New analyst: how are you thinking about learning to model when AI is getting this good?
I started as an analyst a few months ago. One thing has been on my mind that I'm curious how others are handling.
AI tools like Claude are already very good at financial modeling, and even faster/better than me and other analysts. I can't really picture where they'll be in 3 years
I want to become good at all kinds of modeling, not just valuation. But my worry is spending hours and hours grinding at it, only for that skill to become commoditized and handed off to AI entirely a few years from now. I know you need to understand a model to check it. But I'm not sure how to think about where my time is best spent.
For newer analysts: how are you approaching this? Still grinding through building everything yourself, or using AI and focusing on understanding and checking the output?
For more experienced people: what do you think stays valuable?
I want to make sure I'm putting my time into skills that will still matter. If I'm thinking about this the wrong way, please don't hold back. I'd rather hear it now.
The concern about AI overtaking financial modeling is valid, but here's how you can approach it strategically:
Focus on Understanding, Not Just Execution: Based on the most helpful WSO content, the value of modeling isn't just in building the model but in understanding the why behind it. For example, knowing how a 100 bps change in rental growth impacts IRR versus a 50 bps interest rate jump is what separates a skilled analyst from someone who just inputs numbers. AI can automate processes, but it can't replace the judgment and intuition required to interpret results and make strategic decisions.
Build Models from Scratch Early On: Experienced professionals on WSO emphasize the importance of building models manually for your first 10-15 deals. This process helps you internalize the mechanics and develop an intuitive grasp of how inputs affect outputs. Even if AI tools are available, this foundational knowledge is critical for spotting errors and understanding nuances.
Leverage AI as a Tool, Not a Crutch: Use AI to speed up repetitive tasks or generate initial drafts, but focus your time on reviewing, refining, and understanding the output. This ensures you're not just relying on AI but actively engaging with the work.
Develop Complementary Skills: As AI takes over repetitive tasks, skills like critical thinking, deal structuring, negotiation, and understanding market dynamics will become even more valuable. These are areas where human judgment and creativity shine.
Stay Adaptable: The finance industry is constantly evolving. By staying curious and open to learning new tools and methodologies, you'll remain relevant even as AI continues to advance.
Double-Check and Validate: A recurring theme in WSO advice is the importance of accuracy. Always double-check your work, whether it's AI-generated or manually built. This habit will set you apart as someone reliable and detail-oriented.
In summary, don't shy away from learning to model. Instead, focus on mastering the logic and decision-making behind it. Use AI to enhance your efficiency, but prioritize developing skills that AI can't replicate, like strategic thinking and market insight.
Sources: Scared Shitless: How Did You Actually Learn How To Competently Build Financial Models, Acquisitions Analyst Advice - New Analysts, https://www.wallstreetoasis.com/forum/hedge-fund/machine-learning-taking-over-hf-research-analyst-roles-in-near-future?customgpt=1, AI in fundamental investing
I'd still learn to build models from scratch. Not because I think manually linking cells is going to be some irreplaceable skill in 3 years, but because it's hard to catch a bad assumption if you don't understand how the model actually works.
The bigger distinction is between knowing how to build a model and knowing whether the model is telling you anything useful. AI can get you to an answer faster, but it doesn't make the underlying assumptions, accounting, or business judgment any less important. Garbage in, very efficiently garbage out.
I'd use AI aggressively for the grunt work, but make sure you can still build the thing yourself, explain every major driver, and tell when the output doesn't pass the smell test. The analysts who probably have the most to worry about are the ones whose entire value-add is turning comments into Excel formulas. That was never a particularly durable moat to begin with.
Thanks for taking the time to answer. How do you use AI? What type of grunt work does AI do that makes you more efficient? Does your bank allow you to have AI in Excel, PPT, etc?
If not, does Co Pilot help? If yes, how?
Are you suggesting that if you get a comment you disagree with, you should flag this? Confused as to how there’s any other option than turning a comment into an excel formula when your job is to do what you’re tild
How is an unattributed quote by someone that doesn’t say much interesting? And who gives a shit if it can pull a model together?
