How to get people to like your chatbot

People dislike AI chatbots (vs humans), unless you frame it right - or pass on the savings.

Topics: AI | Customer Experience | Website & App | Ecommerce
For: B2C. Can be tested for B2B
Research date: April 2023
Universities: University of Alberta, Erasmus University, University of St. Gallen, Open University of the Netherlands

šŸ“ Intro

The benefits of using a chatbot to handle customer questions and communications are obvious: lower cost, higher (or even unlimited) capacity, and immediate response times.

But customers don’t feel the same way. Generally, customer satisfaction falls when interacting with a chatbot compared to a human.

It doesn’t have to be this way - new research shows two straightforward ways you can replenish customer satisfaction when using a chatbot - or even increase it.

šŸ“ˆ Recommendation

When offering your customers a chatbot to interact with (vs a human), make clear what benefit that gives them. For example, 24/7 availability, immediate responses, or better responses (e.g. this chatbot quickly resolves 95% of conversations).

People will not be unhappy that you are offering a chatbot instead of human interaction - as they normally would - and might even be happier.

If you can’t offer a clear benefit, you can offer a discount to customers who use the chatbot (vs a human agent) for a similar effect.

šŸŽ“ Findings

  • People are less satisfied with and less willing to recommend a company when service is given to them by a chatbot compared to a human agent, even if they’re given identical service by both.

  • As part of 6 experiments, they found that:

    • When a human-like chatbot was used to take orders at a coffee shop, people were 8.8% less satisfied and likely to recommend the place compared to when a human took their order (fun fact: men were more critical of the bot than women)

    • When people were told in an online experiment that their service was given by a bot, their satisfaction and likelihood to recommend the service decreased 8.46%.

  • The effect restores to a comparable level when people interacting with a bot (compared to a human) are given a discount:

    • There was only a 0.39% difference in people’s views of a chatbot, when given a 20% discount, compared to using a human agent without a discount.

  • The effect reverses if a chatbot provides clearly better service than a human agent:

    • When a bot provided clearly superior service than a human agent (faster and just as accurate and effective), those who interacted with the bot (vs a human) were:

      • 75% more likely to stay with the brand

      • 166% more likely to sign up for newsletters

      • 110% more willing to join a loyalty program

    • When told that a bot would provide responses without a delay and experiencing this themselves, people were up to 37% more satisfied and likely to recommend the company compared to those who dealt with humans giving slower responses

🧠 Why it works

  • We assume there’s a tradeoff for companies between pursuing profit and providing the best service to customers, and that companies are more focused on profit.

  • When we deal with an AI chatbot, we assume the company is using them to save costs, probably at the expense of the service experience.

  • If we receive exceptional service from a chatbot, better than we’d expect from a human, we’re less likely to think the chatbot is being used to save costs.

  • Similarly, if the cost savings from using a chatbot are shared with the customer, through a discount, we see it as a fair exchange, as the company might be saving money by using the chatbot, but isn’t making more profit.

šŸ¤– The Science Says guide to AI best practices

Get the best results from applying AI to your marketing with the science-based playbook of AI Best Practices.

We worked with Stefano Puntoni, Co-director of AI at Wharton Business School and 14 leading scientists in AI consumer research to turn 1,000+ pages of research into 50 practical insights, 3 field guides, and 8 deep dives.

Use the playbook to:

  • Use AI more effectively in your marketing

  • Create successful AI chatbots

  • Design your AI to boost sales


āœ‹ Limitations

  • As chatbots are becoming more common in customer service and rapid advances in AI are improving the abilities of most newer chatbots, people are likely to become more used to them and probably to have more positive views of their experiences with chatbots.

  • The study only focused on two services - ordering coffee and changing a cellphone plan. Other research shows people’s views on bots is dependent on the type of interaction they have, as people respond better to bad news when it’s given by a bot than a human.

  • The research focused on immediate reactions to an interaction - it’s unknown if this reaction would hold as more time passes, or if it would change for customers who interact with the chatbot repeatedly.

šŸ¢ Companies using this

  • Brands in almost all industries use chatbots for their customer service, highlighting their role in providing better, faster service than human counterparts:

    • Apparel brand H&M’s chatbot provides information on the availability and sizing of items, as well as details on delivery options, refund queries and more

    • Office furniture brand Wulff Beltton uses its AI chatbot to provide faster answers to technical questions customers have about their products

    • AirHelp, which helps travelers navigate flight delays and cancellations, is able to collect requests, prioritize their urgency, and assign complex cases to humans

    • Bank of America highlights how their AI chatbot, Erica, ā€œhelps clients to take advantage of benefits soonerā€, emphasizing the superior performance of the bot

  • Instead of offering a discount for using a Chatbot, some companies, including almost all US telcos, charge a fee to speak to a human customer agent for tasks available through their IVR self-service helplines.

  • Pepper, a humanoid service robot, was used in physical outlets by over 2,000 companies for uses including giving visitors directions, working as a customer service agent and as a receptionist. Pepper even testified in front of a parliamentary select committee in the UK on the use of AI in classrooms.

  • Plenty of companies offer chatbots that can be easily integrated into your website, including:

    • Drift which features a conversational AI to enhance the quality of replies

    • Freshchat which can provide support in most commonly used languages

    • Gorgias which includes features to segment customers as well as sentiment analysis

    • CRM makers Hubspot and Zendesk’s chatbots

Supermarket Whole Food’s Facebook Messenger chatbot provides faster, more personalized service than human agents, including giving recipes based on customers’ requests.

⚔ Steps to implement

  • When you use chatbots for customer service, make sure that their service is as good - ideally better - than a human agent. This can include providing faster responses, 24/7 availability or more personalized and responsive service using AI models.

  • When possible, let your customers choose whether they want to speak to a bot or human agent, with the option to switch to a human agent during their conversation.

  • Make sure to highlight the benefits of using your bot versus a human agent before your customer is connected to the bot. For example:

    • Highlight specific queries or personalized information the bot can offer that humans can’t (e.g. a clothing brand offering to build complete outfits based on a single piece of clothing the customer has).

    • Faster response time through the bot (e.g. showing estimated wait times for both human and bot options)

  • You can also pass on the savings of using the bot by offering the customer a discount for using them. For example, if you have a product that is highly personalized or specific to a customer’s needs (e.g. curtains/blinds or upholstery) you can charge less for finalizing an order and giving measurement information through a bot compared to a human agent).

  • Learn more about when - and how - to best use AI chatbots, including situation-specific adaptations to make through the Science-based Playbook of AI Best Practices.

šŸ” Study type

Online experiments and a field experiment (using a Pepper robot at a coffee shop in the Netherlands to take customer orders)

šŸ“– Research

Understanding and Improving Consumer Reactions to Service Bots. Journal of Consumer Research (April 2023)

šŸ« Researchers

  • Noah Castelo, Alberta School of Business, University of Alberta.

  • Johannes Boegershausen, Rotterdam School of Management, Erasmus University.

  • Christian Hildebrand, Institute of Behavioral Science and Technology, University of St. Gallen.

  • Alexander P. Henkel, Open University of the Netherlands.

Remember: This is a scientific discovery. In the future it will probably be better understood and could even be proven wrong (that’s how science works). It may also not be generalizable to your situation. If it’s a risky change, always test it on a small scale before rolling it out widely.