The idea of the project seems fairly straightforward - you have a chatbot design in some data format, you have some services for sending and receiving SMS messages, so find a way to engineer them together.
After further research, it seemed that both of these seemingly simple tasks had many unknowns. Neither Voiceflow actually supported any kind of data export, nor was there an SMS service that supported the bot capabilities we needed.
Indeed, the first few months were invested solely in research to answer these unknowns. Every time we learnt something, we developed hypotheses, and consequently proved or disproved them.
We took an agile approach to manage that. There wasn’t anything certain to step on - any aspect of the project could change based on new information. Are we developing a browser extension? A Windows native app? A server API? Do we have a backend? Can we even use Voiceflow (remember: no data export)? If not, do we have to make our own super intuitive bot design interface? Making plans at this stage was meaningless.
As such, there wasn’t a first question that we had to answer, a fundamental hypothesis that we had to validate. We had to try and answer everything at the same time, so we went researching in parallel.
This section’s articles share our research findings and processes, grouped by theme, but not by timeline. Everything went in parallel, and the findings of one type of research influenced the decisions for another.