The traditional consulting answer to "where do we start?" is a discovery phase. Two weeks of interviews, a deck, a roadmap, a steering committee. Nobody has ever pinned a steering committee to their wall.
We do something different. Week one, we ship a v0 of the thing. It's not the right thing. It might not even be a useful thing. But it's a thing the client's actual users can click, and the conversation we have in week two is night-and-day different from the conversation we'd have had over a deck.
The trick is to ship a v0 that is correct about the seams (where data comes from, where it goes, who decides) and wrong about everything else. Wrong is fine — wrong is easy to fix. Misaligned seams are expensive to fix.