Read next
Review Financial Models With ChatGPT
How to use ChatGPT to review financial models more strategically while keeping human judgment in control.
Open article →Digital subscriptions are easy to start and easy to forget. A monthly bank statement can contain software, SaaS tools, advertising platforms, and services that are no longer useful.
The test was simple: upload the same business bank statement to three AI tools and ask each to identify spending on digital goods, software subscriptions, and SaaS tools.
ChatGPT-4o identified several digital expenses and returned the results in a clean table. It also created exportable data.
Strengths:
Limitations:
Claude Sonnet 3.5 performed best in this test. It identified more relevant expenses, calculated the correct total, and presented the output in a publishable format.
Strengths:
Perplexity performed poorly for this specific task. It misclassified some services, missed important entries, and produced an incorrect total.
Strengths:
Limitations:
AI can make a subscription audit faster, but the output is not a financial record. Treat it as a first draft.
For financial work, always compare the AI output against the original statement before acting on totals, classifications, or recommendations.
Suggested reading
Read all notesRead next
How to use ChatGPT to review financial models more strategically while keeping human judgment in control.
Open article →Read next
A workflow for turning long PDFs into structured, verifiable insights without reading every page first.
Open article →Read next
A practical note on when ChatGPT is useful for search and why important facts still need verification.
Open article →