How to make users chat with an entire platform?
The platform DataSebrae has been growing fast since 2016. New digital solutions are created almost every week and we need to keep easiness of navigation in mind. In this case, we wanted users to simply ask their questions, in natural language, and the system would tell them where to find the answers.
That’s how we created the virtual assistant of DataSebrae, initially called Sofia. Based on the technology of Google for the development of chatbots, we created something that would comprehend the user’s necessities, and guide them up to the information.
Sofia’s role and personality
Ok, that’s consensual that chatbots need to look like robots, otherwise, we would be deceiving the users, as if a human being were there. On the other hand, we also wanted to produce a smooth dialog and not a search query. Thus, Sofia was created as a character. She was a 24 years old girl, who works in the team of DataSebrae and is pretty sympathetic and talkative. Just like someone people would like to chat with. In fact, we used to know when our team was testing her by sudden laughs. A few supernatural answers were usual…
Going further, we took a strategic decision. As we had a very limited team working on her, we decided that she wouldn’t give direct answers. Instead, she would give links where people could find the information. The reason is simple. Direct answers would require a big time for development. That time could be used to cover a much larger amount of available content in the platform, and not only a few specific questions.
Maybe in a further cycle of development, we can aggregate those specific answers, and that’s all about Scrum: first, what gives more value, and second, what gives less value.