Where does the money go

Note: general recommendations only, not professional advice. You are responsible for checking before use. For money, VAT and tax matters, consult a certified accountant or tax adviser. The full text is at the end.

What you'll have at the end of the hour

• A clean transactions file for three months, with no account numbers and no identifying details.
• A categorized table in Google Sheets: date, merchant, amount, category, fixed or variable.
• A list of recurring charges with monthly and yearly cost, and one column you fill in: "Do I still use this?".
• A pivot table and a chart of spending by category and by month.
• A formula-only version, for anyone who doesn't want to upload data to any AI tool.
• Three actions for next month: cancel, downgrade, or negotiate.

What you need

• Access to your bank's and credit card company's website (or app), with an option to export transactions to Excel or CSV.
• A Google account (for Sheets).
• A free AI tool: ChatGPT, Claude or Gemini. Optional: step 6 has a version without AI.

The kit, step by step (60 minutes)

Step 1 (10 minutes): Export three months

What to do: Download the transactions for the last three months, from the bank and from each credit card separately.

How: On the bank's and the card company's site, on the transactions or monthly statement page, look for a button like "Export", "Download to Excel" or a file icon. Pick a three-month range. If you can choose between CSV and Excel, Excel usually keeps non-English text intact better.

In Google Sheets, create a new spreadsheet and, before importing, set File > Settings > Locale to your country. That way a date like 05/03 is read the way your bank means it. Then File > Import, one tab per file.

Common mistake: counting the card twice. The bank statement shows one monthly charge per card, and the card company's file shows all the purchases behind that charge. If you upload both, delete the card's monthly charge lines from the bank file.

Step 2 (10 minutes): Remove identifying details before uploading to an AI tool

What to do: Keep only what's needed: date, description, amount. Everything else goes.

How:

1. Delete columns for account number, branch number, card number, account holder name, ID number and balance.
2. Transaction descriptions sometimes include long numbers (reference, customer number). To hide them, in an empty column next to the description column, type the formula below, drag it down to the end, and copy the new column over the original (Paste special > Values only).

=REGEXREPLACE(B2, "\d{5,}", "XXXX")

3. If the file has two amount columns (debit and credit), keep both. The prompt handles it.
4. Put everything in one tab called "Transactions" with all rows from all files, and one extra column "Source" (bank / card 1 / card 2).

Before uploading: check your AI tool's settings for whether your chats are used to train the model, and turn it off if you can. If you're not comfortable uploading even a cleaned file, go straight to step 6.

Step 3 (15 minutes): Sort into categories

What to do: Copy the "Transactions" tab (or upload it as a file) into the AI tool and run the categorizing prompt. You get a table ready to paste back into Sheets.

Here are transactions from the bank and credit cards of [a household / a small business in ...], for three months. Columns: date, description, amount (or debit and credit), source.
[paste the table here, or attach the file]

Task: categorize every transaction.
Household categories: Housing, Bills (electricity, water, local tax, phone and internet), Groceries, Eating out, Transport and car, Health, Insurance, Education and kids, Subscriptions and digital, Shopping, Leisure, Transfers and withdrawals, Income, Other.
Business categories: Materials and stock, Payroll, Rent and office, Software and subscriptions, Advertising and marketing, Travel and car, Suppliers and contractors, Bank and card fees, Taxes and government payments, Income, Other.
Use only the matching list. If a transaction is unclear, "Other", and in the note column say why.

Rules:
1. Do not add, delete or merge rows. The number of output rows equals the number of input rows.
2. An expense is a positive number, income or a refund is a negative number. If there are debit and credit columns, merge them into one column by this rule.
3. "Merchant": a short, consistent name for the same business (the same supplier with different descriptions gets the same name).
4. "Type": fixed (repeats every month at a similar amount) or variable.
5. Don't guess what was bought. Categorize only by the description.

Return one code block, tab-separated, with the headers:
Date	Merchant	Amount	Category	Type	Source	Note
Below the block: rows in input, rows in output, total expenses in input and total expenses in output. If they don't match, say so in the first line.

Putting it back: create a tab "Categorized", click A1 and paste. If everything lands in one column, Data > Split text to columns.

Common mistake: skipping the check line at the end. If the row count or the total doesn't match, something got lost. Ask "return only the missing rows" and don't move on until the numbers match. For a big file (hundreds of rows), send one month at a time.

Step 4 (10 minutes): Find subscriptions and recurring charges

What to do: In the same chat, run a prompt that looks for charges that repeat. This is usually where the surprises are.

Based on the categorized table, find recurring charges: the same merchant appearing in two or three months, at a similar amount.
For each, return a row: Merchant | Category | Average monthly amount | Yearly cost (x12) | Months seen | Last charge date | Price change (if the amount went up between months, by how much) | Do I still use this? (leave empty).
Flag in particular: small charges that are easy to miss, charges in a foreign currency, and two subscriptions that look like duplicates (for example two music services or two storage services).
Sort from most to least expensive. Don't suggest cancelling anything, just show it. At the end: the total of all recurring charges per month and per year.

