ChatGPT Mastery Pro ~8 min read Updated · September 18, 2026

Your spreadsheets, interrogated.

Every business runs on spreadsheets nobody has time to actually read. Upload them and ask questions in English — ChatGPT runs real code against your real numbers. Here's how to do it without fooling yourself.

01 This is real computation, not vibes

When you upload a spreadsheet and ask a question, ChatGPT doesn't guess from the general shape of your data — it writes and runs actual analysis code against the file (that's Code Interpreter working under the hood). Totals are computed, not remembered. This matters: it means the arithmetic is trustworthy even though the interpretation still needs your judgment.

02 The conversation ladder

Orient: "What's in this file? Describe the columns, the date range, and anything weird." Always start here — it catches wrong-file and broken-export problems immediately.
Clean: "Find duplicates, blanks, and formatting inconsistencies. Fix what's safe to fix and list what you changed."
Summarize: "Break down revenue by month and by customer. What are the top five customers as a share of total?"
Probe: "What changed between Q1 and Q2? What's driving it? Show me the three customers behind most of the difference."
Deliver: "Make a chart of the monthly trend I can paste into a deck, and give me the cleaned file back as an Excel download."
Verify like a pro

Pick one number you already know — last month's total, the row count, a customer's balance — and check ChatGPT's version against it before trusting anything else. One known-good anchor validates the whole pipeline. If the anchor's wrong, the export or the column mapping is wrong, and you just saved yourself from presenting garbage.

03 Questions worth stealing

Update · September 18, 2026 — the other route: ChatGPT inside Excel

Everything in this lesson is the upload route: bring the file to ChatGPT. There's now a second route in the other direction. ChatGPT for Excel and Google Sheets (available globally since May 5, 2026) puts a ChatGPT sidebar inside the spreadsheet itself, where it can build trackers and budgets, write and explain formulas, work across multi-tab files, run scenarios, and clean up messy sheets in place — and because it edits the workbook rather than a copy, the “formulas don't survive the trip” problem in section 04 goes away. The same Microsoft add-in now covers PowerPoint (generally available July 6, 2026) and, as of September 17, 2026, Word as well.

Limits, per OpenAI: Free and Go get limited usage; Plus and Pro share the same agentic usage limits as Codex, with extra credits purchasable. Install from Microsoft Marketplace (or Google Workspace Marketplace for Sheets), sign in with your ChatGPT account, and keep the verification habit — OpenAI's own guidance is to review outputs before relying on a formula or an analysis. When to pick which: one-off analysis of an export → upload route (this lesson); a living workbook you'll keep editing → sidebar route. The Word side of the add-in is covered in Writing with ChatGPT.

04 Limits and hygiene

Very large files can hit limits — filter to the relevant slice first. Formulas don't survive the trip: ChatGPT works on values, so treat outputs as new files, not edits to your master. And the standing rule: customer data, salaries, anything sensitive — know your plan's data policy before uploading, and when in doubt, anonymize the columns that matter. (Full data-hygiene guidance in Privacy & what not to paste.)

Try it now

Export something real — sales, expenses, inventory, your bank's CSV. Run the whole ladder: orient, clean, summarize, probe. Twenty minutes, and you'll know things about your own numbers you didn't this morning.

Open ChatGPT →

This week's challenge

Take the spreadsheet you're supposed to look at every month but don't. Upload it, verify one anchor number, then ask: 'Give me the five things in here I most need to know, ranked by how much money each one represents.' Put the answer where you'll act on it.

Up next in ChatGPT Mastery

Meetings, notes & summaries

Before, during, and after — the meeting workflows that give you back your Tuesdays. Read the lesson →