What AI Revealed About My Money That I Couldn't See Myself

I've always been good with numbers. Or so I told myself.

I have dashboards. Spreadsheets. A structured view of assets, liabilities, and monthly cash flow. I track CPF, equity positions, credit lines, and school fee schedules. I know my numbers.

What I didn't know was what my numbers meant about me.

There's a difference between tracking data and understanding it. Between knowing what happened and understanding why. My financial tracking was excellent at the former. It was doing nothing useful for the latter.

The experiment

As part of my AI audit across all life domains, I decided to paste three months of transaction data into an AI and ask it a deceptively simple question: What patterns do you see that I might be missing?

Not "categorise this." Not "give me a pie chart." Not "tell me where I'm overspending." Those are tool-level prompts. They get tool-level answers.

I wanted the thinking-partner version. I wanted it to look at the data the way a sharp, neutral, unafraid advisor would, someone with no incentive to protect my ego.

What it found

The AI identified three patterns I hadn't consciously registered.

First: my discretionary spending spiked consistently in periods of high travel or professional stress. Not random spikes. Patterned ones. The months I was keynoting most heavily were the months I was spending most unconsciously. I was rewarding the performance of high output with financial decisions that quietly undermined the stability I was working toward.

Second: my investment discipline was excellent in isolation, but I had a habit of treating credit facilities as a secondary buffer rather than an emergency option. The line between leverage as a strategy and leverage as avoidance was blurrier than I'd admitted to myself.

Third: the data showed a clear emotional dimension to my spending that mapped almost directly onto school fee cycles. The months where education costs peaked were the months where I was most likely to make small, comfort-driven purchases, nothing dramatic individually, but cumulative in a way that mattered.

None of this was catastrophic. But none of it was visible to me when I was inside the data, because the data didn't tell me the story. It just showed me the events.

The distinction between data and story

This is the gap I talk about with organisations all the time. Companies have more data than they've ever had. They can tell you what happened with extraordinary precision. What most organisations, and most people, struggle to do is contextualise it. To see the narrative in the numbers.

AI is genuinely good at this when you ask it the right questions. The problem is that asking the right questions requires you to be willing to hear the answers.

That's not an AI challenge. That's a human one. And it's the most important one.

Practical AI: Let AI read your spending

Try this. Export or copy the last 3 months of your bank or credit card transactions (strip out account numbers for privacy, keep categories and amounts). Paste it into your AI of choice and use this prompt:

"Here are my last 3 months of transactions. I don't want you to categorise them, I want you to look for behavioural patterns. What does this data suggest about how I relate to money emotionally? What patterns are here that I might be too close to see?"

Then sit with what comes back. Don't defend it. Don't dismiss it. Just ask: Is this true? Today, I have a significantly more advanced workflow. I built a local agent that captures all financial files, such as bank statements, purchase receipts, CPF statements, trading activity, and more, as PDFs (since almost no banks in Singapore provide raw data). It strips the data, normalises it, then enriches all those useless descriptions before storing it in a markdown file, something that, in the AI world, is 95% more efficient. From there, my financial models can work their magic.