Will AI Kill Spreadsheets?
Andrew Chen vs. The Internet
What will happen to spreadsheets in the age of AI?
Already, a transition is underway. The big AI labs are building their own integrations for agents in Excel and Google sheets. Startups are getting in on the action too, with teams like Fundamental Research Labs (SR002) earning traction by reimagining the spreadsheet experience to be AI native.
Last week, a16z speedrun General Partner Andrew Chen kicked a hornet’s nest on X with a post arguing that “anything that is currently modeled as a spreadsheet is better modeled in code.”
Here’s the full post, to save you a clickthrough:
AI code gen means that anything that is currently modeled as a spreadsheet is better modeled in code. You get all the advantages of software—libraries, open source, AI, all the complexity and expressiveness.
Think about what spreadsheets actually are: they’re business logic that’s trapped in a grid. Pricing models, financial forecasts, inventory trackers, marketing attribution—these are all fundamentally *programs* that we’ve been writing in the worst possible IDE. No version control, no testing, no modularity. Just a fragile web of cell references that breaks when someone inserts a row.
The only reason spreadsheets won is that the barrier to writing real software was too high. A finance analyst could learn =VLOOKUP in an afternoon but couldn’t learn Python in a month. AI code gen flips that equation completely. Now the same analyst describes what they want in plain English, and gets a real application—with a database, a UI, error handling, the works. The marginal effort to go from “spreadsheet” to “software” just collapsed to near zero.
This is a massive unlock. There are ~1 billion spreadsheet users worldwide. Most of them are building janky software without realizing it. When even 10% of those use cases migrate to actual code, you get an explosion of new micro-applications that look nothing like traditional software. Internal tools that used to live in a shared Google Sheet now become real products. The “shadow IT” spreadsheet that runs half the company’s operations finally gets proper infrastructure.
The interesting second-order effect: the spreadsheet was the great equalizer that let non-technical people build things. AI code gen is the *next* great equalizer, but the ceiling is 100x higher. We’re about to see what happens when a billion knowledge workers can build real software.
The post racked up over a million views and hundreds of replies, with not everyone in agreement.
Richard Pham argued that much of the value of spreadsheets is in the grid-based visualization itself:
Allie Harris came at this question from a different angle:
Tom Robbins was skeptical as well, pointing to the continued ubiquity of PDF files:
This was a a comparison Andrew was happy to run with:
The question is whether spreadsheets will follow the same trajectory—still around, but no longer the default way a billion people interact with data and business logic.
An interesting framing in the thread came from Dan Hockenmaier, who pointed out that there are really two kinds of spreadsheet use. The “mini software” kind—dashboards, trackers, attribution models—is almost certainly getting replaced. But the other kind where you develop a financial model you build to hone your own understanding of a business is stickier.
One top reply came from Robert L. Peters, who offered a full-throated defense of the spreadsheet:
Would bet my net worth with 100x leverage that front office finance jobs will still use spreadsheets 10 years from now
Can you run an LBO model with python, audit each cell (F2), make micro adjustments to formulas with instant feedback, toggling an assumption back and forth to see how it impacts a sensitivity table that is 1000 calculations later in the chain? And do this with someone who just got out of college a couple months ago?
The act of making the spreadsheet in many cases forces the user to think about assumptions, and reason about the drivers. You may not always have it fully formed in your head in advance as a prompt until you see the data, create an operating build, etc.
If an MD draws on their iPad “make a box here, increase spacing in these three areas, change the calculation for this one year of EBITDA to xyz, and make an italics EBITDAR line underneath, break out product lines in revenue build, shift to monthly, etc” they can give to an analyst and can get back exactly what they want to see in an hour. They don’t need to wait 12 months for a SaaS platform to maybe add the feature. Or prompt an LLM that may asymptote on what you want but isn’t perfect.
There are also so many edge cases for bad data from private companies where you really need to manually look at the data points to understand the gaps. Python doesn’t allow you to do that and prompting doesn’t provide a visual surface.
I’m sure you will be able to prompt a lot of this stuff but for many use cases the form factor will continue to be a spreadsheet - the analyst will be able to make tweaks to the spreadsheet output the LLM makes, check the logic easily, make edits where it’s faster to make it the way you want vs type out prompts, etc
In front office finance jobs (maybe other than VC which is less precise with modeling), one single calculation error can dramatically change the answer and cost you $1 million or $1 billion. 95% right is 0% useful. Python and prompting do not allow for full verification of the calcs and data through to the primary sources.
