AI in finance: what graduate analysts should learn now
Goldman's intern class shrank by 100. The job changed more

Goldman Sachs took 2,500 summer interns this year, about 100 fewer than in 2025. And its CEO says AI will "probably to some degree reduce the number of people that we start with".
(also: under 1% of applicants got in, for the third year running)
AI in finance isn't a future problem for graduate analysts anymore. It's in the interview room.
And the analysts who do get in will spend less time building the deck and a lot more time checking what an AI built.
What's actually happening with AI in banking
The picture is mixed, which is exactly why you have to read it carefully:
- Fortune reports McKinsey's QuantumBlack saying banks are cutting junior analyst classes by as much as two-thirds
- In the same piece, Bank of America is still bringing in 2,000 summer interns and 2,000 full-time recruits
- In the UK, High Fliers found banks planned to hire 2,365 graduates in 2026, up from 2,069 in 2025
- And per Reuters, a KPMG survey in June found 51% of banks piloting AI agents
And "agents" here means actual work:
→ BNY gives its "digital employees" logins, daily tasks and a human manager each → Goldman partnered with Anthropic this year to build agents for trading, transaction accounting, client vetting and onboarding → Barclays has summarised more than 8 million customer calls with generative AI
So fewer seats in some places, but every seat now comes with an AI sitting next to it.
One more McKinsey number worth sitting with: banks source roughly 62% of their AI talent from the same junior cohorts they're shrinking (Fortune).
So the juniors who can actually work with these tools are exactly the ones banks want to keep.
What Goldman actually told its interns about AI
At the intern town hall in August, David Solomon told them to "really try to understand what these things do and don't do", and that "you have to challenge what the models give you".
Goldman's head of human capital, Jacqueline Arthur, said it even more plainly to Fortune: "what's much harder to instill are qualities like judgment, adaptability, and critical thinking."
And from McKinsey's Debasish Patnaik: "Banking is an apprenticeship business."
Translation: the tools are table stakes. Checking them is the job.
What a finance degree won't teach you
Most finance and economics courses still grade you on producing the model.
The job is drifting towards checking it.
bad: "I built a DCF in Excel" good: "I built a DCF, had AI build one too, then found where they disagreed and why"
Uni rarely teaches you:
- tracing every number back to a filing
- sanity checking outputs that look right
- explaining a view to a client in plain English
A critical-thinking framework for graduate analysts
Five checks before any AI-assisted number leaves your desk:
- Trace it. Where did this figure come from? Annual report, filing, data provider? If you can't point to it, it doesn't go in the deck
- Smell test it. Is that margin, multiple or growth rate even plausible for this sector? AI is very good at confident numbers that are off by an order of magnitude
- Find the assumption doing the work. Which single input moves the answer most? Flex it and see what breaks
- Ask what would change your mind. Solomon says he uses models to stress-test his own read on events. Do the same: make the model argue the other side
- Say the so-what. One sentence a client would actually care about. If the AI wrote that sentence, rewrite it in your own words
What to do this month
- Pick one listed company. Build a quick model by hand, then with AI, and write down every difference
- Read one full annual report, notes included (that's where the interesting bits hide)
- Practise a two-minute verbal pitch on a stock or deal, no slides
- Prep a real answer to "how do you use AI, and how do you check it?" with one specific example
so TLDR: AI in finance is shrinking the grunt work, not the need for judgement. Learn to check the model, not just build it.
What's one number you'd never put in front of a client without checking it yourself?
Sources
- Exclusive: Goldman Sachs intern acceptance rate falls below 1% for third straight yearFortune · 08 JUN 2026
- Exclusive: Goldman CEO David Solomon on how his dad's advice shaped his career, and why interns should steal itFortune · 06 AUG 2026
- Banks lay groundwork for mass workforce cuts as AI takes holdFortune (with Bloomberg) · 07 JUN 2026
- Investment banks may hire record graduates in 2026, getting a job could be harder than evereFinancialCareers (reporting High Fliers, The Graduate Market in 2026) · 09 FEB 2026
- Wall Street banks ramp up digital assistants in bid to win productivity raceReuters, via Kitco News · 13 JUL 2026
Get the next issue by email
theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.
