This is a story about industry jargon, human conversations and why building a dumb chatbot can be really smart.

Date: April 2016
Client: Gjensidige Forsikring
Scope: 3-week sprint + development

A complex world

«Even I don't understand all of it!» 

Sitting in one of the conference rooms in Gjensidiges new and stylish headquarters we were a little surprised by this outburst from the senior employee.

Gjensidige had prior to this meeting presented us with a list of the most urgent problems that they and the insurance industry were facing. We picked what we thought was the most user-centric: helping young adults understand insurance.

We discovered early that insurance is complicated. Very complicated. Industry people were telling us it represented a community coming together to help each other in case of unforeseen incidents. However, talking to customers we learned it was more recognized for its ability to hide important information in plain sight using industry jargon and complicated legal documents.

Narrowing it down to car insurance, we saw that people got stuck on questions like «What’s my bonus?» and «What distance do I drive in a year?» The abstract nature of insurance and the potential consequences of a mistake made it really uncomfortable for new buyers to make decisions.

A visit to the customer support center
turned out to be our breakthrough

Customer support

«Hi! Welcome to Gjensidige how may I assist you?» The voice of the phone support was calm and kind. «You bought a new car? Congratulations! Let me help you get it insured.»

A visit to Gjensidige's customer support center turned out to be our big breakthrough. Listening in on the support staff's conversations with customers we learned that they had already solved our problem. When asking questions they used relatable comparisons that made everything feel much less intimidating. Their ability to empathize with customers had made them experts in helping people understand insurance.

Now, all we had to do was replicate their success in a way that was more accessible to our users.

An interesting phenomenon

OK, let's stop for a moment and look at some context. At the time of writing people are making chatbots for everything (though extremely few are any good). But this story takes place in the beginning of April 2016, roughy two months after Quartz launched a news presenting chatbot, and just a week before Facebook unveiled their own chatbot API for Messenger. Neither Quartz nor Facebook was first at what they did, but both mark important events in what would be become the year of the chatbots.

So why is this relevant? It's interesting because our situation illustrates a common phenomenon in technology, where different parties spread around the world are working on more or less the same solutions, all believing that they are the only ones. Back in the days when the computer was invented, scientists in different places on the globe, totally isolated from each other, came to the same technological breakthroughs at roughly the same time. Walter Isaacson's The Innovators has some great information on this topic.

«A chatbot for news is cool,» we thought «but a chatbot for customer support makes perfect sense!»

It was the obvious choice.

Dumb in a smart way

So what's different about Bable? Well, Bable is a chatbot in its simplest form. Most of the time just giving you a question and two options. A or B. Then another question. And some times you get an input field, custom designed for that specific question. 

We think that most chatbot makers are trying too hard - that they have been allured by the technological opportunities in AI and language comprehension. For the most part, the result is a chatbot that never really get what you meant and a lot of dead ends. On top of this, most people don't know how to write to a robot. And how can they, when they don't know the bot's limitations?

That's why all possible conversations with Bable are scripted. No AI, no magic tricks. Just a lot of work by Jørgen, the project's chief of story. After creating a linear storyline he focused all his energy on writing in a whole bunch of shortcuts and detours. In a way, when you're having a conversation with Bable, you're really talking with Jørgen. 

While Jørgen wrote the story the rest of the team
worked on Bable's user interface

First meeting with Bable

New insights

After testing the Bable on we got a lot of new insights. In the first iterations we intended Bable to be an alternative to existing solutions, like information pages and order forms. What we learned was that Bable would be far more effective as a supplement than as an alternative. 

Obviously, it's great that you can get your car insured through a single conversation with Bable, but it's just as important to help the ones filling out traditional forms or that have trouble navigating all the information.

Our experience working on Bable has truly underpinned our belief that a big idea only gets you so far.

It was not the idea of solving customer problems with a chatbot that made Bable great, nor was it the technological achievements. It was all the details and nuances based on empathic insight.

Update: August 2017

The Bable project is now incorporated in a larger initiative to digitize Gjensidige's customer service. It's future depends on several technological and organizational decisions, and until those are made Bable is enjoying some well deserved time off.


Script Jørgen Kvalsvik
Design Jens Obel
Tech Ole Strøm
Facilitators Ludvig Nevland, Andreas Solbakken
Additional design Alexander Schüssel
Tech partner Capra Consulting
From Gjensidige Tom Engen, Per Kåre Olsen, Cathrine Saxebøl, Sofie Selstad