This firstly, is definitely a problem worth solving and the company made its first versions even usable on some non-critical use cases, which is amazing! If this works, will bring huge benefits (time, effort) compared to traditional surveys (take days, weeks, or even months). Of course, good things take time. There are a lot of barriers to be solved (trust and accuracy are the top priority, I believe).
Some suggestions come to mind at this point:
+ Consider improving UX and making the data explanation as easy to understand as possible (normal people don't understand semantic search results or vector comparing,.. so even if you expose these data to them, may not bring much benefit), su much technical data and so few information for people to confidently put their trust on the result (as I'm writing this at Nov, 2023)
+ I personally see it as a a large-data-warehouse for an effective survey, rather than generated & simulated survey, may the applied AI brings more fancy, but the true value of the product is the large pool of validated data, just that can save people ton of time and effort.
Congrats team and keep up the great work!
