Ask your spending questions without building a spreadsheet

“How much did I spend on dining last quarter?” sounds simple until you open a spreadsheet, filter three months of rows, fix miscategorized takeout, and rebuild a pivot table. Most people abandon the question halfway through. The data already exists in your bank statements — you just need a way to query it in plain language instead of formula syntax.

Vereloop Finance includes an AI assistant that answers questions against your uploaded transaction history. No export to CSV, no SUMIF ranges, no chart wizard. This article explains how it works, what to ask, and how to get trustworthy answers.

From PDF to answerable data

The assistant does not read PDF files directly. It works on parsed, categorized transactions — the same structured rows you see on Analyze after upload. That is why automatic categorization matters: when Groceries and Dining are tagged correctly, answers about food spend split the right way.

Upload two or three months of statements first. Skimpy history produces vague answers; richer history supports comparisons like “April versus May” or “before and after I cancelled Netflix.”

Where to ask questions

Open the Analyze page after signing in. The assistant panel sits alongside filters and charts — ask in the chat box, get a text summary backed by your numbers. For deeper API and integration details, see the developer documentation.

Example questions that work well

• “What were my top five merchants last month?”
• “How much went to Subscriptions in the last 90 days?”
• “Compare my Transit spending January versus February.”
• “Did I spend more on Groceries or Dining in March?”
• “List transactions over $200 in April.”
• “What recurring charges hit my chequing account?”

Specific time ranges and category names yield sharper answers than “how am I doing?” Start concrete, then follow up.

How the assistant uses your data

When you send a question, the system resolves it against your stored transactions — dates, amounts, categories, merchant text — not against generic financial advice. It aggregates, filters, and summarizes what you actually spent. It does not invent charges or pull data from accounts you have not uploaded.

Think of it as a conversational layer on top of Analyze filters. You could click the same filters manually; the assistant saves clicks when the question is exploratory or you are not sure which filter combination you need.

Getting better answers

1. Fix categories first — one review pass after each upload prevents garbage summaries.
2. Name the window — “last month,” “Q1 2026,” or explicit dates beat “recently.”
3. One question at a time — compound asks (“compare dining and subs and also show rent”) split cleanly when separated.
4. Verify surprises — tap through to the underlying transactions if a total looks wrong; miscategorized rows are the usual cause.

What it is not

The assistant is not a licensed financial advisor. It will not pick stocks, approve mortgages, or optimize taxes. It excels at descriptive questions about your uploaded history — what happened, how much, where, when — not prescriptive life planning.

It also does not connect to your bank live. Answers reflect PDFs you uploaded, so refresh data by adding new statements when a month closes.

Spreadsheets versus conversational analysis

Spreadsheets shine when you model scenarios — salary changes, debt snowballs, investment projections. They are poor for ad hoc history questions unless you maintain them religiously. Most people do not. Bank PDFs plus parse plus AI questions give you spreadsheet-grade accuracy on real spending without maintaining a sheet. Use spreadsheets for forecasting; use Vereloop for honest lookbacks and quick audits.

Privacy

Questions and answers stay in your session context tied to your account. Vereloop does not use your chat to target ads or sell insights to third parties. Your transaction text never leaves the product for marketing purposes.