Earlier this week, I wrote about an uncomfortable discovery: Michael and I were spending more than we thought. Our budget was mostly intact, and we were still saving, just not as much as we had expected. A few evenings of combing through statements turned up nothing obvious, so we decided to let AI take a look.
Before we did, though, I wanted to figure out how to do it without handing a chatbot standing access to our bank accounts. (The short version: Share a one-time, sanitized snapshot of your transactions instead of connecting your accounts, and use the service's privacy settings rather than trusting a request to "forget" your data.)
But getting your data ready is only half the job. The other half is knowing what to ask. And this is where I think a lot of people miss an opportunity.
Don't Ask AI to "Analyze My Spending." Give It an Assignment.
If you upload a spreadsheet and type "Can you analyze my spending?" you'll probably get a tidy summary of your categories. Groceries: $1,800. Restaurants: $890. Shopping: $1,250. That's accurate, but it's also exactly what your banking app already tells you.
A better approach is to treat AI the way you'd treat a sharp new research assistant. Tell it what you're trying to accomplish, give it specific tasks, and make it show its work. Here's the prompt I plan to use:
I'm reviewing my household spending because our expenses appear to have increased over the past several months, but I haven't been able to identify a clear cause.
Analyze the attached transaction data and:
1. Calculate total spending by month and category.
2. Identify the five largest increases in spending, both in dollars and percentages.
3. Look for recurring charges, subscriptions, and possible duplicate payments.
4. Identify merchants where our spending has increased significantly.
5. Look for patterns involving purchase frequency, transaction size, or timing.
6. Separate unusual one-time expenses from ongoing spending changes.
7. Suggest three to five realistic opportunities to reduce spending without making major lifestyle changes.
Show the calculations supporting your findings. Do not invent missing information or assume that every expense is unnecessary. Flag uncertain categories and ask for clarification when needed.
Finally, summarize your findings in plain English, prioritizing the issues with the greatest potential financial impact.
Feel free to copy it, and adjust the opening line to fit your own situation.
Why This Prompt Works
Every piece of that prompt is there for a reason.
The opening sentence gives AI context. It knows we're not building a budget from scratch or preparing a tax return. We're hunting for a change, so it should focus on what's different, not just what's large.
Asking for increases in both dollars and percentages keeps the analysis honest. A streaming service that jumps from $10 to $18 a month is an 80% increase, but it costs you less than $100 a year. Groceries rising 15% on an $1,800 monthly bill costs you about $3,200 a year. You want to see both numbers, because the most dramatic percentage isn't always the most expensive problem.
Separating one-time expenses from ongoing changes is probably the most important instruction on the list. A $3,000 home repair in August will make that month look terrible, but it isn't a leak. It's a planned expense. A $150 increase in restaurant spending that shows up every month, on the other hand, is precisely what we're looking for.
The final paragraph sets the ground rules. Telling AI not to invent information, not to assume every expense is wasteful, and to flag what it isn't sure about makes it less likely to hand you a confident answer built on guesswork.
And notice what the prompt doesn't include: my name, our address, account numbers, or our household income. The analysis doesn't need any of it, so I'm not sharing it.
Then Dig Deeper With Follow-Up Questions
The first response is a starting point, not a final report. Once AI flags something, ask it to explain.
Say it reports that grocery spending rose sharply over six months. That alone doesn't tell you what to do about it. So follow up:
Break down the increase in grocery spending by number of trips, average amount per trip, and store. Which of those factors accounts for most of the change?
If the answer is that you're making eight more trips a month, the fix might be a weekly meal plan. If your average receipt went up at the same store, you might be looking at higher prices or a few new habits that crept into the cart. Those are very different problems with very different solutions.
The single most useful follow-up, though, is one of the simplest: "Show me the transactions behind that number." It forces AI to connect its conclusions to your actual data, and it's the fastest way to spot when something has gone wrong.
Prompts for the Most Common Spending Leaks
If you already have a suspicion about where your money is going, you can skip the broad analysis and go straight to a targeted question. A few I'd keep handy:
The subscription audit. "List every recurring charge with the merchant, amount, billing frequency, and the first date it appears. Flag any charge whose price has changed and any services that seem to overlap."
The annual renewals. "Identify charges that appear once a year, such as memberships, software, or warranties, and estimate when each is likely to renew." These are easy to forget because they show up so rarely.
The payday effect. "Compare our spending in the seven days after each paycheck with the rest of the month. Is there a pattern?"
Death by small purchases. "Total all transactions under $25 by month and by merchant. How much do they add up to, and is that number growing?"
Price creep. "For merchants we pay regularly, such as utilities, insurance, and phone service, show how the amount has changed over time."
Check AI's Math Before You Act on It
Here's the part that matters most: AI can make mistakes, and it tends to make them confidently.
It can miscategorize transactions, misread refunds as purchases, or count a pending charge and its posted version as two separate payments. One of the most common errors comes from uploading both your checking account and your credit card. The payment from checking to your card can be counted as spending, on top of the purchases it paid for, which can double a chunk of your expenses overnight.
So before you cancel anything, do a quick reality check. Ask AI to total each month's spending, then compare those totals with your statements. Pick a handful of the transactions it cites and make sure they exist and say what it claims. If the numbers don't match, ask it to explain the difference.
And keep your own judgment in the loop. If AI suggests cutting childcare because it's one of our largest monthly bills, I'm going to politely suggest that it reconsider. Some big expenses are big because they matter.
Let AI Find the Patterns. You Make the Calls.
Michael and I still don't know where our leak is. It could be a stack of forgotten subscriptions, a grocery bill that has quietly crept higher, or a mistake I made when I built the budget in the first place. Whatever it is, I'd rather find it with a clear assignment than by scrolling through hundreds of transactions one more time.
Because the goal isn't to let AI decide what you should and shouldn't spend money on.
It's to let AI find the patterns, so you can make better decisions.