There's a spreadsheet on your computer right now that you can't fully use.
A sales export. A survey result. A list of expenses. A messy CSV someone emailed you with 4,000 rows. You know the answer you need is somewhere in there. You just don't know the formula to pull it out, and you're not about to learn VLOOKUP at 9 PM on a Tuesday.
For most of computing history, that wall was real. The data was useless to you unless you knew the spreadsheet incantations to interrogate it. AI knocked the wall down. You can now drop a file into ChatGPT or Claude, ask your question in plain English, and get the answer, the chart, and a cleaned-up version back, without touching a single formula.
This is one of the most quietly useful things AI does, and most people still don't know it's possible. It's also a skill you can charge other people for, which I'll get to.
What AI actually does with your spreadsheet
Worth understanding, because it explains both the power and the limits.
When you upload a spreadsheet to ChatGPT's data analysis mode or to Claude, the AI doesn't eyeball your numbers and guess. It writes actual Python code behind the scenes, runs that code on your file in a sandbox, and reports back what the code found. You ask "which region grew fastest last quarter," it writes the code to calculate that, runs it, and tells you the answer in a sentence.
You don't have to read the code, though it's right there if you want to check it. You don't need to. You're talking to it like you'd talk to a competent analyst who happens to work instantly and never sighs at your questions.
That's the shift. The formula knowledge used to live in your head. Now it lives in the AI, and you just need to know what question you want answered. And knowing the question turns out to be the easy part.
The four prompts that do almost everything
Most spreadsheet work comes down to four moves.
The first: understand what you're even looking at.
"Here's a spreadsheet. Before I ask anything specific, tell me what's in it: what each column represents, how many rows, the date range if there is one, any obvious data quality issues, and three interesting things you notice at first glance."
This is the one people skip and shouldn't. Half the time, the file has problems (blank rows, inconsistent formatting, a column that's secretly text instead of numbers) that you'd want to know about before you trust any analysis. This prompt surfaces them.
The second: ask your actual question, in plain words.
"Using this data, answer: [your question]. Show me the number, explain how you got it, and tell me if anything in the data makes the answer less reliable than it looks."
The "explain how you got it" and "less reliable than it looks" parts matter. They turn a black-box number into something you can sanity-check.
The third: clean the mess.
"This export is messy. Standardize the date formats, remove duplicate rows, fix the inconsistent capitalization in the names column, flag any rows with missing values, and give me back a clean version I can download."
Data cleaning is the most tedious task in any office, and it's the one AI handles most happily. What takes a person an hour of find-and-replace takes AI about thirty seconds.
The fourth, and the one that quietly makes you better: learn the move.
"Show me the formula or the steps I'd use to do this myself in Excel or Google Sheets next time, and explain it simply enough that I'll remember it."
This is the difference between outsourcing your brain and upgrading it. Ask for the formula, read the explanation, and over a few months you'll actually pick up the spreadsheet skills you've avoided, painlessly, one real problem at a time.
Which tool for which job
Three options, each with a sweet spot.
ChatGPT's data analysis (built into Plus at $20 a month, with a limited free tier) is the most accessible. Upload, ask, get charts and answers. Best for quick one-off analysis. The catch: files are ephemeral, so each new session starts fresh with no memory of the last one.
Claude is the strongest for reasoning across messy or multi-sheet data, and for when you want it to think carefully through a multi-step question. Several 2026 tool comparisons rate it the best overall for serious spreadsheet reasoning. The catch: a smaller upload limit, 30MB per file, so very large files can be a problem.
Julius AI is the no-setup "upload and ask" option built specifically for this, with clean charts out of the box. Worth knowing if you do this often enough to want a dedicated tool.
For most people, whichever AI you already pay for is the right answer. You don't need a new subscription. You need to know the file-upload button exists.
The part where this makes you money
Here's the service angle, because it's real.
Every small business owner has data they can't read. The shop owner with two years of sales they've never analyzed. The nonprofit sitting on survey results nobody has touched. The freelancer who wants to know which clients are actually profitable once you subtract the headaches. They have the data and no idea how to question it. And you can be the person who does.
The service is simple: they send you their messy spreadsheet and the questions they want answered, you run it through AI, sanity-check the output, and deliver a short, clear report with the answers and a chart or two. A few hours of work, and most of it is the AI's.
You're not selling Python skills. You're selling the willingness to sit with someone's data and turn it into plain-English answers they can act on. Price it per project to start. The same "show, don't pitch" move that works everywhere else works here too: offer to analyze one dataset for free, deliver something genuinely useful, and the paid work follows.
This pairs naturally with any admin, bookkeeping, or marketing work you already do for small clients. It's a strong add-on, not just a standalone gig.
The honest limits
Three things to keep you out of trouble.
Verify the numbers, always. AI runs hidden code, and if it misreads a column (treats a text field as a number, misunderstands what a column means, silently drops blank rows), it can hand you a confident, wrong answer. For anything that matters, spot-check the result against something you can verify by hand. The "explain how you got it" prompt is your first line of defense.
Watch the file size. AI chatbots hit their context or upload limits well before Excel's actual row limits. Somewhere past tens of thousands of rows, they slow down or quietly truncate your data, which means you might be analyzing only part of the file without realizing it. For genuinely large datasets, you need a purpose-built tool, not a chatbot.
It's a conversation, not a saved system. In ChatGPT and Claude, the analysis lives in the chat thread, not in the spreadsheet itself. Close the session and the working is gone, even if you kept the answer. For one-off questions that's fine. For something you'll redo every month, save your prompts so you can rerun them cleanly.
If you want the broader system for handing AI the office busywork you've been doing by hand (data, documents, email, planning, all of it), I just published Your AI Operating System: The Beginner's Field Guide to Letting AI Do Your Busywork on Gumroad. Volume 1 of my AI for Real Life library. Volume 4 will go deep on AI for office work, where spreadsheet and data workflows like this one get the full treatment. Launch price for Volume 1 is $19, and existing buyers get every future volume free as I release them.
You've been treating "I'm not good at spreadsheets" as a permanent fact about yourself.
It was just a missing skill, and the skill is now a button. Ask the spreadsheet what you want to know.
Tags: Productivity, Artificial Intelligence, Data Analysis, AI Tools, Spreadsheets
Discussion
Comments
Share your thoughts on this article. Comments are reviewed before they appear.
No approved comments yet. Be the first to start the discussion.