Your Expense Tracker Is Sleeping Data. AI Wakes It Up. (Prompt Finance)

My wife and I have 13 years of budget data in YNAB. Every kopi, every Grab ride, every impulse Lazada purchase. All tagged, categorised, and synced.

Since 2013, the best I could do with that data was look at what was available as a feature in the app, or use basic searching and filtering. “You spent $X on groceries this month.” Thanks, I noticed.

I decided to ask AI to build me a local database. One afternoon later, I could ask my data anything. Not “how much did I spend this month” but questions like:

• “How has my relationship with cash changed over the last decade?”
• “Which ride-hailing app won my wallet?”
• “How much have I spent on food delivery since it existed?”

The answers were more interesting than I expected. Some were eye-opening. One was uncomfortable enough to change my behaviour.

The 3 Questions I Got to Ask

The point of having a local database on my laptop isn’t the database itself, but the questions I can now ask.

This represents a shift in how I access my own data. Instead of searching by keywords or scrolling through filtered lists, I can enquire using plain English. No code, no formulas, no query language. I type a question, AI answers it using my data.

Once my 13 years of YNAB data sat in a queryable format, I started asking things I’d never thought to ask. Three examples stood out.

The Cashless Revolution (Lived, Not Observed)

My prompt → What percentage of my transaction counts were cash, year by year?

In 2013, about 70% of my transactions involved cash. By 2022, that number dropped to single-digit percentages. The transition to a largely cashless society was accelerated by the Covid pandemic, and my data tells that story in granular detail.

Now and then, there are talks about Singapore going cashless. Surveys, government reports, news articles. This wasn’t a survey. Every row in that screenshot is a moment I reached for something other than cash to pay for a product or service. Thirteen years of a national shift, captured in my spending habits.

I didn’t set out to go cashless. It happened gradually, one PayNow and one contactless tap at a time. Seeing the trend laid out year by year made it feel real in a way no headline could.

Interesting pattern. But what about the apps competing for that cashless spend?

The Ride-Hailing Wars (Through My Wallet)

My prompt → What does my ride-hailing spend look like by app, year by year?

Grab took an early lead and dominated, until Gojek entered the picture in 2018. By 2024, Gojek had surpassed Grab in my personal spend. ComfortDelGro’s booking app? It almost vanished from my records after 2021. (I wrote about the ComfortDelGro vs Grab battle back in 2018.)

This is market disruption, viewed through my wallet. I didn’t need an industry report. I could see which app earned my money, when the shift happened, and how fast Gojek gained ground once it arrived.

For reference, I’m not loyal to any platform. I pick whichever app gives me a better price at the moment. The shift in my data reflects pricing competitiveness more than brand preference.

Two patterns down, both interesting. But here’s the one that made me put my phone down and think.

The Food Delivery Shock

My prompt → How much have I spent on food delivery each year? Indicate percentage of each merchant, and total annual spend.

I knew the number would be high. I wrote about spending $5,000 on GrabFood and Foodpanda in 2022. But seeing the full year-by-year breakdown across every food delivery platform, from the first order to last month, was different.

The first two examples are interesting, but this one is actionable. Knowing the exact cumulative number forced me to reconsider how often I hit “order” instead of walking downstairs. Some truths are easier to ignore when they stay as a vague feeling. A specific dollar figure is harder to argue with.

To be clear, the data doesn’t quite reflect my current reality. Increasingly, I’ve been using food delivery apps for in-store pickups, which helps decrease the total cost of food. But that’s probably another story for another post.

The three queries above took minutes. The data had been sitting in my expense tracker for years. I’d recorded every transaction but could only see what the app’s built-in reports and features allowed.

A local database that I have ownership of changed that. But a database doesn’t build itself. That part took an afternoon.

One Afternoon, One AI, One Database

Truth be told, I didn’t know how to set up a local database. I track expenses, not write code (usually).

So I asked Nei, my AI assistant. That’s what I call my CTO, short for Chief Turtle Officer, haha! I told her what I wanted: a local mirror of my YNAB budget data that I could query. Nei gave it a fancy name, designed the workflow, checked the official YNAB API documentation when she was unsure, walked me through each step of the process, tested its own creation, wrote documentation on how to use it, and sifted through my data to flag typos and misspellings for clean-up.

One afternoon. No prior database experience required.

The setup isn’t a one-time thing either. Since new data flows into YNAB daily, I asked Nei to create one-click sync tasks so I wouldn’t have to remember a single thing:

• Full Sync (complete mirror from day one, slower)
• Fast Sync (last 12 months, quick).

I run the fast sync regularly, the full sync occasionally, and my local database is instantly updated and brought up to speed.

Could I have figured this out on my own? Probably, with enough Googling and weekend afternoons. The point is I didn’t have to. I described what I wanted, AI handled the how. That’s a pattern I keep coming back to in this Prompt Finance series. (In my previous post, I used the same approach to calculate my KrisFlyer miles expiry.)

If you’re curious about the step-by-step setup, that could be a future post. For now, the takeaway: if your expense tracker exports data, you’re one afternoon and one AI conversation away from a queryable database.

But that “if” is doing a lot of work. Not every tool qualifies.

The One Criterion Nobody Tells You

If you’re using an expense tracker or budgeting tool, or shopping for one, I look at three things.

• Mobile-first. You’re going to track expenses on the go. If the app isn’t fast and painless on your phone, you’ll stop using it within a month. Friction kills tracking accuracy, and without data, none of this is possible.
• Desktop or web version. Your phone is for recording. A larger screen is for housekeeping. Correcting wrong categories, splitting transactions, reviewing trends. Try doing that on a 6-inch screen and you’ll understand why both matter.
• Exportable data. This is the one most people skip, and it’s the one that makes everything in this post possible.

Your app’s built-in reports show you what the developer thought you’d want to see. A local copy of your data lets you ask anything. The cashless trend, the ride-hailing shift, the food delivery reality check. None of those queries existed in YNAB’s interface, and keyword searching wasn’t powerful enough. They only became possible because YNAB lets me export my data.

Cloud-based tracking is convenient, but local data you can query is a superpower.

Personally, I use YNAB. It fits all three criteria. Legacy pricing puts me at about S$10 per month on the annual plan, which isn’t cheap, but ease-of-use plus years of queryable financial history is worth it to me. Your expense tracker or budgeting app doesn’t have to be YNAB, but it has to be able to export your data.

Kevin

Kevin started Turtle Investor after hitting rock bottom at negative $25,755 net worth. Today, his investment income and side hustles cover his expenses and travels. He shares what actually works (and what doesn’t) for Singaporeans building wealth. Learn more about Kevin here.