Excel Forever
Claude, Excel and the Future of Market Intelligence
“[Excel is] like this interactive toy, where you can kind of program it without knowing you’re programming.” — Ray Ozzie, former Chief Software Architect, Microsoft.
As an analyst, I was always good at spreadsheets. Not World Championship good, but I could knock up an earnings model fairly quickly. Before grasping that less can often be more, my spreadsheets could get quite unwieldy. The largest still sitting on my hard drive is a 10MB monster that spits out earnings estimates for Deutsche Bank while ingesting 25 years of history across five business segments.
As my friend Stephen Clapham points out, it can be oddly therapeutic bashing numbers into a spreadsheet. But it isn’t all fun, so when Anthropic released an add-in that integrates its AI assistant Claude into Excel, I thought I’d give it a spin. “Claude listens carefully, follows instructions precisely, and thinks through complex problems,” its marketing promises.
I chose Factset as the target of my analysis, an $8 billion company listed on the New York Stock Exchange whose stock is down 46% over the past six months because … well, we’ll get to that later. Opening a blank workbook, I told Claude to produce me an earnings model. I wanted ten years of history and three years of forecasts, and it goes without saying that it should look pretty.
What unrolled before me was magical – like when IT support takes control of your computer and your cursor flies across the screen, possessed of its own intelligence. Within five minutes, Claude had visited the Securities and Exchange Commission to fetch historical data and populated a very simple model forecasting revenue and operating income using company guidance that it directly sourced.
As a first draft, it was a bit lacking in granularity, so I asked it to analyse key earnings drivers and try again. After navigating usage limits (which I spent $65 to overcome) and context limits in the chat (frustrating, but since solved via the Claude Opus 4.6 update), Claude produced a serviceable model in the time it took me to finish a cup of coffee. Without knowing it, the model settled on earnings estimates that fell within 2% of consensus. That may tell you something about the way analysts generate earnings forecasts: extrapolating growth rates and margins is not a difficult algorithm to automate. Also, Factset is a relatively simple company to model – I have yet to set Claude loose on Deutsche Bank. But as a first attempt, it proved remarkably capable (see what you think below the paywall).
I’ve written about Excel here before. When I started out in equity research, it was ten years old and there was just one holdout in my department, a European autos analyst, who remained loyal to pencil and paper. Decades on, it is ubiquitous among financial analysts. Spreadsheets themselves have become commoditised along the way – firms make them readily accessible and few analysts derive competitive advantage from them – but they embed a common language for analyzing and thinking about company performance.
If there’s one drawback, it’s that while Excel makes its data visible, its formulas are hidden. That’s fine for visualizing how variables interact – like client numbers and annual subscription values in my Factset case. But when the data gets so big that you can’t look at it all, or when a spreadsheet becomes infested by circular references (we’ve all been there), it makes more sense to look at formulas. Trouble is, few financial analysts are comfortable venturing into an integrated development environment (IDE) where the formulas are exposed but the data hidden. And even if they are, that’s not the common language of financial analysts.
This is where Claude comes in. Nick Lin, head of product for Financial Services at Anthropic, has this to say: “Our file creation feature that was launched a few weeks ago [in October] that enables Claude to create Excel documents and PowerPoint is essentially Claude accessing a virtual machine within which it can run Python code at scale to edit, analyze, and create Excel documents and create these perfect DCF models.”
In particular, Lin recognises that Excel is where finance professionals live and so rather than disrupting it, he needs to accommodate it: “Financial models themselves, they’re not just these beautiful Excel sheets, right? They’re a way for finance analysts to inject their own judgment of what the future looks like and what the proper valuation looks like for that company, right? So, with that in mind, we want Claude to be really good at understanding these core finance concepts and manipulate systems like Excel and spreadsheets to be able to do that calculation.”
So could this be the killer app to inject AI into finance workflows? This week, software stocks – including market intelligence names like Factset – have cratered as investors question whether AI renders their business redundant. The launch of Claude Opus 4.6, with its expanded context window and improved reasoning capabilities, only sharpens the question. To explore what this all means for financial analysts, for Excel and for legacy tools like Factset, read on.