Turn spreadsheet chaos into data you can trust

The pack covers the four jobs every spreadsheet cleanup needs. A profiling prompt maps the mess before touching it: every date format, phone style, placeholder blank, and suspected duplicate, ending in a cleanup plan you approve. A normalization prompt standardizes formats without changing meaning — ambiguous dates get flagged instead of silently reinterpreted, and blanks are never filled with guesses. A dedupe prompt catches exact copies and near duplicates (name variants, reordered words, the same email spelled two ways) and proposes which row to keep rather than deleting anything. A validation prompt then hunts for broken emails, impossible dates, empty required fields, and rows that contradict themselves.

It is deliberately tool-agnostic. The same prompts run in ChatGPT, Claude, or Gemini, so it works with whichever assistant your team already pays for.

Back up the original file first, then export the sheet to CSV. Work in batches of a few hundred rows at most; smaller batches keep accuracy up and make checking feasible. Run the four prompts in order inside one chat, and edit the profiling prompt's proposed plan before approving it — you know the data, the model does not.

The recipe's verification step is the part most people skip and should not. Reconcile row counts in versus out. Then sample rows the model cannot anticipate: every 10th row plus a handful you pick yourself, compared field by field against your backup. You are hunting values that changed meaning rather than format, and values invented for blanks. If a sampled row fails, you add a rule to the prompt and re-run the batch instead of patching one cell.

This pack is the right tool for a one-off cleanup. It is the wrong tool when the same mess regrows every month, because recurring mess is an intake problem: forms without validation, systems that do not sync, the same record typed into two tools. Cleaning downstream of that is a treadmill. The durable fix is validation at the point of entry, continuous deduplication, and pipelines that keep your systems in sync without a human shuttling CSVs. That is what we build, and a free AI opportunity audit will tell you whether your data problem justifies it yet.