AI for Financial Modeling: Will It Kill Excel and Entry-Level Investment Banking Jobs?

I’ve covered the topic of artificial intelligence several times on this site, including why I stayed away from enterprise AI startups last year and a quick review of ChatGPT for investment banking a month after it launched.

But a lot has changed over the past few years.

These tools have advanced and become multimodal (i.e., they work with text, images, files, and video), many vertical AI apps have popped up, and now there are tools that can generate both PowerPoint and Excel output from prompts and company filings.

I’ve also seen an uptick in panicked messages from students wondering which jobs are “safe from AI,” and whether industries like tech, finance, and law are even worth it anymore.

So, it seemed like a good time to revisit this topic.

I’ll explain below the immediate catalysts for this article, present both sides of the “AI will kill white-collar jobs” debate, and attempt to reach some conclusions.

I won’t provide an upfront summary, but you can skip to my conclusions and predictions if you don’t want to read everything.

But I’ll start with a quick “market update”:

AI for Financial Modeling: What “Prompted” This Article?

As of mid-2025, a bunch of new AI tools geared toward use cases in finance have been released:

• Claude for Financial Services – It’s a modern, more automated/AI-driven version of Capital IQ or Bloomberg. And you can find some cool demos on YouTube.
• Shortcut AI – Automates the process of building 3-statement models and valuations via a platform that is “compatible with Excel” (traditionally, large language models were quite bad at spreadsheets).
• ProSights – This one’s a bit narrower in scope and is more about automating specific process steps, such as extracting tables and creating “smart charts.” It’s more of a super-powered macro package.
• Hebbia – This one’s more of a “knowledge worker” solution for various industries (finance, law, corporate) that speeds up processes such as screening for buyers and sellers, coming up with due diligence question lists, “generating” entire pitch books, and drafting initial CIMs. Auquan is similar but more for credit and asset management.
• OffDeal – Claims to be an “AI-native investment bank” that can automate the process of marketing and selling small businesses. They have some great media coverage, including this clip with Andrew Ross Sorkin. But to be fair, what they’re doing is more in the realm of “business brokerage” and not investment banking.

There are probably dozens of others; it’s almost too much to keep track of.

Initially, large language models (LLMs) were quite good at generating text and answering questions more naturally than Google searches.

Then, they got much better at coding with tools like Cursor and Claude Code, and they’ve now come for PowerPoint (easier) and Excel (harder).

As I see it, these developments create two key questions:

1. Investment Banking Business Model – Will “AI-native banks” disrupt the incumbents and turn IB into more of a commoditized/automated business?
2. Entry-Level Investment Banking Jobs – If these AI tools can generate entire presentations and financial models, will banks still hire Analysts and Associates? Will there even be entry points into the industry anymore?

I cannot properly address both in one article, so I’ll focus on question #2 here.

I’m much more skeptical of question #1 because I don’t think OffDeal’s model will scale to larger companies.

Yes, someone with a $1 million per year nail salon in Arizona might sell it with an automated approach, but the CEO of a $1 billion company will always want humans involved.

Disclosures & Disclaimers: Where I Stand

I’ll say upfront that I have very mixed views of this AI hype cycle and all the funding and startups associated with it.

I have experimented with these tools for different tasks (image generation, scanning filings, suggesting model approaches, getting feedback on writing, etc.), but I’ve found them to be… less than amazing?

They have saved me effort, but not necessarily time.

I never use AI to generate content and would never publish any AI slop on this site (hopefully, you can tell).

My view is that these tools are best for automating boring and repetitive processes, such as lip-syncing characters in movies or games or updating numbers in models based on PDFs.

They should not replace thinking, judgment, or decision-making, or we’ll end up with generations of even dumber humans.

So, while I’m not as skeptical as someone like Ed Zitron, I also do not “drink the Kool-Aid” in the same way Tech Bros and Big Tech executives do.

The Bull Case: AI for Financial Modeling Will Severely Reduce the Number of IB/Finance Jobs

In this version of the future, senior bankers will still exist, but they simply won’t need Analysts, Associates, or other junior employees.

They’ll be able to type a company or deal name into a prompt and instantly get a full valuation, pitch book, and lists of buyers or sellers.

These automation tools won’t get everything right, but the senior bankers can keep tweaking the prompts and asking for re-dos until the output matches what they want.

And since junior bankers spend most of their time on modeling and presentation work, banks simply won’t need them anymore.

Most models and presentations follow predictable formats, so it’s not as if dozens of edge cases will flummox these tools.

Also, unlike in buy-side roles, there isn’t much “critical thinking” in most IB assignments.

You’re not going out to conduct channel checks to form your own views on a company’s projections, nor will you “challenge” management’s views or forecasting approach.

Most of the job consists of following the client’s wishes/demands, which means the workflow is even easier to automate.

Junior bankers won’t be eliminated because banks still need to groom senior bankers, and parts of the job do involve tasks outside of Excel/PowerPoint/Word.

But the headcount will fall dramatically, so the large banks might hire classes of dozens rather than hundreds in the future.

