Our Users Forget They are Chatting With a Machine
Do you enjoy taking surveys, does anyone? I know as soon as I open one I switch into a mode of “get through this as fast as possible,” which means my responses are minimally useful and I leave feeling a bit grumpy because I couldn’t speak my mind.
Other so called “conversational data gathering” products fail here. They are just a survey in a conversational form, and people can tell. The result is still survey-quality data that can tell you something is wrong, but not how to fix or improve it.
Our approach mimics an experience everyone enjoys, talking with people, and as a result, gathers much richer insights. To do this we started not with a survey, but a human interview, and then taught our AI to mimic it. Some of the ways we achieve this are by having our Conversational AI notice when a user is struggling and help them out. It usually starts with a person saying something like:
“I don’t know”
“What do you mean?”
Having looked at tens of thousands of conversations I can say this happens about 10% of the time. You ask someone a question and they can’t think of an answer or the question doesn’t make sense to them. Our Conversational AI, like a human, detects this and follows up as a human would. It can clarify the question, try asking it in a different way, or move on if the person does not want to answer. Critically, it also understands when follow-up questions don’t make sense to ask.
Q: “What’s causing the issue?”
A: “I can’t think of anything”
Q: “Hmm, how about what initially caused this issue?”
A: “Oh, it was the customer engagement policy”
Q: “What do you see as the biggest issues in remote work?”
A: “I’m not sure, what do you mean?”
Q: “I’m looking for things that made it hard to work from home.”
A: “Oh, well my internet wasn’t fast enough for video chat, that was very frustrating”
Being aware of what a person does and doesn’t understand can keep them in the mindset of talking with another person, resulting in a more detailed conversation and the person feeling like they were heard.