Continuing the discussion from Do you feel it’s important for younger generations to be able to read an analog clock? If so why?
More curiosity as someone who doesn’t create spreadsheets much -
To those who do -
How much are you entering in formulae now vs just telling Claude or CoPilot what to do?
How do you see AI changing your process?
I enter the formula myself.
I feel way more confident in my ability to tell Excel what to do than in my ability to tell AI to tell Excel what to do.
In my field (web software), I only occasionally use spreadsheets at work. I’ve experimented with Gemini in Google Sheets a few times, and it’s OK. Sometimes it works. It’s nice when I can just tell it “I need to find the corresponding Y in Sheet2 based on the X in this range” and it’ll write a vlookup or index/match for me. Makes creating pivot tables a bit easier too.
But sometimes it’s completely off-base. Sometimes it goes and fetch completely irrelevant things from my other Google things (looking for PDFs in GDrive, etc.).
Overall it’s not a great experience. Feels like a shimmed-in afterthought.
Sometimes I just manually prompt ChatGPT instead, like “How do I ______?” and that usually works very well, but takes more effort.
In terms of models, I think Gemini and ChatGPT are pretty close (especially for simpler spreadsheets and trivial formulas). In terms of the user-facing product though, I vastly prefer the simple elegance of the ChatGPT desktop app vs the diarrhea trails of Gemini leaking through every Google service even if you don’t want it.
Maybe it’s better in Excel & Copilot (or not? no idea), but I certainly wouldn’t trust Microsoft (of all people) to do it better, and I wouldn’t pay for MS Office just to find out. Like Google, Microsoft’s approach to cramming in Copilot everywhere is a huge turn-off, and I just reflexively disable it anywhere I can now. Microsoft’s been on an “AI everywhere” quest as of late, and my answer to that has been mostly a shrug and a “fuck off”.
But, overall, I think the OP kind of assumes the premise in that in the age of AI, there is less of a reason to reach for spreadsheets to begin with. It’s like asking it “How do I write this paragraph in Microsoft Word?” when it can just write the entire chapter for me instead.
There is a class of problems where I’d probably just reach for AI to start with, altogether bypassing spreadsheets, and just ask it to either directly analyze the sample data (if it’s simple enough) or else work with it step by step to write a simple program (like in Python or Javascript) to analyze it and explain the math along the way — that’s a lot easier to keep track of (for me at least), algorithmically, than trying to trace every cell connection in a multi-page spreadsheet. So in that sense, AI is changing the process even more in favor of code vs spreadsheet for structured data analysis.
But that’s just my field. The last time I seriously used spreadsheets was in school (actually, mostly in middle school, where we did a semester on Excel). Since then, I casually use GSheets maybe a few times a month. Haven’t had to work with a Excel file in a decade or so now.
If I worked in finance, accounting, etc., I imagine it’d be quite different.
If I enter the formula myself, I can be confident that I entered it correctly. I’m not handing that off to something that thinks “strawberry” has two Rs: AI’s mistakes are opaque, and I’ll be unlikely to catch its hallucinations in a dataset.
If I need a more complex formula than what I currently know, I Google instructions for writing that formula: by learning how to do it myself, I’ll gain a deeper understanding of the data and of the analysis I want to conduct.
That said, I work on spreadsheets at a level above most casual users, and far below most professional users. If I didn’t understand basic formulae, or if I understood complex coding, my answer would probably be different. But for the level I’m at, entering the formulae myself makes the most sense.
Slight tangent:
Also, for the use case of “I need somewhere to store a table of structured data” (but not necessarily to do serious number-crunching, more just to store them and look them up in relation to each other), there are also a class of products that are kind of a mix of spreadsheet and database-lite, like https://www.airtable.com/. Those kinds of services better fill a lot of the niches that simple spreadsheets struggle to meet, and are especially useful for the non-Microsoft world where Access isn’t usually available or desirable.
Coincidentally, they also tend to integrate better with external services and AIs, because 1) they’re built with third-party integrations in mind, and AI is just another integration and 2) they’re not built by Microsoft, who caters more to bureaucratic enterprises than small businesses and startups.
