Will AI Replace Financial Analysts?

The model came back in two days. Three weeks of work a first-year would have done, handed back by the tool over a long weekend, good enough that the managing director barely marked it up.

Your managing director mentioned hiring fewer first-years per partner. A peer got cut when her bank rolled up its research function. Two weeks of that, and the question stopped feeling abstract.

The sharper read isn’t about whether AI is coming for the work. It’s about which layer of the work.

The clearest framing is geometric, not apocalyptic. The career has always been a pyramid. The base (cleaning data, building base-case models from a template, writing first-draft memos, formatting decks) was the apprenticeship that taught you the craft. The top (figuring out what question to ask, defending an investment thesis, owning the recommendation) was where the career compounded.

Those two layers are now moving in opposite directions, and that’s the actual story.

The Bureau of Labor Statistics projects financial analyst roles to grow about 6% through 2034, faster than the average job. The Anthropic Economic Index has financial and compliance analysts among the highest-Claude-usage occupations on its platform. Both are correct because the work AI is doing concentrates at the base of the pyramid, while the demand concentrates at the top.

The pyramid is losing its base

The base of the pyramid (the first three years of the job) used to fill the analyst’s week with a small set of repeating tasks:

• Pulling and cleaning data from disparate sources. AI does this in minutes with the right access, and improves every quarter.
• Building a model from a template. Last quarter’s model with new inputs, the standard comp-set, base-case DCF with reasonable assumptions.
• First-draft memos and decks. The one-pager for the senior person to mark up, formatted comparable-company tables, draft earnings call summaries.

None of that work was strategically valuable in itself. It was valuable because doing it for two or three years taught the analyst how the business worked, how a model breaks, and what good output looks like.

That work is now a Claude or ChatGPT conversation away. The model isn’t perfect, but it’s good enough to skip several hours per analyst per week, and it improves every quarter.

The analysts who come through the Workplace AI Institute and get ahead are the ones who spent the freed-up time moving closer to the senior work, not doing less of it overall. The first group gets promoted faster. The second runs into the shortened apprenticeship ladder the WEF Future of Jobs report flags.

Does your day sit at the base of the pyramid or near the top? The 3-minute readiness check places you on it.

What AI cannot do, and is not about to

Three parts of the job belong to the senior analyst, and each gained value as the base compressed:

• Asking the right question. The market is full of facts. The senior analyst knows which two or three facts will move the recommendation, and which thirty are noise. AI cannot do that step without being told what to look for.
• Defending the call. When a portfolio manager pushes back on a thesis, a senior analyst needs to stand behind their numbers, address the counter-argument, and update the view live. That’s a conversational skill, not a modeling skill.
• Owning the relationship. Buy-side analysts have relationships with management teams. Sell-side analysts have relationships with clients. Corporate FP&A analysts have relationships with operators inside the business. Those relationships are how the non-public context gets into the analysis.

What’s actually changed, in our reading, is that the apprenticeship is shorter and the bar to clear it is higher. The three years a junior used to get before contributing on the senior side is now eighteen months or less.

Where the tools speed the analysis

An analyst can push three kinds of work to these tools:

• Compressing the data-pull and modeling phase of every cycle. The first-day work of any new name is the part AI handles best.
• Stress-testing the thesis. Have AI argue the strongest objection back at you, name the one assumption most likely to be wrong, propose the two stress scenarios that would expose a silent failure.
• Drafting deliverables for senior review. The first version of the memo, the standard comp-set table, the earnings call summary. The senior edits and signs.

Save this for the night before you defend a thesis in a portfolio review.

You are a senior financial analyst stress-testing the model below. Company is [NAME, INDUSTRY]. The base case assumes [LIST 3 KEY ASSUMPTIONS]. Identify the three assumptions whose failure would most damage the thesis. For each, propose a specific datapoint, document, or call topic that would confirm or refute it.

That prompt doesn’t produce a recommendation. It pushes you toward the work AI cannot finish, which is choosing which of the three flagged risks is the one that actually matters and going to find out.

Climbing off the base of the pyramid

The firms that attribute EBIT impact to AI cluster in finance and tech functions, McKinsey’s State of AI finds, and the value tends to come from senior people doing more rather than from junior headcount cuts.

Three things to do before next bonus season:

1. Spend the time AI saves you on the parts of the job AI is bad at. If a research draft now takes you ninety minutes instead of half a day, the saved hours should reappear in conversations with management teams, in deeper sector reading, and in pre-recommendation testing of your own thesis.
2. Build evaluation, not just generation. The analyst who can spot a wrong-but-confident output gets a premium that didn’t exist three years ago. Practice breaking the model.
3. Pick one or two industries to build genuine depth in. AI value compounds where context is hard to compress. Industry expertise is one of the densest forms of that context. Generalist analysts who try to cover everything with AI assistance are competing with software.

Push yourself toward the judgment end of the pyramid, where the recommendations get made and defended, and the climb is worth it.

So will AI replace financial analysts?

The BLS projects more analyst jobs ten years out, not fewer, and the Anthropic data shows the role is one of the heaviest current users of the technology. Both facts point at the same conclusion.

The shape of the career is changing, the entry rung is harder to reach, and the top of the pyramid is in higher demand than it was. Where you stand on that pyramid in 2027 is decided by what you do with the next eighteen months.

To place your own role on the pyramid instead of guessing, the readiness check works through a week of your tasks.

The structured version, the AI for Financial Analysts course, runs from data compression through thesis defense.

The model will keep coming back over the weekend, finished. What won’t is the person who decides which question the model should have answered and defends the call in the room. Climb toward that seat, because it is the one the tool can’t sit in.