Banking (Un)fortunately Still Needs You: My Case for Bankers in the AI Era

Yesterday, I finished a model end to end in two days that would have taken me at least a week otherwise.

AI agents like ModelML, Hebbia and Rogo have made life far simpler for analysts / associates. The hard part of the job has always been figuring out things that seniors knew automatically, while you somehow had to learn everything through osmosis. AI has commoditised this.

On a broader scale, my concern is that capital will flow away from workers and towards AI companies / asset owners even faster than it already does. This will accelerate further when humans figure out physical AI, which feels like the logical next blue ocean.

So, are we cooked, chat!? I personally don’t know how I would have felt if I were still a student, so let me try to think of it this way.

I still see a need for MDs and Partners. But “someone needs to manage capital” isn’t a good enough argument on its own. AI could probably do a lot of that too.

The stronger argument is that having the answer and getting people to act on it are two very different jobs.

I can see AI taking over more of public markets. In private markets, the information disparity could shrink enough that parts of the market start looking more like publics. But knowing what an asset is worth doesn’t mean you can buy it. Someone still needs to convince the owner to sell, get an IC comfortable, negotiate terms and keep the whole thing from falling apart when someone suddenly decides they don’t like the deal anymore.

And by natural extension, there is still a job for MDs in banks.

AI might know that the partner at KKR absolutely abhors this other dude working at Bx. But knowing that and getting the two of them to transact anyway are different things. MDs know which egos need massaging, which conversations need to happen off the record and when “we’re not interested” actually means “come back with a better price.” It is their job to know.

Rubbing shoulders and grifter sales. Somehow, still a moat.

There is also the question of responsibility. An AI can recommend a deal. Someone still has to put their name behind it, defend it when it goes wrong and ask the same investors for money again. You can automate the analysis much more easily than you can automate being the person everyone holds responsible.

At the end of the day, I love the irony that some of the little inefficiencies and flaws in the job might be what helps humans keep their jobs.

Being technically good still matters. It just becomes less of a differentiator when everyone has access to technically good.

And for you all, the world still needs Partners and MDs, which (un)fortunately means it still needs some of you.