Many of the people I’ve worked with as colleagues and customers would reach for a tool like this rather than a pure spreadsheet.
I generally live in spreadsheets at work, and I manually deal with the formulae. I have consulted with gemini when it comes to doing any scripting work to automate the updating of those spreadsheets however, but it has missed some subtleties in selecting correct add-ons and libraries that I’ve had to go find out for myself.
Apologies for possible hijack, but what are we using as the definition for “spreadsheet”?
Until the other day, I did not realize the term primarily related to computer generated charts/records - like Excel. In my ignorance, I assumed the ledgers I studied in accounting classes were spreadsheets. As were the charts I compile using various word processing software, using tabs to create columns. I would have thought that the stock market pages and sports boxscores in the newspapers of decades ago were “spreadsheets.” Is these misuses of the term?
IMHO This is a similar workflow to asking an AI which formula to construct — as the first step, not the final word. It’s how I prefer to use AI, just as a replacement for a search engine and not as something that takes over the whole workflow.
It’s really, really good at “translating” a natural-language construct like “how do I get the last four characters out of this column in this other sheet, then hyperlink it” into its spreadsheet equivalent, and saves time vs having to manually navigate through the SEO minefields and blogspam hellscape that is Google in 2026. It’s trained on all that anyway, and this is the kind of thing LLMs are actually good at doing, vs trying to count letters in a word (which just isn’t how their tokenizations work, so it’s kinda a worst-case scenario for them).
Anyhow, this still has a similar learning effect of letting me know about formulas which I might never have known about (while also being very good at explaining what they are and how they work), but without replacing manual entry altogether. I’m still the one who enters the formula, now that I understand how they work.
But, I suspect people like you and me are not long of this world. The AIs are the bleeding edge are all semi-autonomous agents now that do all the work on their own, in lieu of other software (and the people who would’ve operated them). In those workflows, the people (if any) are more like shepherds than knowledge workers.
It could just be an organic evolution in usage over time? Wikipedia differentiates “spreadsheets” (computerized) from “worksheets” (dead trees), but that might just be pulled out of some editor’s ass 
The cited source for that statement is just some dubious blogspam article.
In the OP’s context:
It’s clear that they were asking about the computerized version.
But maybe that wasn’t always the natural assumption? I’m 40-ish years old, and as far as I’ve been alive, spreadsheets were always computerized, even when I was a kid working in Lotus 1-2-3. If there was a table on paper, I’d just have called it a table (or more likely, a “printout”).
If that wasn’t the case in your experience, and you remember a time when spreadsheet more generally referred to paper documents of various sorts… for us youngins, it would be a very interesting anecdote to hear! You should share your war stories (but maybe in a linked thread?)
EDIT: Armchair etymologists debate the origin of the word here: https://english.stackexchange.com/questions/541763/etymology-of-spreadsheet-an-anachronistic-use and the top-rated answer is particularly informative (that it refers to a specific sort of wide paper layout where a ledger is easily visible in the full).
I am a software engineer. I am competant in C#, VB.NET, Java, PHP (alas) and a variety of now dead languages.
This is the the biggest mistake, in my opinion, since Microfoft allowed ODBC connections to MS Access databases.
What is the “this” here? That Microsoft is trying to integrate AI into its spreadsheets, that people are trying to use AI to learn spreadsheets, or both…? Or something else?
“On two occasions, I have been asked [by members of Parliament], ‘Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?’ I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question.”
― Charles Babbage
A lot of people today seem to have the same misapprehensions about AI today that, allegedly, faced Babbage about his famed difference engine.
To the extent that AI can help me with how I want a spreadsheet do what I want, it’s a marginal improvement over Google but most of my time spent with spreadsheets is figuring out what I even want in the first place and that’s a non-trivial process of discovery and experimentation.
I’ve never used Claude Code inside Excel, but I’ve used it to write other software and it’s really amazing.
When I have to track down some really complicated thing in Excel, I turn to Claude and it writes the most elegant formula, using the most current (very powerful) new functions. I imagine Claude Code in Excel is even better.
Connections between Excel and Access are really handy and powerful when dealing with moderately large datasets (larger than Excel can comfortably handle, but not so large that Access starts to choke). Get the data into Access, have it do the organization, and return the results to Excel.
