You will be able to get transaction data out of your bank and remove details that should not be shared.
Priya has tried to budget three times. Each time she started with a blank spreadsheet, typed in a week of spending from memory, got bored and stopped. Her bank already has every transaction she made last month, with the date and the amount, in the right order. She has just never taken it out.
This module builds a working budget from that record. An assistant does the tedious sorting, a spreadsheet does the totals, and you stay in charge of both. The first step happens before any AI is involved: getting the data out and stripping it down to what the job needs.
Most banks and card issuers let you download your transaction history from internet banking. Look for an option near your account history called something like download, export or statement, and check which formats it offers. CSV is the one you want. It opens straight into Excel or Google Sheets, one transaction per row.
Some accounts only offer PDF statements. That is more work but still usable. You can often select the transaction table in the PDF, copy it and paste it into a spreadsheet, then tidy the columns. Be careful with free online PDF-to-spreadsheet converters: you would be uploading a full statement to a service you know nothing about, which defeats the point of this lesson.
Start with one account and one month. If you have a savings account and two credit cards, pick the one that carries most of your everyday spending. You can add the others once the method works.
Open the file. A typical export has more than you need: your account number at the top, a running balance, reference numbers, transaction codes, sometimes a column for the branch or channel. The budget needs three things.
Date, so you can tell which month a transaction belongs to. Description, so the assistant can work out what it was. Amount, so the spreadsheet can add it up.
Delete everything else. Delete the header rows that show your name, address or account number. Delete the balance column, because it tells anyone who sees it how much you have. Delete reference codes, because some of them can identify a specific transaction or account.
If your export puts money out and money in into two separate columns, keep both for now, or combine them into one amount column with spending as positive numbers and money in as negative. Pick one convention and stick to it, because the totals in lesson 2.4 depend on it.
Look at what is left in the description column. You'll see things like a supermarket name cut short, a food delivery service with a long order code, a PayNow transfer with a phone number fragment or someone's initials, a card payment to a merchant whose trading name is nothing like the shop sign, and a GIRO deduction to an insurer.
You might be tempted to clean these up too. Don't, beyond removing anything that identifies a person or an account. Removing a phone number or a full name from a PayNow line is right. Rewriting "SUPERMKT TAMPINES CTR" to "groceries" is the job you're about to hand over.
This is the part an assistant is genuinely good at, as lesson 1.2, Four money jobs AI does well, set out. It has read enough text about shops and payment services to make a sensible guess at most of these. You'd spend an evening on two hundred rows. It will take a few seconds, and you'll spend your time checking instead of typing.
Save your work as a new file with a plain name, such as "march cleaned.csv". Keep the original download as it came, somewhere private, and never paste it anywhere.
The reason for two files is simple. When something looks wrong later, you can go back to the original and check it. And when you open an assistant, there is only one file you are allowed to use: the cleaned one. If you have to stop and think about which file you're about to paste from, the system has already failed. Make the cleaned file the only one that ever leaves your folder.
Before you save, scroll through every row once more. Look for anything you missed: a name in a transfer description, a full card number in a refund line, an address in a delivery charge. Lesson 7.1, What never goes into an assistant, lists the categories in full, but for a transaction export the rule is short. Nothing that identifies you, another person or an account.
Priya did this for March, on the credit card she uses for almost everything. It took her twelve minutes, most of it spent finding the download button. Your turn is next: one account, one month, three columns, saved as its own file.
Download one month of transactions from one account, reduce it to date, description and amount, and save it as a cleaned CSV.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).