AI is excellent at messy-text jobs that are tedious by hand: categorising bank narrations, standardising names, pulling fields out of free text.
Classify
“Categorise each line into Salary, Rent, Utilities, Vendor payment, Customer receipt, Bank charges or Other. Return a two-column table.” Give the fixed list of categories, or you’ll get a different set each time.
Standardise
“Standardise these company names: remove M/s, fix abbreviations, Title Case.” Ask for a mapping table (original → clean) so you can review and reuse it with XLOOKUP.
Extract
From addresses, emails or descriptions: “Extract city, state and PIN code as a table; null if missing.”
Scale: ask for the tool, not the answer
For 50 rows, paste and clean. For 50,000 rows every month, ask: “Give me Power Query steps (or an Excel formula / Python script) that does this for any number of rows.” Then the cleaning is repeatable and auditable.
💡 Spot-check 10% of AI-classified rows, focusing on the ‘Other’ bucket and borderline cases, before using the results in a report.
⚠️ Remove account numbers, PAN, Aadhaar and personal details before pasting data into any AI tool that isn’t approved for confidential data.
Practice
Try the three exercises with anonymised data.
📎 Practice files for this article
📄
Lesson 5 worksheetPrompts and exercises for this lesson.