Paste the result into a "Recurring" tab, and go through the "Do I still use this?" column yourself. Only you know the answer.

Step 5 (10 minutes): A pivot table and a monthly chart

What to do: Build a category by month table in Sheets, and a chart that shows it.

How:

1. In the "Categorized" tab, add a "Month" column. In H1 type "Month", in H2 paste the formula below. It fills the whole column by itself.

=ARRAYFORMULA(IF(A2:A="", "", TEXT(A2:A, "yyyy-mm")))

2. Select the whole table and choose Insert > Pivot table > New sheet.
3. In the pivot editor: Rows: Category. Columns: Month. Values: Amount, summarized by SUM. Filters: Amount, by condition "Greater than 0", so income doesn't offset expenses.
4. Click inside the pivot table and choose Insert > Chart > Stacked column chart. Each column is a month, each color is a category.

Want the AI to draw it? Some tools can create a chart from an uploaded file. You can ask:

From the categorized table, create a stacked column chart: X axis months, a color per category, expenses only (positive amounts). Below it a small table: for each category, the total per month and the monthly average.
If you can't create a chart, write me exact steps to build it in Google Sheets instead.

Step 6 (alternative, no AI at all): The formula version

What to do: The same categorizing, with a rules table and formulas. The data never leaves Sheets.

How:

1. A new tab called "Rules", with two columns: A keyword, B category. One row per rule, for example: "supermarket" → Groceries, "electric" → Bills, "fuel" → Transport and car. A keyword is part of the description as it appears in the file.
2. In the "Transactions" tab, arrange: A date, B description, C expense amount. If the file has debit and credit columns, the debit column is C. If there's a "Source" column, move it to G. In E2:

=IFERROR(INDEX(FILTER(Rules!$B$2:$B, Rules!$A$2:$A<>"", ISNUMBER(SEARCH(Rules!$A$2:$A, B2))), 1), "Uncategorized")

Drag it down to the end of the table. Every "Uncategorized" is a missing rule: add a row in the "Rules" tab and the category updates by itself.

3. Month column in D2, as in step 5 (with A2:A).
4. In how many months each merchant appears, in F2 (drag down). 3 or more is probably a recurring charge:

=COUNTUNIQUE(FILTER($D$2:$D, $B$2:$B=B2))

5. A category by month summary, in a new tab, in A1:

=QUERY(Transactions!A1:E, "select E, sum(C) where C > 0 and E <> '' group by E pivot D", 1)

If expenses appear as negative numbers in your file, write C < 0 instead of C > 0. From here the chart is the same as in step 5. Next month, just paste new transactions and everything updates.

Step 7 (5 minutes): Three actions for next month

What to do: Pick three actions from the lists: cancel a subscription, move to a cheaper plan, or ask a supplier for a better price.

I want to [cancel / downgrade / ask for a better price] with [supplier name]. I pay [amount] a month today, and I've been a customer for [how long].
Write a short, polite message to customer service, up to 5 sentences, that asks for exactly this and asks for written confirmation.
Don't invent competitor offers and don't threaten to leave unless I wrote that I'm ready to leave.

Note in the "Recurring" tab what you decided and when. Next month, run the kit again and check that the charge really went down.

Full example: Noa Design Studio (a fictional business)

Everything in this example is fictional: the business, the merchants, the names and the amounts. "PixelSuite", "CloudBox" and the other names are made up for illustration.

Noa exported three months from her business credit card and her business bank account. She deleted the card's monthly charge lines from the bank file, so nothing is counted twice, and removed the account and card number columns.

Excerpt from the categorized table (output of prompt 1):

The check line: "Rows in input: 87, rows in output: 87. Total expenses in input and output: equal."

The recurring charges list (output of prompt 2, shortened):

The AI's flags: DriveSpace and CloudBox look like duplicates (two storage services). FontHub is charged in a foreign currency. StockPix is a small charge that's easy to miss.

Noa's three actions for next month: cancel DriveSpace, find out what StockPix is and cancel it, and ask PixelSuite whether a yearly plan costs less. Together that's ILS 64 a month less from the first two, before PixelSuite even answers.

Checklist

• Three months, from the bank and from each card
• The card's monthly charge lines deleted from the bank file
• Sheet locale set before importing
• Account, card and ID numbers, names and balances removed
• Long numbers in descriptions hidden
• The check line: row count and total match
• You went through "Do I still use this?" yourself
• Pivot table with an expenses-only filter, and a chart
• Three actions chosen and written down with a date

Disclaimer

This content is general recommendations only, not professional, legal, accounting or tax advice. Every business and organization is different. You are responsible for checking that it fits your situation, and for checking the data, prompts and code before you use them. Outside tools, prices and terms of service change, so check them at the source. For anything involving money, VAT, tax or bookkeeping, consult a certified accountant or tax adviser before you rely on the result.

What's next

Here you export, clean and paste by yourself once a month. A process can be connected that takes the files, cleans and categorizes them by your rules, flags a new charge or a price that went up, and sends you a short monthly summary.