This complaint is at least a year out of date. Current AI implementations do hallucinate sometimes of course, but they are much more powerful and consistent than they were just a few months ago. Just getting better and better.
I know nothing about your spreadsheet knowledge, so forgive me if I insult you by mistake.
Everything, in my opinion that can go in a “real” database should be in a real database, either SQL or NoSQL, as is the need.
While Access and Excel have “SQL like” query languages, which are reasonably close to T-SQL, I … have reservations. I have seen insanity in TSQL stored procedures, but at least it was written in multi-line T-SQL, which, despite T-SQL’s apalling debug ability, do allow sone accountability.
Excel’s single line input is also a bastard. I typically write SQL Stored Procedures over at least 20 lines of code, whereas Excel can get realy complicated.
Excel, is of course supported by ODBC, so I can write code that references a specific spreadsheet, but just not concurrent requests.
The specific example is a year out of date–but the basic complaint is not. I see no hard evidence that hallucinations have reduced to the level that their output is fundamentally reliable at this point, and plenty of evidence–both academic and anecdotal–that not only do hallucinations continue, but that they’re also a fundamental and inevitable aspect of the technology. Absent strong statistical analysis that refutes this conclusion, I’m unlikely to change it.
Lol, thank you, but no worries there… spreadsheets are not something I have any sort of emotional investment in
I was just trying to understand what you meant.
Oh I see… so you’re more talking about this from a “when all you have is Excel, everything looks like a spreadsheet” perspective? You’d prefer “purer”, single-purpose tools instead of the gradual feature creep that led into the flawed overlap between them (spreadsheets and databases)?
I’m a developer too (but in web, not desktop) and it’s something I also see at work, where less-experienced customers frequently use the wrong tool for the job and end up miring themselves in very contorted situations that are hard to work with and impossible to maintain. But to them, I think the mental barrier to learning the right tool is still higher than having to deal with the wrong tool inefficiently every single time… maybe the constant need for optimization is a symptom (or cause?) of “programmer brain” that regular people don’t have? 
To clarify to both you and @Dinsdale - my common understanding of “spreadsheet” is same as @Dinsdale but this thread is specifically asking about the Excel sort. @puzzlegal’s comment:
was one for me to seriously consider. I am in the, not small, group of knowledge workers who do not use them. I get presented them sure. But when I was in my administration role I made requests of what data I wanted and sometimes grabbed specific information to number crunch by hand for my very specific questions.
I was honestly completely ignorant that this tool is actively used by so many.
Yes I do suspect that most will be in my position in the very near future: knowing what meaningful questions they want answered, what data comparisons they want to make, what trends they want to look at, what distracting garbage they don’t want, and querying the agent to create it. The agent will just be much more rapidly responsive than any of my administrative support was!
Clearly we are not quite there yet.
Thanks for the reply. I asked my question, because the OP was an offshot from a thread discussing “old” tech - analog clocks.
This is one of those things where I have no clear recollection of what we called spreadsheet-type charts “back-in-the-day. I’m 65, and cannot recall a time that I didn’t see information presented in that manner. Even if the term “spreadsheet” started with Quicken, that was back in the 80s. I certainly have no recollection of saying “spreadsheet” before that.
In the 80s-90s, I played on a softball team and in a golf league. One guy on our team and another in our league kept all manner of computerized stats. I imagine they did them using Quicken or some other program. But the information was instantly understandable to me. In no way did their products impress me as something novel that I had not previously learned how to understand.
For whatever reason, I’m thinking that my earliest awareness of such carts was related to expanded timelines - setting forth historical/sociological eras, and tracking various developments in those eras. Or, as I noted, sports boxscores in the paper and on trading cards. Hell, a sports scoreboard is a type of simple spreadsheet. Sure, not exactly the same thing as a current spreadsheet, but certainly provided me all the tools needed to decipher a spreadsheet.
My wife uses spreadsheets for our household budgeting. A BIL and his son are softwar engineer and accountant respectfully. I recall them discussing the intricacies and wonders of Excel some years back. Impressed me as something that could be used by many people for many purposes - but for my purposes, I could stick to word processing and pen/paper.