The biggest gotcha is that if existing products were developed over a decade or more, that is decade of iteration over details and customer feedback. You can see the final result, but not the rationale behind 10+ years worth of decisions and discussions. The LLMs are almost guaranteed to get something wrong without that context, which means you final product won't be competitive. Unless you understand the nuance of which features are table stakes vs. market choices vs. regulatory requirement or other such fixed functionality, you might spend all your energy building something that is not even viable.

That doesn't mean you cannot build a newer, better, competitive product. You surely can. But you need to build the understanding of the market yourself so you know when the LLMs go off the rails and get them back on track.