To use an AI prompt to total medical expenses by family member, paste a plain list of bills with a date, who it was for, what it was and the amount, then ask for a table of totals by person and month. Always check the result by adding the numbers yourself or with a spreadsheet, because AI assistants can misread a line or miscount.
A pile of pharmacy bills, lab receipts and consultation slips is hard to total by hand, and an AI assistant can sort it quickly. It is a helpful first pass, not an authority; you stay responsible for the figures.
What should you prepare before pasting?
- Make one line per bill: date, family member, type, amount. For example: 04-Mar, Amma, pharmacy, 1850.
- Use first names or initials, not full names. Leave out account numbers, card numbers, addresses and any identifiers.
- Keep amounts as plain numbers without symbols, and use one consistent date style.
- Add a condition label only if you want totals per condition, such as 'BP' or 'knee'.
Check what the tool you use says about how it handles the text you paste. Do not paste anything you are not comfortable sharing.
Which prompts work well?
- Total by person: 'Here is a list of medical bills. Give a table of the total amount for each family member. Show the sum of each person and the grand total. Do not guess missing values.'
- Per month: 'Group the bills by month and person, and show a table with months as rows and people as columns, with row and column totals.'
- Per condition: 'Using the condition column, give the total per condition and per person, and list any bills with no condition.'
- Checking: 'List any lines you could not read as a date, a person or an amount.'
Ask the assistant to show the lines that went into each total. That makes errors easy to spot.
A small example
Take four made-up lines: 04-Mar Amma pharmacy 1850; 09-Mar Papa lab 2200; 15-Mar Amma consultation 800; 22-Mar Papa pharmacy 1400. Amma's total is 1,850 plus 800, which is ₹2,650. Papa's is 2,200 plus 1,400, which is ₹3,600. The grand total is 2,650 plus 3,600, or ₹6,250.
| Person | Pharmacy (₹) | Lab (₹) | Consultation (₹) | Total (₹) |
|---|---|---|---|---|
| Amma | 1,850 | 0 | 800 | 2,650 |
| Papa | 1,400 | 2,200 | 0 | 3,600 |
| Family | 3,250 | 2,200 | 800 | 6,250 |
If the assistant's table disagrees with these numbers, the error is on its side, and a reply asking it to recheck line by line usually fixes it.
How do you check the answer?
- Paste the same lines into a spreadsheet and total with =SUMIFS($C$2:$C$100, $B$2:$B$100, "Amma"), where C is the amount and B the person.
- Compare the count of lines the assistant used with the number you pasted.
- Spot-check two or three large bills against the original receipts.
- Look for a missing line, a duplicate or a misplaced decimal; these are the usual faults.
Is there a more direct way?
If your records are in SpendixAI, Claude or ChatGPT can be connected through MCP. You choose the scope, can start read-only and can disconnect at any time, so the assistant can answer questions about your own records without a copy-and-paste step. Whether you connect it is your choice, and the product's AI and privacy information explains what is shared.
A short checklist
- One line per bill, with initials only.
- Ask for the lines behind each total.
- Re-add the numbers yourself.
- Keep the originals and confirm anything that affects a claim.
Useful ideas start with a clearer picture of your own spending.
Explore your reports


