The bright future of chatbots
An interview with Kindly’s chatbot creators
Text by Hans Joakim Lien and Siv Fjørtoft. Photos by Siv Fjørtoft.
A few times a year, we like to gather all our customers to talk about what’s new with our product Kindly, and where we’re headed in the near future. Our user forum is an excellent chance to catch up with the people who actually use Kindly on a regular basis.
This time around, we got to chat with the people working on chatbots for DNT (The Norwegian Trekking Association), Adecco and KF (a supplier of knowledge and information products for municipalities in Norway).
Firstly, we wanted to find out what kind of experiences our customers have had with Kindly and with using chatbots.
Mikael Svensson from Adecco is in charge of their chatbot, Jobbot, which helps users apply for jobs. He says they’ve had good experiences with their chatbot.
“We work with recruiting people, so it’s hard to find concrete numbers when it comes to measuring conversions. But we see that a lot of people are using the chatbot, and they often tell us when they come in for interviews that they’re pleased with how the chatbot performed.”
Heidi Grindstad, leader of membership development in DNT, says they’ve only launched a chatbot for the website UT.no so far, but their experiences are very good.
“We’ve managed to teach our users how our new functionality works, and that’s our best result so far,” she says.
“We’re about to start creating the chatbot for DNT and offered most of our organizations to use it on their DNT pages. And almost everyone we’ve talked to is very open to the idea. We think the bot can contribute even more, especially for our support staff. We’d like to see if we can receive more membership questions through our chatbot rather than via email or phone,” Heidi adds.
KF Chatbot is a chatbot developed by KF and is used by various municipalities all around Norway to help citizens with various municipal issues. Stig Andre Dignæs Rundhaug is KF’s robot trainer and he also has positive experiences with using a chatbot.
“I’m especially happy with how open Kindly is compared to other platforms. That is really important to us seeing as we work in the public sector, and municipalities often have their own integrations, systems and different rules and regulations they need to follow. They often also have their own names for different services, so being able to add information relevant to a specific city or county, is incredibly helpful. It’s also great that you don’t need a degree in IT to use Kindly.”
Chatbots are challenged with longer questions
In the few months that DNT has used a chatbot, very few challenges have become apparent.
Heidi says that a lot of users are good at asking short questions that are to the point, which is what makes a chatbot thrive. They do, however, receive some longer and more complex questions, which can in some occasions confuse the bot. In those cases, users often ask to speak to a person and actually end up having a good experience after all because the employees at DNT are quick to pick up these conversations and take over the chat.
“It’s a useful channel even when the bot isn’t able to answer a question because our employees are quick and users will then receive help faster than what is possible with phone and e-mail,” Heidi says.
Mikael, on the other hand, doesn’t see any specific challenges with chatbots. And as far as the future goes: “The sky is the limit,” he says.
Stig, working in the public sector, sees challenges with chatbots specifically when considering older citizens who may not be used to technology. But he has solved this by creating a user manual in Norwegian, with pictures and descriptions.
Many ways to succeed
At DNT, several employees were trained in using Kindly, which has turned out to work really well. Having more than a few people know how to create good questions and answers for the bots, is especially valuable when making a second chatbot.
“Our expertise is completely different now than what it was when we created the bot for UT.no. We’re not making as many answers, but we’re trying to make truly good answers with a lot of corresponding questions. Our communal understanding of what kind of language to use and how to answer our users is very important,” Heidi says.
“Time, accuracy and a little inaccuracy at the same time,” are Mikael’s best tips for success. Adecco also launched their chatbot early, long before it was actually finished. That secured them a lot of valuable data that they could use to further improve the bot.
“I’d recommend this to others as well,” he says. “We see that a lot of people can tolerate a lot if the right premise is given in the chatbot’s greeting.”
Managing the user’s expectations is important, and the chatbots for UT.no, Adecco and KF all do a good job of explaining how they should be used. And the more realistic a user’s expectation is, the better a chatbot will perform. Stig is certainly as optimistic as Mikael and Heidi when it comes to where chatbots are headed.
“I think the future of chatbots is bright. And Convertelligence is doing the right things. I’m happy you have people with expertise in languages, IT and UX all working together. It’s a dynamic and organic way of thinking,” says Stig.