The Bear Case: AI for Financial Modeling Will Make an Impact, But Won’t Reduce Headcount That Much

The strongest counterarguments to the above case are:

1) Much of the work in IB is already based on templates, such as existing models and presentations, and little time is spent creating anything “from scratch” (which is what all AI tools are best at).

2) Historically, automation tools such as Capital IQ, FactSet, ARGUS in real estate, and even Excel itself have not eliminated jobs; they’ve just allowed bankers to do more with less.

3) While junior bankers do spend a lot of time in Excel and PowerPoint, much of that time is spent refining content and making small tweaks. This is much harder to automate because it requires careful checks and reviews of tiny details.

Sure, maybe Managing Directors can enter prompts to get presentations, but it still takes a lot of time to review the output and triple-check everything.

And no MD wants to do that; they aim to spend ~100% of their time on clients, potential clients, and deal negotiations.

So, if anything, these AI tools could produce more work that the junior bankers need to check and fix (“Why use just 3 scenarios? We can run 30 now!”).

Junior bankers will spend less time “doing” and more time “checking,” but the headcount will not fall dramatically.

Likely Changes in Investment Banking in the “Base Case” Scenario

I think both these positions have some validity, so I’m not sure there is a clear “winner.”

But I think many people have been thinking about this issue incorrectly – it’s not about whether entry-level jobs will “go away,” but rather how the entire hierarchy will change.

For example, you could make a case for any of the following outcomes:

• Many senior bankers will leave to start their own firms, and they’ll be able to advise on much larger deals with smaller teams. Headcount might not change much, but it will be distributed over thousands of additional firms.
• Many expensive mid-level banker roles will disappear because junior bankers will now perform similar functions more effectively.
• There will be fewer entry-level roles because the mid- and senior-level bankers can just generate all the work and check it themselves.

I predict that more entry-level and mid-level roles will “merge,” so we’ll see more Associates who are effectively acting as Analysts and VPs at the same time.

This already happens in many emerging and frontier markets, where there’s usually one “Negotiator” and one “Doer” on each team.

It will take a long time for the biggest banks to adapt, but it could happen more quickly at the regional boutiques and middle-market banks.

Investment banking is a professional services industry, so the MDs and Partners ultimately control the cost structures.

If they can get by with 2 Analysts instead of 3 Analysts, 1 Associate, and 1 VP, they’ll do it.

I don’t think they can forego hiring junior bankers altogether because they do need to groom successors and keep the firm going in the long term.

But I would expect to see the following changes over the next 5 – 10 years:

• A flattening/compression of the hierarchy, with more focus on hiring junior bankers who want to stay in the industry.
• More competitive recruiting due to fewer spots, but also faster advancement for those who perform very well and can credibly win deals.
• More time spent editing and checking output and less time writing/building anything from scratch or messing around with manual data entry.
• More importance placed on “debugging” Excel files and finding problems in the links, assumptions, and drivers from generated models.

AI for Financial Modeling: What Can You Do About This?

Whenever an industry becomes more automated, you have two main strategies:

1. Get *really* good at your skill set and figure out how you can add value above and beyond the automated tools.
2. Go into fields that are less likely to be disrupted, such as the trades (electricians, plumbers, etc.), certain areas of medicine, or even dentistry (Can you imagine ChatGPT ever drilling holes in someone’s teeth?).

If you attended a 4-year university, strategy #2 is probably not appealing…

…but, honestly, I think it’s a reasonable approach.

Careers in medicine have plenty of issues, but if you asked me to make a list of “high-paying jobs that are less likely to be disrupted by AI,” surgeons would be near the top.

If you’re not interested in physical labor of any sort and want to continue in industries like finance, tech, or law, all I can suggest is strategy #1.

This might take the form of developing better judgment, better relationships, or more creativity.

For example, could you propose a deal structure or a potential buyer that no automated tool would “think of?”

Or could you find clear problems in auto-generated models and fix them more quickly than other people?

But I don’t know; I’m just making guesstimates here.

I don’t think there is a great solution to any of this, and I expect some rough times over the next few years.

Why I’m Taking a Fairly Relaxed Attitude

You might read all this and ask:

“How are you so relaxed? This shift toward automation might destroy your entire business since no one will buy modeling courses anymore! They can just enter a prompt and get everything!”

Good question.

First, even if the work becomes more automated, you’ll still need to understand how models work to check everything.

A model that is “90% correct” is 100% useless in many cases.

Maybe you’ll be doing less data entry/updates and formula entry, but you’ll need to be much better at understanding the output.

But if I’m honest, there’s another factor: I never expected this business to last this long in the first place.

When I started M&I and BIWS around the time of the 2008 financial crisis, I expected to run them for 3 – 4 years and then get another office job or launch something new.

I never expected to be here doing the same thing almost 20 years later.

I assumed that new technology would make Excel skills irrelevant or that students would lose interest in finance careers.

But these didn’t really happen – or at least, not to the extent I expected.

Since my expectations were low, I view the continued interest and sales from ~2013 onward as a “discretionary bonus.”

If this bonus eventually stops, it will be disappointing, but not the end of the world.

Maybe I can even get an AI agent to come up with some new business ideas and launch Version 1.0.