Uncanny valley: How to avoid creepy AI in customer service
Last week, I was chatting online with customer service, and my agent, Barb, was so friendly and helpful. Our chat exchange was moving along, but then something strange happened. My lovely conversation pivoted and she sent me a series of messages that felt like a canned, impassive set of instructions. Gone were the expressive comments and banter. Initially, I felt a little strange about this cold response. What happened to my friendly Barb? Her profile photo was still smiling at me, but her tone no longer matched that smile. Within a few seconds, I realized Barb was not real!
Creepy AI and the uncanny valley
Artificial intelligence (AI) is slowly becoming integrated into many facets of our lives, and most of the time it has become so seamless that we don’t even realize that AI is churning away in the background. This phenomenon is called creeping AI. Because AI usually solves problems and makes everyday tasks faster and more efficient, most don’t mind this creeping effect. That is, until something seems off. I was happy with Barb when she was helpful. I was comfortable when I thought she was human. Oddly, I was also fine with Barb once I realized that she was a bot. It was that moment in between that caused unease. That moment of unease is the theoretical place called the uncanny valley. The term describes the point at which an object resembles a human being to the point of causing a person psychological discomfort or eeriness. The humanoid or bot is not quite perfectly human-like and we can sense slight discrepancies. In the uncanny valley, AI becomes creepy AI. In the customer service world, a creepy bot is equivalent to a rude human agent. Both are a huge turn off for customers.
Avoiding uncanny valley – the Creepy AI
With the demand for personalization in customer service increasing, the use of AI is essential for creating efficient and scalable solutions. Leveraging AI is a great way to build an intelligent, conversational user interface (UI) that improves customer service while meeting business goals. Digital agents using AI have the ability to perform tasks such as assisting customers with questions or problems or giving product suggestions based on customer buying habits. So, how can AI be incorporated seamlessly into customer interactions while avoiding the trappings of “creepy AI”?
Instead of trying to make the agent increasingly more human-like and risk falling into the dreaded uncanny valley, embrace the bot identity and clearly communicate that identity to customers. “Hi! I’m your digital agent, Barb, here to help you.” Being honest with customers creates a relationship of trust and patience from the customer.
Studies show that AI seems creepier when elements of the design are inconsistent. Barb’s mix of human-like characteristics and bot-like characteristics caused uncertainty and unease, as did the transition from human-like to clearly bot behavior. This inconsistency stands out to customers and runs the risk of an AI solution appearing disjointed.
It takes time for humans to adapt to new technology. Some people may not initially like conversing with AI bots, but once users become accustomed and see the benefits, the dislike will fade. Remember how adverse people were to accepting cookies during internet searches? That faded when the benefits to cookies were realized.
Un-Creepy Bot Creation
Once your business determines that developing an AI bot or conversational UI is the right solution for better engagement with your customers, it is important to take the time to guarantee that this new digital assistant represents your business properly. There are many tools available for developing bots, but it is important to plan your strategy before embarking upon the development phase.
One important component of successful AI is giving the machine appropriate data that it can then use to predict and perform. Determine what data will be fed to the bot to achieve the results that best assist customers.
A bot should match the overall company voice and be an integrated part of the entire customer journey. Take the time to research customers’ needs and then map out the all of the possible interactions between the bot and the customer. The content for the bot should be written in a way to give customers the information they need in a clear, concise way, but also to engage in a friendly manner. Don’t forget, though, that transparency is important for maintaining customer trust. Conversation should be human-like, while also acknowledging the bot’s existence and limitations.
QA and Test
Before introducing a bot to the world, it should undergo extensive QA and testing. In addition, AI should be constantly monitored to ensure that both inputted data sets and outputted results are following protocol, avoiding bias, and fitting into the overall customer care strategy.
As AI creeps into customers’ lives, it is important to make sure that the contribution of AI is positive, meaningful and relevant. It’s possible to avoid agents like Barb drifting into creepy AI territory with better design and testing. As long as a business is fitting AI into its overall customer service strategy appropriately and following through with constant monitoring and maintenance, AI tools can be powerful for improving customer engagement and overall business goals. Customer demand for personalization is increasing. Make sure your bot can keep up.
Interested in conversational AI for your organization?
Check out how Logic20/20 can help.
Liz Gillette is a director of Digital Transformation at Logic20/20. She is responsible for overseeing complex engagements, ensuring delivery excellence focusing on best practices and standards, and growing talent within the technology organization.
Follow Liz on LinkedIn