FREE PRACTICAL AI LEARNING · LEARN BY DOINGNO SIGNUP NEEDED · YOUR WORK STAYS YOURS
Open PracticePRACTICAL AI ENGINEERING
EARN BY DOING · BUILD 04 · DATA & SPREADSHEETS

Clean a real spreadsheet safely

Cleaning a client’s spreadsheet is easy to get wrong in ways nobody notices until money is involved. Here you work on a copy, write down every rule before you apply it, and finish with a validation report that compares before and after. AI can suggest rules and formulas; the checks are done with counts and formulas, not trust.

⏱ About 20–30 minutes to draft · the real-world steps take days · Works with or without AI · No signup

You will make

  • Rules log
  • Cleaned file
  • Validation report

You will practise

  • Writing cleaning rules before touching the data
  • Protecting personal data while you work
  • Proving a clean-up with before-and-after counts

Download the practice pack (.zip)See a finished exampleOffer contract (.md)

No AI key needed. The guided template works for everyone.

  1. 1Brief

    Specific answers make better results — with or without AI.

  2. 2Make

    Pick a route. The template gives you a structure with [[blanks]] to fill in yourself. Your own key drafts it for you — then you still check it.

    See the exact prompt — copy it into any AI chat you already use
  3. 3Check
    The same rules run on every draft, human or AI. Most first drafts fail something — that is the point.
      Self-check — judgement no checker can make
    • 4Improve
      Fix one item at a time, then check again. Notice which fixes made it better — that is the skill.
      1. Run the check first.
    • 5Ship
      Use it for real: send it, post it or save it. Record what happened — that is your evidence of learning.
    A finished example
    The business, people and numbers are invented. It passes every check above, so you can see what “done” looks like before you start.
    Show the worked example
    > Worked example. The business, people and numbers are invented to show the shape of a finished deliverable.

# Clean-up plan: a hardware shop’s stock list

## Safety first
- I work on a copy named stock-clean-2026-11-04.xlsx. The original is never edited.
- Personal data: supplier phone numbers in column H. I hid column H before asking AI for formulas and never pasted any row.
- Size at the start: 1,240 rows, 9 columns

## Rules log
Problems found: duplicate products, prices typed as text, supplier names spelled three ways.

1. Merge exact duplicates — column B (product code) — why: the same code appears twice — method: sort by code, compare rows, keep the newer count
2. Convert prices to numbers — column E — why: 312 prices were text — method: =VALUE(SUBSTITUTE(E2,"$",""))
3. One spelling per supplier — column G — why: "Hardline", "Hard Line" and "HARDLINE" — method: lookup table in the tab "suppliers"

## Validation report
- Rows before: 1,240 · rows after: 1,198
- Duplicates removed: 42, listed in the tab “removed”
- Values changed: 391, listed in the tab “changes”
- Check total before: 1,240 rows of stock · after: 1,198 · difference explained: the 42 duplicates, each checked
- Spot checks (three rows compared by hand): rows 15, 402 and 1,103 match the paper stock book
- The goal works: total stock value by supplier → a pivot table on column G, total US$18,406

## Handoff
- The client receives: the cleaned copy, the rules log and this report
- The client must check: the 6 products where the two duplicate counts disagreed

## Boundary
- Nothing was deleted without being listed. The original file is untouched.
- No personal data was pasted into an AI tool.

    Pro moves
    Methods this build practises
    Each method in the directory has a starter kit: the offer prefilled, first-sale steps, proof checks and the offer contract.
    METHOD 16Data-cleaning and spreadsheet repairBuyer: small organizations with messy customer, stock or expense sheetsMETHOD 23CRM cleanup and follow-up rescueBuyer: firms with neglected lead sheetsMETHOD 27Document-to-database setupBuyer: organizations with repeated forms/invoices/catalogues
    Questions people ask
    Can AI clean the spreadsheet for me?
    It can suggest rules and formulas, and that is useful. But the clean-up must be done with formulas you can check, and personal data must never be pasted into an AI tool.
    What does a client actually pay for here?
    For trust: a file they can rely on and proof of what changed. The validation report is what makes the work worth paying for.
    Will this make me money?
    Nobody can promise that, and Open Practice never does. This workshop helps you make work that is worth paying for and keep the evidence. Whether anyone buys depends on your market, your outreach and your work.
    Do I need an AI key to use this?
    No. The guided template teaches the same skill with no AI at all. If you add your own free key (Groq, Google Gemini, OpenRouter, Mistral or Cerebras), the tool can also draft and revise. The key stays in your browser and requests go straight to that provider.
    Where is my work saved?
    In this browser only (local storage), until you copy or download it. Nothing you type is sent to Open Practice.
    Keep building
    NEXT BUILDBuild 05 · Turn one recording into a clip packYou’ll make: Clips, Captions, Rights log, Approval sheetFREE TOOLGoogle Sheets Script GeneratorAutomate a Google Sheet with a safe Apps ScriptACADEMY COURSEEarn by Doing: useful work, honest offersThe full build path: guided activities, a hosted lab and a combined capstone.