Meet Kamu
Co-designing a chatbot for immigrants
Similar to the situation in many other countries, immigration rose in Finland over the last few years. In 2017, the population of Finland was 5.1m people of Finnish origin and 380,000 from foreign backgrounds (source: Statistics Finland). The number of yearly new arrivals has risen especially during the height of the European refugee crisis in 2016. At the same time there is a rising need for even more immigrants to join the labor force in Finland. In response to this change in Finland, a co-design and innovation team within the Finnish Immigration Service (in Finnish Maahanmuutovirasto, Migri for short) was established in August 2017.
Our co-design team, called Inland Design set out to improve the services of Migri, together with immigrants as well as civil servants. Migri makes about 94,000 decisions on residence permits, asylum cases, registration of eu citizens as well as Finnish citizenships yearly. On top of that it oversees the operations of reception centers for asylum seekers in Finland.
Since May 2018, Migri has had a chatbot, which answers commonly asked questions by providing pre-written answers. Inland Design was involved in setting up the working practices of the chatbot development team. The design team also had an impact on how the chatbot answers questions.
In the beginning of the project in 2017, our customer service phone line had an answer rate of around 20% — there were more calls than staff could answer. The project was therefore not about replacing humans through robots, but much more about giving additional support to the human personnel, so that they don’t have to answer the commonly asked questions any longer. This also means if the chatbot cannot answer a customer’s question, our human personnel takes over during the opening hours of our live chat. During a co-design workshop in September 2017, one of the biggest concerns of our employees was: will our clients trust a machine?
Together we identified four success factors, which led the development of the chatbot:
- Trust in the answers
- Understandability of the answers
- Same information in all channels
- Easy enough to use
To live up to the first two factors especially, we spent a lot of time developing the personality of our chatbot. We used co-design methods with staff and immigrants and led a workshop with our telephone customer service staff, who answer questions from immigrants on a daily basis. We ran a survey and did user tests with immigrants, and we had Migri staff vote for the name of the chatbot. Through this process the chatbot team and us, the designers, learned a lot about users’ expectations. We learned about the language level we could expect from users, and if Kamu could use emojis (the answer is yes and no). As a result we designed a Kamu personality poster and content guidelines. By doing this from the beginning, we tried to make sure that the several content designers who design Kamu’s logic and answers speak with one voice and personality, so that the users do not see a difference. We had our content designers do training in plain language, a need which only became obvious after several months.
In the beginning of the project, Migri’s staff had divided opinions. Some looked forward to the chatbot because they hoped to not answer the most common question many times a day: “When is my decision ready?” Others were sceptical if a chatbot could answer with the same standard as our trained human staff. Most of our staff is now in favor of Kamu due to how we designed and implemented it. We included all staff in workshops, in choosing the name, and openly communicated about the reasons behind our decisions. We also ran an internal pilot first, way before the immigrants could use the chatbot themselves. The internal pilot helped secure us the support of many of the sceptics.
Our metrics show the impact Kamu has had. In the second half of 2018 the answer rate for our customer service phone line increased to about 75%. Other projects have contributed to this raise, however, in part it is also attributed to Kamu.
While in the beginning we had to convince that Kamu can answer questions about this or that topic automatically, now different parts of the organisation ask the chatbot team to add content from their area of expertise to Kamu. In the beginning we used co-design methods to prove Kamu is useful for our organisation, we are now facing a different challenge: wow to prioritise? More and more often we ask ourselves: which content brings the biggest value for Finland’s immigrants? At the same time we are experimenting with different content interactions: a voice-based version of Kamu, multimedia content in the answers and whether Kamu can help immigrants to fill in forms as part of their conversation. Throughout the development of Kamu we’ve included end users and have proven within the Finnish Immigration Service that technology can be applied in a user-centered way.
More about Inland Design and Kamu: www.inlanddesign.fi
———
Suse Miessner is design and prototyping lead at Inland Design, the innovation team within the Finnish Immigration Service. She’s also an artist, gardener and urban enthusiast.
