I tried tracking my expenses with an Excel sheet for a few months and it worked as long as I was diligent. I even installed Excel on my iPhone. Some days I was able to document each purchase as soon as it was made; on other days multiple small purchases piled up over several days, so eventually I had to spend a lot of time going back and forth between the Excel sheet and my banking app, deciphering transactions.

So I went looking for an app that could make this whole job easier. I tried a few from the recommendations and all of them allowed detailed tracking of personal expenses, “buckets,” and other cool stuff that felt advanced. Most of these apps had appealing design, looked polished and inviting. However, none of them removed that manual friction point — either diligently tracking a purchase right away, or spending hours backtracking multiple expenses made over several days.

This problem got me thinking about how to reduce this friction. One way, I thought, would be to detect each occasion when an NFC was used for a payment with the phone. It turns out to be technically unfeasible, as for security reasons both Android and iOS protect wallets from supervision by other apps. Another way would be to connect to users’ bank data, so at the moment of payment, an app would fire. This type of integration with banks would be possible under PSD2, but requires a license as an account information service provider (AISP). A heavy investment for a side project like this one.

The most effective way, it appeared to me, would be to change the data input — to move away from manually typing each expense to using voice recognition. What if I could just talk my expenses into an app, and AI would handle the rest?