Comment by Priotecs
2 days ago
That’s fair — and I agree if enough context exists.
The key limitation is that a raw bank transaction usually contains very little semantic information: amount, merchant name, date. From that alone, an LLM can only guess based on patterns or prior behavior, not actually know what the expense was for.
“$100 at a supermarket” could be groceries, pet food, a household item, or something work-related. An LLM can infer probabilities once it has enough historical data and feedback, but that still means the user has to correct or confirm things at some point.
So I see LLMs as very helpful for assisting categorization (suggestions, defaults, learning over time), but they can’t fully replace intent unless the underlying data becomes richer than what bank statements provide today.
Is there any chance it could become richer? What governs the content of credit card and bank statements? Is there anyone pushing for them to be more useful?
I think (granted, this is from a quick bit of research so I could be wildly wrong) - the message you see in your credit card app with a transaction is usually mainly the merchant name and location which is part of ISO 8583, so it may be a bit hard to extend it to include an arbitrary message in a way that works without merchants having to replace card reader/POS systems en-masse.
Why not build a receipt scanner then?