Sign up for our newsletter and get the latest big data news and analysis.

Interview: Dr. Susan Hura, Chief Design Officer at Kore.ai

I recently caught up with Dr. Susan Hura, Chief Design Officer at Kore.ai to discuss the bank-end work that goes into developing an intuitive conversational AI-supported chatbot. She’ll also dispel some of the myths behind developing and introducing a CAI-empowered chatbot into a business’s digital platform. Whether used out-of-the-box or customized, a chatbot’s design plays a more strategic role than one might think and requires an immense amount of human input to create.

insideBIGDATA: What are the often overlooked backend complexities and intricacies of programming a CAI-supported chatbot?

Dr. Susan Hura: Customers are often surprised by how much work goes into developing these solutions because people rarely take the time to truly understand the capabilities of conversational AI (CAI).  Oftentimes the media sells CAI as magic when it is not; there are complex processes behind the scenes that must be taken up by conversation designers and natural language processing (NLP) analysts to create interactions that seem simple. Any response that the bot gives, someone has programmed into it. 

For instance, machine learning (ML) plays an integral role in developing a CAI chatbot. Machine learning is a technique that allows us to take a large set of user utterances and analyze the range of ways users ask about different topics. The algorithm ensures that the CAI bot will understand that “What’s my account balance?” is the same intent as “Could I get my balance please?” based on the similarity in the way the user phrased the question. The tricky part is when the user asks in a completely different way. As humans, it’s clear that “How much do I have in checking?” means the same thing, but it’s not automatic for a bot–someone has to do the mapping to flag the similarity. 

The reason behind this is that bots don’t speak English–or any other language for that matter. When a person tells a chatbot, “I think my credit card was stolen,” the bot might respond with a kind statement like, “I’m sorry to hear that. Let me help you lock your card so no one else can use it.” On the surface, it might appear that the bot is empathetically understanding the user’s experience, but in actuality, that response was programmed into the solution by a conversation designer and NLP analyst. The more data these experts take in, the easier it is to create a bot that can deliver a natural conversational experience in spite of its limitations.

insideBIGDATA: How intelligent are ready-to-use chatbots? Are they truly intuitive right out of the box?

Dr. Susan Hura: Conversational AI is further along than I thought it would be when I was studying it in grad school. With that being said, CAI empowered chatbots are very simple entities that require a lot of human input. Out of the box, these solutions may seem smarter than they truly are. Most organizations understand that setting up a chatbot takes additional time and assistance. But there’s inevitably a point where the company will want the bot to perform a new task, and the solution is unable to assist. The organization gets frustrated because outwardly, the bot can do all of these other cool and valuable tasks; why can’t it automatically do something new? Well, you haven’t specifically trained it to accommodate these new functions. Organizations need to think of their bot as a new employee. They’d never tell a call center agent to suddenly take on a new range of skills without proper coaching, so why would they assume their chatbot is capable of doing so? People think bots are much smarter than humans when in actuality, they are far less intelligent. 

insideBIGDATA: What elements are crucial to creating a CAI chatbot that enhances a brand’s reputation and improves a customer’s overall experience?

Dr. Susan Hura: It’s easy to believe that CAI technology itself will ensure a successful CAI chatbot, but conversation design truly is the most imperative element. Design ensures that the bot’s sound and feel instantiates brand values and engages end users by creating trustworthy, frictionless experiences. Lots of people believe that conversation design is just about writing prompts that sound good, but in reality, it’s focused on ensuring that the bot meets the needs of end users. These elements are what drive business outcomes such as increasing self services rates and reducing operational expenses. There has to be a compelling reason for users to want to engage with these technologies and if the design is poor, they simply won’t use them.

About the Interviewee

Dr. Susan Hura is a conversational user experience designer and strategist with 35+ years of experience in linguistics, user-centered design and speech technologies. She is the Chief Design Officer at Kore.ai, a leading conversational AI software company, and has previously worked on IVR and voice user interface implementation for companies like Lucent Technologies Bell Labs and Human Factors International; she has also founded Banter Technology, discourse.ai, and SpeechUsability. She holds a Doctorate in Linguistics from the University of Texas at Austin, and BA in Linguistics from Ohio State University.

Sign up for the free insideBIGDATA newsletter.

Join us on Twitter: @InsideBigData1 – https://twitter.com/InsideBigData1

Leave a Comment

*