I don't see the difficulty as finding the problem (schlep blindness) or an idea of the solution. Rather, it's always imagining realistic ways of implementing a solution.

In payments, for example, the problem is never "building the thing". It will always be gaining permission (whatever it means for your application) of hooking into the system. It might require depositing some amount of capital; and it's usually an amount larger than any startup can reasonably get. Most such barriers are there by design.