I have limited experience with that field. It looks right to me, though. I'll add that metaprogramming has untapped potential far as interfacing with the legacy systems. Code the prototype in a familiar language, carefully enter it into a transformation system (eg Semantic Designs, Racket Scheme), and generate a robust version of that nonsense you'd rather not write yourself. There's still work to do the initial transformation rules and/or integration. However, it pays off as you don't have to do it again.
The same logic applies to fintech. The opportunity comes from the incumbents using outdated technology and concepts.
Yep. I can't wait to see them devour SWIFT's customer base. :)