Interviewing chatbot architecture. How to?

Hi! I’m looking for some info about architecture for interviewing/advising (proactively) chatbots when the goal is to find out/evaluate a user’s situation/level via conversation and come up with advice vs common Q&A bots.
I could get a quite complicated flowchart/state machine/graph which is apparently would be needed to direct the conversation but it’s still unclear how to combine it all together. What has to be the chunks? How to embed the conversation rules and make the bot proactive? How to keep track of the context?
Looked into MongoDB chatbot but not sure it can be used like that.
Thank you!