Tech ppl perennially have a fundamentally flawed mental model of Excel (“why not use Python????”) because they have never had exposure to how it’s used in the real world, and it’s why so many companies with the business model “move your OUTDATED spreadsheets to my MODERN B2B SaaS in your browser!” that for the past decade have spammed PE firm inboxes have crashed and burned
Spreadsheets are a better form factor than python for business users and LLMs won’t kill spreadsheets
Peters’ reply gets at something the original take arguably doesn’t address: The act of building the model is how you develop conviction about the inputs. You can’t just skip to the output.
Chen pushed back:
What you’re talking about is the “IDE” user experience of spreadsheets -- being able to inspect the logic, checking out values/cells/etc, debugging, etc. I both agree that you’ll need to do that BUT also that this is all possible, and frankly, better, in software as well
In software you also get unit/integration tests, version control, coding harnesses, AI everything, and much more. You could even ask the LLM to describe the code in a grid if you want to try to be quasi-backwards compatible.
Chen is saying those workflows migrate to a better substrate.
But you can’t win ‘em all. Joel Grus came off the top rope with this zinger:
So, after a day of fighting in the reply guy trenches, Chen posted a followup (full text below):
Reading the replies -- a lotta folks hate this prediction!
A lot of folks can’t imagine programming the logic/variables/inputs without the spreadsheet grid paradigm
My primary counterpoint:
Much complaining is from people trained on keyboard shortcuts in Windows Excel on Thinkpads during their glorious banking days, swearing that they’ll never switch to anything else. Late adopter normie finance bros. Soon to be disrupted
My actual counterpoints:
Programming itself has changed its UX many many times. Punchcards, typing into files, IDEs, and now LLM coding harnesses. Spreadsheets are not the only way to encode business logic -- there are better ways, while gaining all the power of software
The grid UX might remain in some form, but might be more of a display. Just as you code in Codex/Claude but then still pull up a webpage. Or maybe you’ll have a grid as a DB but then build apps on top, but still want a querying UI for the data
LLMs will make going between logic in code and logic in spreadsheets interchangeable. So maybe you’ll edit in a grid but then hit “deploy” and it’ll build a webapp in the cloud. And just as we have VLOOKUP() they’ll be LLM() that can encode AI logic
Anyone who works with software knows its infinitely better and more expressive and more powerful. AI code gen is a blessing for all the non-technical excel wizards who can now take their work to the next level
The history of computing really is a story of interfaces evolving beyond what the previous generation thought was possible. Punchcards felt essential until they didn’t. Command lines felt essential until GUIs arrived.
The spreadsheet grid might be next.
What do you think about the spreadsheets debate? Let us know in the comments. And for more weekly dives into the world of early stage startups, subscribe below.
The "pro-code" camp is right about the flaws. Spreadsheets lack version control, they are prone to "broken link" nightmares, and they struggle with scale. But the "pro-grid" camp is right about the human element. You cannot "feel" the data in a Python script the way you can in a cell.
At NeutronTech (neutrontech.ai), we agree that spreadsheets are often "janky software," but we also believe the grid is an irreplaceable cognitive interface for auditing and iteration.
We didn't build a better spreadsheet; we built a software engine that looks like a spreadsheet.
- NeutronGrid isn't a web of fragile references. It’s powered by a data engine.
- The Grid is an IDE. We kept the grid because it is the best UI ever designed for data interaction. It’s not "shadow IT." It’s a visual interface for high-performance software.
- The AI is a bridge. Because it is structured like real software, our AI doesn’t just "hallucinate" numbers. It uses the dedicated data engine, validates against 3,000+ tests, and provides answers you can actually trust.
- It's built for Apple Silicon. We optimized by using Metal -- for real GPU/hardware accelerated analytics and 20 chart types with 3D rendering.
It's wicked fast, AI understands your cell reference and data structure, and it's currently in beta test.
Andrew Chen is directionally right—but the future isn’t spreadsheets vs code, it’s convergence. Tools like NeutronGrid from Neutrontech.ai treat the grid as an interface, not the logic layer—combining the inspectability and intuition of Microsoft Excel with the power, testing, and scalability of software. Instead of replacing spreadsheets, this approach compiles them into AI-native, auditable systems—preserving how humans think while unlocking what code can do.
-Grid as an interface, not the source of truth
-Code as the execution layer, not the user experience
-AI as the translator between intent, logic, and output.
NeutronGrid is effectively turning spreadsheets into:
1. Inspectable, testable systems (like software)
2. Composable building blocks (not fragile cell webs)
3. AI-co-authored models that remain fully auditable.
We're just few days away from launching this revolutionary platform.