonishing results with people putting too much trust in recommendations provided by robo advisors. We went as far as proposing objectively wrong portfolio recommendations to knowledgeable investors. Despite the objectively wrong advice and additional warnings to investors, they were three times more likely to accept an incorrect recommendation from a humanized robo advisor than from a traditional web interface. Figure 2 clearly depicts the greater trust in the robo-advisors and substantially fewer rejections of the recommended portfolio (red dots indicate rejection of the recommended portfolio).
In another study, allowing consumers to customize a dress shirt via a humanized chatbot rather than a traditional ecommerce website also led to a significant increase in customers’ willingness to pay for the custom-made dress shirt.

Designing better chatbot touchpoints
The implementation of bot interfaces varies tremendously in current industry practice. They range from human-like bot interfaces to merely presenting a brand logo or a digital avatar. Some applications provide a maximum amount of information with limited turn-taking between the user and the interface, others offer only short pieces of information and require more turn-taking. Often, the implementation is a function of the default settings provided by chatbot providers and platforms. Instead of simply implementing the default option, companies should consider very carefully how the specifics of the chatbot interface might affect the user experience. Our work provides clear guidelines for developing better customer-bot touchpoints ─ for both customers and companies. Four effective design “mechanics” may serve as a guideline.

• Make the interaction feel natural
  Our studies reveal that greater turn-taking between the user and the interface is an important lever for generating trust. This critical trust-building mechanism has positive downstream consequences on brand perception and upselling potential. Managers should move beyond “out-of-the-box” chatbot services. Simple mechanics such as increasing the frequency of interactions leads to greater trust and a more enjoyable user experience. Yet, managers should carefully monitor and detect the flow of the conversation and provide witty answers to random and unrelated questions. Should the conversation with the chatbot get stuck nevertheless, companies should have hybrid systems at the ready that quickly connect users to a human representative.
   
• Adapt chatbots to consumer characteristics
  While we have only investigated a few ways of personalizing chatbots – based on consumers’ gender, using similar-sounding first names, mimicking the type of language used – we see great potential for brands in providing a personalized chatbot experience. Our findings suggest that personalizing chatbots to basic consumer characteristics increases trust perception and improves the perception of intimacy between the customer, the chatbot and, ultimately, the brand.

• Adapt chatbots to your brand
  A chatbot should not only adapt to consumer characteristics, but it should also reflect the key characteristics of the brand. While we did not investigate this aspect in this research, we expect that chatbot “personalities” can be specifically tailored to the desired brand personality. Not only structural aspects of a chatbot interaction like the extent of turn-taking, but also more contextual factors such as the personality expressed through the type of language will affect both brand perception and consumer choice. Thus, designing the personality of machines more generally will be an exciting avenue for future research and brand strategists alike.