AI Isn’t Coming for the Analyst. It’s Coming for the Analyst’s Monday.

Every finance student has heard the warning: AI is going to replace analysts. The truth is less dramatic and more useful. AI isn’t replacing the analyst. It’s replacing the parts of the job that nobody got into finance to do.

Think about what a junior analyst actually does on a Monday. Pull the 10-K. Copy the income statement into Excel. Reformat the dates. Hunt through the footnotes for the lease adjustment. Rebuild the comps table because a peer reported last week. Draft the first version of a memo that a VP will rewrite anyway. Very little of that is analysis. Most of it is data handling and formatting, and that’s exactly what large language models and the tools built on them have gotten good at.

What’s changing

Three things are moving fast.

First, data extraction. Models can now read a filing, pull the segment data, tag it correctly, and drop it into a structured table in seconds. The work that used to eat a Sunday night is becoming a prompt.

Second, first drafts. Earnings summaries, variance commentary, the “what happened this quarter” paragraph at the top of a research note: AI writes a competent version instantly. Not a final version. A starting point.

Third, pattern recognition at scale. A human can read maybe a dozen earnings calls in a week. A model can read a thousand and flag the ones where management’s tone shifted, where a guidance range got quietly widened, where a word like “transitory” showed up for the first time.

What’s not changing

Judgment. The model can tell you that free cash flow conversion dropped from 85% to 60%. It cannot tell you whether that matters, whether it’s a one-time working capital swing or the beginning of a structural problem, or whether the market has already priced it in. That’s still the job.

Accountability. When a model is wrong about a number, someone still has to own the recommendation. Clients, portfolio managers, and regulators don’t accept “the AI said so.”

Relationships. Deals get done because people trust each other. Capital gets raised because someone picked up the phone. No model does that.

What this means if you’re building a career

The analysts who struggle will be the ones whose value was speed at mechanical tasks. If your edge was being the fastest person at building a comps table, that edge is gone, and it’s not coming back.

The analysts who thrive will be the ones who treat AI as a leverage tool and spend the recovered hours on the parts of the work that compound: understanding the business model, forming a view, being right more often than the consensus.

Practically, that means a few things. Learn to prompt well, which mostly means learning to ask precise questions. Learn to check the model’s work, because it will be confidently wrong on occasion and the errors are subtle. Learn enough about how these systems work to know where they fail: they’re weak on long arithmetic chains, prone to inventing sources, and they don’t know what they don’t know.

And keep building the fundamentals. A model can generate a DCF in thirty seconds. Only someone who understands a DCF can tell you whether the terminal growth assumption is nonsense.

Zoom out

Finance has gone through this before. Spreadsheets were going to eliminate accountants. Bloomberg terminals were going to eliminate research. Algorithmic trading was going to eliminate traders. In every case, the headcount shifted, the job changed, and the people who adapted ended up with more interesting work than the people who came before them.

AI is the same story at a faster pace. The Monday morning grind is disappearing. What you do with the time is up to you.