AI Guides › Money & Business
Using AI To Track Where Your Money Actually Goes, Not Where You Think It Goes
By Nigel Guy · 2 min read
Most people carry a rough mental model of their spending — "marketing is our
biggest cost," "software is cheap" — built from memory and impression rather
than actual transaction data. That mental model is reliably wrong in
specific, checkable ways, particularly around small recurring charges nobody
re-examines once they're set up.
The rule: your mental model of where money goes is built from memory, and
it's reliably wrong in specific, checkable ways — use AI to categorise your
actual transaction history, not to guess at your spending pattern from a
description of your business.
The mechanism: the categorisation pass
- Export actual transaction data for a real period — three to six
months. Real data, not a description of your business written from
memory.
- Have AI group the transactions into consistent categories, and flag
anything recurring you may have forgotten you're still paying for.
- Write your own guess for each category's total before looking at the
actual figures. Comparing guess against actual is the honest test of
where your mental model is wrong, and it's the whole point of the
exercise.
- Investigate the biggest gaps between guess and actual specifically —
that's where the useful finding is, not in the categories where you were
roughly right.
- Repeat this quarterly, not once. Spending patterns and subscription
creep continue after the first clean-up; a single pass only tells you
about a single period.
What to skip
Skip relying on the categorisation as fully accurate without spot-checking a
sample yourself — miscategorisation happens, and it compounds if left
unchecked across periods. Skip stopping at the summary totals without
opening the list of individual recurring subscriptions specifically; that
list is usually where the actual surprise sits. And skip treating a single
clean-up as finished business — cancelling what you find this quarter
doesn't stop something new creeping in next quarter.
Guardrails
- Accuracy depends entirely on the completeness of the data you export;
missing an account or a card produces a confident-looking picture that's
actually incomplete.
- This is a diagnostic exercise for your own understanding, not a substitute
for an accountant or bookkeeper for tax or compliance purposes.
- Treat the categorisation as a first pass to review, not a final source of
truth for financial reporting.
All 751 AI guides · JulieMango plans from £17/mo