AI Data Analysis: From Messy Sheets to Reports
Cleaning, formulas, pivots, charts and conclusions — one complete spreadsheet pipeline.
Path goal
Finish able to turn a messy raw sheet into a report page that states a conclusion.
Who it's for：Operations, business analysts, admin roles
-
Step 1: Clean before you analyse
Deduplicate, handle missing values, normalise formats, flag outliers and clean text. Define the rules first. AI writes the formulas or pandas code; only you can decide what counts as wrong.
💡 Keep the original columns and add flag columns so you can trace any mistake back.
-
Step 2: Have AI write formulas, not answers
Multi-condition sums, lookups, text splitting and distinct counts: describe the need and let AI produce a formula you run locally. Pasting a whole sheet into a chat is slow and error-prone.
💡 State the business definition first, then the formula you want — that minimises rework.
-
Step 3: Turn results into one conclusion
A conclusion sentence is a judgement, a number and a recommended action. Charts show trend; prose delivers the judgement — do not repeat one in the other.
💡 A report without definitions or recommendations leaves the reader asking "so what".
-
Step 4: Write a page that survives review
Headline as conclusion, body in three lines, numbers verified by you. Weekly, monthly and retrospective reports need different structures — do not force one template.
💡 Verify every number yourself — one wrong figure undermines the rest of the report.