The scarcest person in AI right now isn't someone who simply understands AI.
It's someone who understands the business deeply enough to know what AI should actually build.
Hedgineer recently explained, in public, why they killed their own forward-deployed engineer role.
The FDE model needed unicorns: engineers who could also speak twenty years of finance fluently. Out of thousands of engineers, they found five or six people who fit. Every one of them cost seven figures.
So they flipped the model. Hire the domain expert first, someone who already has the taste and context from sitting on a credit desk or covering TMT at a fund, then let AI translate that expertise into the technical layer. Not the other way around.
That resonated with me.
I've spent seventeen years in credit trading. Knowing what works, what doesn't, and where the real bottlenecks are is the part that doesn't get replaced. AI doesn't replace that experience. It dramatically increases its leverage.
Domain expertise without AI increasingly leaves value on the table.
AI fluency without domain expertise produces polished answers to the wrong problems.
The combination is where the value is moving.