I've been having quite a few debates with friends on colleagues and how to architect a new solution and would like to seek a wider group of potential opinions. There's certainly more context to this debate, but at the core, the part that sparks the most disagreements is how to handle all of the potential iterations of questions and follow-up questions. For example, consider this line of questioning:

​

Are you old enough to drive? If yes, do you have a vehicle? If yes, what make, model, and color? And can you upload a pic?

​

In the actual use case, the potential number of follow-ups could be endless and unknown, we could go 2-3 levels deep, we could go 7-10.

​

The system will need to evolve and be trained based on responses and categories of responses (e.g., acceptable or standard potential answers vs. non-standard). To a certain degree, we're just collecting data, but at the same time, we are looking for determine a level of "quality", some answers are better than others (i.e., I drive a car that's 2-3 years old vs. 12+ years old).

​

Based on the above, you may initially think of bringing in a AI / ML component, that's certainly on the table, but looking to consider traditional logic first, don't want to over engineer the solution if not necessary.

​

Thanks in advance for your opinions and feedback.