Stop tagging me, retard
I should sincerely hope that a piece of software is faster at modeling than a junior analyst. Also, simply stating that something is interesting doesn't make it so, and anyone with the tiniest bit of critical thinking understands that (which is exactly why you don't, you've stopped thinking and pretended like using a chatbot is the same thing). Who is this random internet stranger you are quoting? How experienced or good are they at their job. How easily impressed are they? Have they put any critical thought into the product they're claiming is so good? As is pretty clear, the person you quote clearly conflates the concept of "good modeling" with doing it fast. I remember my banking boot camp, where the rock stars were the people who could build DCF models in like 90 seconds. Who cares? Why is that a valuable skill for an AI? It impresses someone whose job is to build lots of mostly boilerplate spreadsheets, but what value does that provide?
As I say, over and over and you repeatedly refuse or are unable to answer, how does this justify the trillions and trillions of dollars being spent on it? "Oh we've automated away 5% of an analysts job!" Wow!
And we can have a decent idea where AI tools will be in 3 years, by looking at where they were 3 years ago. Pretty much in the same spot.
Too many extreme one-sided takes – it will not be in the same place in 3 years from now lmao. Agree with OP that modeling in and of itself is not as important; but the second and third order thinking will become all the more relevant and required earlier for bankers. Analysts usually just do what they are told and build, but now that AI can do that part, think onus is increasingly on analysts to think thoughtfully about assumptions and the story – whereas that would usually be the associate / mid level responsibility.
It is like learning how to spell. I use autocorrect for everything - but obviously fucks up quite often. As mentioned earlier the main focus is for you to be able to derive actionable insights from said models, even if you realize you inherited a frankenstiened piece of shit, you have to know it’s limitations and be able to confidently speak to the output. The human skills will be that much more valuable imo.
You're a few months in. Of course Claude is better at modeling than you. So is every decent second-year analyst.
Build models from scratch while you're junior. The point isn't to become a human Excel macro. It's to learn what drives a business, which assumptions matter, and when the numbers are bullshit.
Use AI as a coach. Build the model yourself, then have Claude tear it apart. Ask it to find mistakes, challenge your assumptions, and tell you what you missed. Don't have it rebuild the model for you.
Get your reps in. That's what the analyst years are about.
Chat GPT mogs
The reality is that profit maximizing firms will want things done as quickly as possible, whereas your instinct is correct that building things from scratch and or updating manually (since let’s be realistic no one ever builds anything from scratch in the wild outside of rare circumstances) is the best way for you to actually learn. You should spend time on a case by case basis deciding how much deadline flex you have relative to how much time you’d like to spend learning. One thing that will definitely make this harder and probably entail a need to spend your own time getting good is the almost certain increase in velocity of deliverables. Personally I’m already experiencing seniors / clients asking for stuff at 3x / 5x the turnaround time before. Manual scrubs of information that used to take two days now take Claude 2 minutes to pull and a junior 30 / 60 minutes to scan. That doesn’t now allow me to spend a day genuinely understanding the work product becuase I am now juggling 4 of these asks.
It remains to be seen how long junior human in the loop is needed to clean and audit things, but that’s certainly going to be the next 12-24 months. Realistically, once the data gets so good juniors don’t need to review, really forward thinking clients are going to start wondering if you even need the senior to explain it versus just plugging into GSAI-API on your corp dev side. Of course, the usual caveats about relationships are what really matter for seniors, etc. apply.
Sequi sit provident recusandae aut quia unde doloremque aut. Consectetur consectetur odit dolorum earum ea.
Repellendus voluptas fuga praesentium dignissimos voluptatibus ut. Ipsam expedita quibusdam maxime est veritatis eligendi harum. Ut qui perferendis voluptatem temporibus facere blanditiis velit.
Incidunt corporis at architecto tenetur voluptatem voluptas. Est eligendi numquam et iusto. Quia et voluptatibus sint velit officiis nobis in.
Velit sit ut dolor enim aut omnis. Ipsa qui vel enim ut ducimus tempore velit.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...