FinanceThe long run

Quid of AI in the finance professions?

Analysts, bankers, fund managers: artificial intelligence is not replacing financiers yet. It is already changing what we expect of them.

2 min read

Art Deco illustration: at night, in a black, green and red office, a man in a 1930s suit sits thinking at his desk; in front of him, a typewriter unrolls a long ribbon of paper down to the floor. Through the window, lit skyscrapers and a bridge.
The machine writes on its own; the man does the thinking. That may be the whole question.

Every investment bank has a rite of passage nobody truly misses: the all-nighter spent lining up comparables, reformatting charts and checking that page 47 says the same thing as page 12. That is precisely the work artificial intelligence can now do quickly, and rather well.

The question is no longer whether AI is entering the finance professions. It is already there. The question is what it leaves to humans, and what that changes about how we train them, hire them and pay them.

What gets automated first

The first tasks to go are the ones that turn one document into another: summarising an annual report, pulling figures out of a prospectus, drafting a first credit memo, writing up a call with management. Nothing spectacular, but it is most of a junior analyst’s day.

Research and monitoring come next. A portfolio manager following dozens of stocks can ask an assistant to flag what changed in a chief executive’s language from one quarter to the next. The work is not done for them; it is cleared.

AI does not remove judgement. It removes the excuse for not having time to exercise it.

What holds out

The heart of the job holds firm. Deciding whether to invest, persuading a committee, negotiating a valuation, sensing that a founder is hiding something: all of this is judgement that carries responsibility. And in finance, responsibility is not delegated to a model, if only because the regulator will not allow it.

There is a more prosaic limit, too. Models are wrong with confidence. In a business where a misplaced decimal is expensive, a tool that invents a source with elegance remains a colleague whose work must be checked.

The real question: training

Here is the paradox. The tasks AI takes over are precisely the ones that trained beginners. You learned to read a balance sheet by copying it out, to understand a deal by building the model line by line. If the machine does it, how do you train the next generation of financiers?

The most clear-sighted firms are starting to answer by moving the learning elsewhere: less production, more critical review, earlier contact with clients. It is a bet. It assumes you can learn to judge work you have never done yourself.

Predicting the end of a profession would be reckless. Predicting a change in its content is safer. Tomorrow’s financier will spend less time producing and more time asking the right questions, checking and deciding. In other words, doing what was always expected of them, without ever quite having the time to do it.