AI + Financial Modelling
I'm an ex IBanker turned corporate strategy at major telco. I've been tasked with making our team as AI pilled as possible as the organization wants to be on the frontier of AI and automation. We have CoPilot in Teams, agent builder and CoPilot in Excel. My perception is that there was an initial wave of FOMO with AI, boards pushed big spends driving enterprise uptake on CoPilot/Claude/GPT which boosts their top line revenue but, outside of software eng and acting as a more robust Google Search, email drafter and verbose legal contract interpreter, I have not seen meaningful uptake to completely replace entire manual workflows. Granted I think the incentive to streamline and automate work is less in a 9-5 versus IBD where time is of essence, so perhaps the use cases I am seeing are less creative thus far.
I'm looking for inspiration on ways it has been successfully integrated into analyst style workflows in Excel such as financial modelling - where do you find it useful, where do you find it weak and what tools - Tracelight, Shortcut, Endex, Claude in Excel, CoPilot, others? - have you found useful... or perhaps you have had issues with it, or you think it's better to do things by hand?
I’ve used ChatGPT, CoPilot and Claude. Claude is by far the best, hands down.
I moved from Banking to strategy at a Consumer brand that is all-in on AI and I use Claude every minute of my working day. I no longer update ppts manually, dashboards come together just by talking to Claude code for 5min and I automate everything I was doing in excel with /skills so eventually all of my repetative workflows will be plugins.
I work in a generally project based role so anytime the CEO gives me a whimsical insight he wants analyzed, I just talk (yes talk) to Claude CoWork pointing to a folder dedicated to Claude projects and ask it to help me figure out how to make it happen. If a dashboard is the best output, it will give provide me with a .md file or prompt to put into Claude Code.
Eventually, you can pair it up with a web hosting server to share it with other people in the company to update the sources etc. over time.
So many use cases. Too many to write avout
Super interesting. Thanks for sharing.
When you say dashboards - do you effectively mean take a raw dataset and transpose it into a set of charts/tables/outputs that enable you to engage with and understand the data? Have you tried any integrations with PowerBI etc where you can automate the build of the front end?
Also curious to know beyond dashboards in Excel, have you had any experiences using it with any other tasks like financial modelling for example?
bump
I mean Copilot is cooked from the beginning tbh. Used to work at McKinsey and I definitely got addicted to Claude Opus. I think it got to a point where I could not do anything without it anymore. It is ridicously good at problem solving in the corporate context and also helping you write pages. Literally wouldn't write a single paragraph myself anymore. Give it the context, tell it how the page should look like and it does it.
For more meaningless stuff (i.e., pitches) I even managed to create templates for deliverables (i.e., 4-5 pages that have to look the same in which you describe a bit the context on the LHS and then on the RHS how it will look like in the future at the client). I build one of those with Claude (on cursor) then gave it the template and said now replicate this 5 times for other deliverables (all deliverables where I really had limited background, some growth, marketing & sales transformation).
It took it like 10mins to build those pages. Associate partner asked me to do that until EoD so I quickly reviewed the pages, went to the Gym and sauna and slotted them in 3-4 hours later.
He was very happy with limited comments and so was the client.
In other words, consulting is completely cooked (my thesis).
Amazing but also scary. So am I correct in saying main use cases is on the PPT / templatized presentation side where you basically give it the instructions on how to fill the template + the template, then each time you want a new PPT, you feed it the context specific to that engagement and of it goes?
Any thoughts on AI in Excel specifically? Have you found it useful in the spreadsheet or data or modelling workflows?
I would not say the main use case is PPT. The main use case is research + problem solving. I.e., you have to build a complex market model forecasting the end-market demand for some random niche industrial products in Latin America. Problem solve with claude!
What is the approach? How could we build up a driver tree? What are the relevant end-markets?
Done that; great. Now please guide me to all the relevant sources I could use ... oh there is this super interesting price-index I would have never found manually? Claude just found it on some national statistics bureau web-site.
Now something more complex. I need to extrapolate the market size figures to other markets. How would we do that? Lets do it by ratio of manufacturing output vs. my focus market + some qualitative adjustment factor. I don't want to do it myself now, so please Claude go ahead and build a nice model for this. 10 mins later you will have a clean "model" with extrapolation factors for every region + a crisp rationale as to why you adjust it higher (lower) than manufacturing output suggests.
