Hey there! This week we're targeting a much requested breakdown: how to clean up a messy spreadsheet with AI.
Use whatever messy spreadsheet is already in front of you: an inventory list, budget, data-heavy sheet, copied table, CSV, schedule, or expense export. First we will help you pick the right AI tool and set it up, then show you how to have AI inspect the full sheet, propose a cleanup plan, make approved changes, and verify the result.
Before we get into it, here's this week at a glance.

Most people use AI for spreadsheets the hard way. They ask a chatbot what formula to write, copy it into Excel or Google Sheets, hit an error, then go back and ask what went wrong.
The better move is to bring AI into the spreadsheet itself so it can read the cells, formulas, tabs, and formatting directly. This week, the workflow is simple: inspect the full sheet, get the cleanup plan, approve it, then verify what changed.
We've used similar workflows to organize 2025 tax data, parse hundreds of lines of business expenses and income, and build a profit-and-loss summary to review with an accountant. The useful part was not that AI magically did everything. The useful part was that it helped organize the mess, then gave us a clear checklist of everything that it did.
Step 1: Set up the right tool
Step 2: Open a messy spreadsheet and ask AI to inspect the whole thing first.
Paste this into the AI sidebar: Review this spreadsheet. Take in every tab, column, formula, and table you can access. Tell me what it contains, what looks messy, and what cleanup plan you recommend. Do not change anything yet. Wait for my approval.
Step 3: Approve the plan or ask for changes.
If the plan matches what you want, approve it and let AI begin. If the plan misses something, ask it to revise the plan first. The plan is what makes verification easier because you know what it intended to change.
Step 4: Verify.
Once the changes have been made, review row counts, totals, formulas, sample rows, and anything AI marked uncertain. If you want more cleanup, summarizing, formatting, or formula help, repeat the same inspect-plan-approve-verify loop.
Other useful things to ask AI inside a spreadsheet.
Help me understand what this spreadsheet is tracking and where my expertise is needed.Design a better version of this spreadsheet from scratch. Ask me questions first, then suggest tabs, columns, formulas, and summary views.Color-code this sheet so I can quickly see missing data, risky rows, completed items, and categories.Turn this messy data into a clean tracker I can keep using every week.Create a summary view that shows the most important numbers, what changed, and anything unusual.
U.S. export controls forced Anthropic to pull or restrict access to Fable 5 and Mythos 5 (their latest AI models). Fable 5 is the same model we covered last week. These models are especially strong in cybersecurity.
Your access to AI models (like Claude) is not guaranteed. It can be taken away. Whether you are an individual or a company, ensure that you understand how to use other AI models and can switch. Do not be entirely dependent on one AI company.
The best AI users are tool-flexible. They know which model they prefer, but they are not stuck if that model is unavailable.
See you next week,

Ky Tomita, The Playbooks AI