Tool: Claude
Time: 5 minutes
Result: An email-ready summary with trustworthy numbers, from a spreadsheet you didn't build.
The export has 38 columns, cryptic headers, and someone wants "a quick summary" by end of day. You didn't build this spreadsheet, but you're the one who has to explain it.
AI Win #11
Upload the messy spreadsheet to Claude and get a plain-English readout of what the data actually contains, the headline numbers, the patterns worth attention, and every data-quality problem flagged before it can skew a figure. It ends with a summary written for whoever asked, ready to paste into an email. Works on CRM exports, finance dumps, survey results, and inventory sheets.
How To Do It
Get the file as CSV or Excel. CSV uploads work on every Claude plan; .xlsx uploads need the file creation feature enabled in Settings. If in doubt, open the sheet and save it as CSV first.
Go to claude.ai, start a new chat, and attach the file with the + button in the message box.
Paste the prompt below, fill in who the report is for and the one-line context (or write "unknown" and let Claude work it out) and send.
Answer any questions Claude asks about cryptic columns, that's just it refusing to guess. Then refine: "shorten the summary to 3 sentences" or "give me the cleaned data back as a downloadable file."
Why This Is Useful
The slow part of spreadsheet work isn't the math, it's decoding a structure someone else built and cleaning it by hand. This prompt makes Claude explain the data and surface quality problems before it computes anything, so every number rests on a stated assumption instead of a silent guess. And because the summary is aimed at a named audience, it comes out readable, not like a data dump.
Copy / Paste Prompt
You are a data analyst who is excellent at making messy data understandable. I'm uploading a spreadsheet export that I need to turn into a clean summary report for [WHO IT'S FOR, e.g., my manager].
Context on the data: [ONE SENTENCE, e.g., "monthly sales export from our CRM" — or write "unknown"]
Work through this in order:
WHAT THIS DATA IS — in plain English: what one row represents, what each column contains, and what period it covers. If you can't interpret a column, don't guess — list it in your questions at the end.
QUALITY CHECK — duplicates, blanks, inconsistent formats (dates, currencies, spellings), and obvious errors. Give a count for each problem and say how you'll handle it in the numbers below.
HEADLINE NUMBERS — the 5–8 figures someone skimming this needs: totals, averages, biggest and smallest, and change over time if there are dates.
WHAT STANDS OUT — 3 patterns or outliers worth attention. For each: the specific rows or values behind it, and one plain sentence on why it might matter.
SUMMARY REPORT — a version I can paste straight into an email to [WHO IT'S FOR]: a 2-sentence overview, the headline numbers, and the standouts. Plain business language — no jargon, no column names.
Rules:
-Every number must come from the actual data. Where a calculation required an assumption (e.g., how blanks were treated), state it next to the number.
-Don't smooth over problems — if the data is too messy for a figure to be trustworthy, say so instead of reporting it.
-End with up to 3 questions whose answers would most improve this report.Got an export sitting in your downloads folder that you've been avoiding? That's your test file.
