Unfortunately, yes/no questions are generally not enough to provide guidance on any specific action unless they are of a specific domain of knowledge. And even a highly focused topic of knowledge will reveal shortfalls that cannot be easily overcome in terms of the codification of the pathways to a conclusion about the question.
This is [partly] why AI and NLP (natural language processing) converge to create smart systems that act like conversational problem-solvers. In that sense, @Paul_Danyliuk’s idea is probably a good place to start.
In terms of practicality, Coda is an ideal place to experience conversational activities with machine learning or with other collaborators, but not well-suited for building such an expert system using formulas. Decision trees based on conditional states would require massively complex formulas. The way forward likely requires integration with a system such as Dialog Flow which would allow you to map linguistic patterns, user intent, and entities to determine how best to respond to conversations.
This approach does not rule out simple yes/no answers to questions, but specifically rules in ways to address queries that people will type unexpectedly, thus providing a more productive experience for getting answers.
But there’s a key requirement for a smart system or any sort of conversational process in Coda - the need for real-time interaction with services outside Coda. This suggests that it would need access to a real-time network (such as Ably or PubNub) and event handlers that make such interactions instant (humans hate waiting for more than 500ms when chatting).
It’s a really interesting idea and you may be able to at least prototype the concept in Coda. Certainly you can design and document the idea in Coda.