AI and MCP

Bulk import transactions with an AI assistant: a beginner tutorial

A step-by-step guide for non-technical users: use an AI assistant to tidy an old expense sheet, then bulk import it into your tracker and check the total.

Editorial illustration for Bulk import transactions with an AI assistant: a beginner tutorial

You have an old Excel sheet of expenses and want everything in one tracker without retyping. A good beginner route is to let an AI assistant tidy the sheet, then use the tracker's own file import to load it. You stay in control, and you can check every step with a simple total.

Step 1: look at your sheet as a stranger would

Open the sheet. Can someone who has never seen it tell what each column is? You want one row per expense and four things on every row: the date, a short description, the amount and which account it came from. If amounts have the rupee sign typed into the cell, or dates look like 3/5 so nobody can tell March from May, those are what the assistant will help clean.

Step 2: ask the assistant to tidy it

  1. Copy the rows into the assistant, or attach the file if your assistant accepts it.
  2. Say clearly: make one row per expense with columns date, description, amount, account. Dates should be in day-month-year. Amounts should be plain numbers.
  3. Tell it the year and that dates in the old sheet are written day first.
  4. Ask it to list any row it was unsure about instead of guessing.
  5. Copy the tidy table back into a fresh sheet and save it as Excel or CSV.

Step 3: do the totals check before you import

Say your old sheet had six rows: ₹1,250, ₹3,400, ₹560, ₹780, ₹2,200 and ₹940. The total is 1,250 + 3,400 = 4,650, plus 560 = 5,210, plus 780 = 5,990, plus 2,200 = 8,190, plus 940 = 9,130.

CheckExpectedWhat a mismatch means
Number of rows6A row was dropped or split
Total amount₹9,130An amount was misread or a minus sign lost
Earliest and latest dateMatch your sheetDay and month got swapped

Step 4: import into the tracker

  • Pick the account first. SpendixAI keeps each transaction against the account you chose before importing.
  • Upload the file. You can upload several in one batch, and the import history shows a status for each.
  • Review the rows shown before you save anything.
  • If an import fails, you will get an Import failed notification, and the file is the first thing to check.

Do I need the assistant to talk to the tracker directly?

No. Many beginners will be happier without that. Claude or ChatGPT can be connected through MCP, but it is optional, you choose the scope and can start read-only, and you can disconnect any time. The file route keeps things simple and gives you a review screen.

Where beginners get stuck

  • The assistant 'helpfully' fixes something that was not broken, such as rounding ₹499.50 to ₹500. Check a few amounts by eye.
  • Rows with a blank amount quietly vanish. Count rows both before and after.
  • Two sheets for the same month overlap, and you import duplicates. Possible duplicates are flagged, so read the flags.
  • Bank statements are better imported directly than retyped. If you have the statement itself, upload that instead.

Fixing a wrong import without panic

If the total is off after importing, do not edit rows one by one at first. Work out the gap. Say you expected ₹9,130 and see ₹8,190: the gap is 9,130 − 8,190 = ₹940, which is exactly the last row of the example sheet. That tells you the last row was dropped, probably because of a blank line in the file. Fix the file, delete the imported rows from that batch and import again.

Doing it this way takes ten minutes. Editing rows one at a time without knowing the gap can take an evening, and you may still miss the real fault.

When you are done

Open the transaction list, sort by date and read down the first and last ten rows. If both ends look right and the total matches, close the sheet. Next month, skip the sheet and upload the statement straight away.

SPENDIXAI JOURNAL

Useful ideas start with a clearer picture of your own spending.

Explore your reports