Now for powerpoint it is also good if you have to fill pages with contents. Don't know how to write, well let it produce a draft. Most of the time the output is quiet good of course with some tweaks that you need to make here and there.
Has anyone been able to get it to work well for actually modeling in excel?
I’ve worked with Claude via excel plugin and only had issues. Can be helpful for quick setups of feeder tabs or side analysis or sample unit economics but in terms of actual model construction I’ve only had issues. Curious if others feel differently or have had success with things like Shortcut
Thanks for sharing. What do you mean by quick setups of feeder tabs, side analysis and sample unit economics? And what type of issues have you encountered?
played around with Claude (chat, not an excel plug-in) to calculate the unit economics of a very famous private company, using quite detailed instructions.
What was good: it found really useful sources for the inputs very quickly.
What was bad: broken links, a few logical errors (surprisingly!), and an awful modelling technique. Ca. 50 rows of modelling were required.
That's super interesting, thanks for clarifying
Claude is finally usable with cowork. What I’ve had our intern work on this summer (five man shop) is systematizing our models (both deal diligence / execution and portfolio valuation) into portable flexible Claude skills. Started with the simplest stuff and it’s done well. You need to give it good templates to work from + the source data in the same folder. Auditing outputs heavily to make sure it gets identical results to our handbuilt models. But WOW early impression is it can compress 5-6 hours of modeling into 30 min. Massive time save standing up a model where now all my time is spent flexing scenarios and assumptions rather than adapting our templates to the deal. Huge time savings on those workbook spanning tasks like linking up sheets of source data. Highly recommend. It’s been accurate with the right instructions and I’ve checked it closely.
Does decently well with written instructions but much better if I just talk into Wispr for 10-15 minutes about what we’re doing and point to what it should reference. The context helps and I have it keep a markdown file of modeling best practices so it learns our style over time. Claude code has been fantastic too we are building an application that runs comps and cap table waterfalls. Gonna try out the tag too.
That's super interesting, thanks for sharing. So just like with the grads, it's all about how much time you invest in sharing context and training them to get it up to speed. Are you in sell-side or buy-side? I'm curious to know what your use case of modelling is. Sell-side it seems like modelling was more about the user understanding the mechanics and outputs so they could explain the numbers to the MD/Client - the model itself was a means to an end. On that basis, I'm wondering how you think about understanding the logic behind the numbers in the outputs if Claude is pumping it out? Or in this scenario is it because you already understand the logic with years of experience that you are encoding into the skill that you are happy to automate the process away and just check the numbers? If it's the latter, how do you think about training the next generation so they have the same intuitive understanding of the numbers that you do, and can understand the calculations and mechanics to explain and check numbers?
I’m in my fourth year on the buy side. Very lean shop so this was just necessary to speed up our early review of opportunities. We get a few data rooms each month and need to quickly identify if it’s viable for our underwriting. Once we’re looking deeply I’m still building custom stuff. There’s no one below me but if I had an analyst I would have them rebuild all our templates to get familiar before plugging them in on tweaking automated workflows. I agree you need the understanding and it’s a good reason to start as a banker.
I tried using AI for formula building. It gave me a thousand unique formulas instead of one flexible one. Completely broke the model. Learning to prompt better now.
That's interesting... did you use =copilot() or was it via the chat bot feature and asking it a question in natural language? What type of formula was it- something abstract/complex?
Deleniti delectus eius nesciunt labore commodi aliquam dolorum. Libero nulla nobis vero fugit exercitationem facilis odit nemo. Et neque assumenda suscipit aut. Ipsum est doloremque aut ut qui. Est cumque voluptate velit quia quae. Sed alias quo et quae omnis.
Omnis impedit amet impedit quae molestiae. Vel et architecto quos. Dicta aliquam modi autem. Nostrum totam unde laboriosam aut minus aperiam. Doloribus ipsum deserunt quae.
Eos et veniam voluptatem aut. Corrupti est tempore ducimus non aut dolorem. Ipsum harum ut inventore corporis non. Est voluptatum vitae facilis quia. Eum soluta sed quo aliquam voluptas possimus reiciendis. Voluptas impedit non ducimus. Ut ut repudiandae rem molestiae necessitatibus qui.
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...