Goldman Sachs accepted fewer than 1% of its summer intern applicants. Third year running. And AI agents can now draft the pitchbooks those interns used to build overnight.

So which finance graduate skills still matter? Honestly, the same three tools as before (Excel, Python, AI), just used very differently.

Here's the toolkit, and how to use each part without letting it think for you.

What's changing for finance graduates right now

Three things happened in the last six weeks.

1. AI agents moved into the junior workflow. On 5 May, Anthropic launched ten finance agent templates:

→ pitch builder, meeting preparer, earnings reviewer, model builder, market researcher → valuation reviewer, general ledger reconciler, month-end closer, statement auditor, KYC screener (know your customer checks) → Claude add-ins for Excel, PowerPoint and Word are now generally available

Read that list again and it's basically a first-year analyst's job description. The catch: Anthropic says users "stay firmly in the loop", reviewing and approving work before it goes to a client.

2. The CFA added AI to the syllabus. The 2027 curriculum update, announced in late May, adds a module on financial data science, AI and large language models, plus Python-based Practical Skills Modules at Levels I and II. It applies from February 2027.

3. Goldman's intern class shrank a bit. Fortune reports about 2,500 interns this summer, roughly 100 fewer than last year. David Solomon said AI will "probably" reduce starting classes over the next few years, but "probably not what you and I would call dramatically".

The line that matters most for you: Goldman's head of human capital management told Fortune technical skills can be taught, but judgement, adaptability and critical thinking are harder. And how candidates apply AI tools is the key differentiator.

The finance graduate skills toolkit

1. Excel: still the language of the desk

Every agent above outputs into Excel or PowerPoint. If you can't read a model, you can't check one.

→ learn: three-statement models, lookups, scenario tables, keyboard shortcuts → the new skill: auditing a model someone (or something) else built. Trace every hardcoded number

2. Python: for the stuff Excel chokes on

Python in Excel has been generally available to Microsoft 365 Business and Enterprise users on Windows since September 2024, so you don't need to leave the spreadsheet to start.

→ learn: pandas for cleaning data, simple charts, pulling prices from a public source → the new skill: reading AI-written code well enough to spot when it's quietly wrong

3. AI agents: the junior who never sleeps

→ use them for first drafts: a comps table, an earnings summary, a meeting brief → the new skill: being the reviewer. That's literally the job Anthropic's own launch describes

What uni won't teach you about AI in finance

Finance degrees are great on theory: valuation, CAPM, portfolio maths. Where they fall short:

  1. almost no model auditing. You build models in coursework. You rarely get handed a broken one to fix
  2. Python is optional. The CFA is now putting it in front of every candidate. Many degrees still don't
  3. judgement isn't assessed. Exams reward the right number, not spotting that the question was wrong

We've also covered why "AI stocks" hype rarely survives contact with the evidence in AI investing: hype vs reality. Same instinct applies to your own spreadsheets.

The critical-thinking move: 4 checks for any AI-built analysis

  1. Where did every number come from? No source, no number. Same rule a VP will apply to you
  2. Does it tie out? Balance sheet balances, cash flow reconciles, totals add up. Agents get this wrong more than you'd think
  3. What assumption is doing the heavy lifting? Find the one input that moves the answer most, and ask if it's reasonable
  4. Would I defend this in front of a client? Solomon wants juniors in front of clients sooner. That only works if you understand what you're showing

What to do this month

  1. Build one three-statement model by hand for a listed company. Then ask an AI to build the same one and compare. Write down every difference
  2. Do one small Python project: pull five years of prices for two stocks and chart them. Put it on GitHub
  3. Practise reviewing: take an AI-generated earnings summary and check every figure against the actual results release
  4. Borrow the analyst stack from our junior analyst AI toolkit for consulting. Research, slides and spreadsheets work the same way in banking

And keep the thinking yours: the method in studying with AI without outsourcing your thinking works for modelling too. More on our finance hub.

so TLDR: Excel, Python and AI are still the toolkit, but the skill that gets you hired is checking the output, not producing it. Learn to audit models, read code and defend numbers.

Which part of a financial model would you never let an AI agent build without checking line by line?

Sources

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.