To analyse food delivery spending with an AI assistant, paste a few months of statement rows, ask it to pick out the delivery-app orders, then show how many there were, the total, and which weekdays they cluster on. Next ask for a gentle limit and one question about what the convenience replaced. Check the count and total yourself.
Which prompts separate delivery orders from other food?
- 'These are my debits for the last three months. List every payment that looks like an order on a food delivery or quick-commerce app. Show date, merchant and amount. Do not include restaurants I paid in person, and tell me which rows you were unsure about.'
- 'Give me the number of orders, the total, and the average per order, month by month. Show the arithmetic.'
- 'Show how many orders fell on each weekday. If the rows carry a time, group by morning, afternoon, evening and late night; if not, say so.'
- 'Tell me which category my other food spending falls in, so I can compare delivery with groceries and with eating out.'
What might the answer look like?
With invented numbers, say 14 orders in a month averaging ₹420. The total is 14 × 420 = ₹5,880. The weekday split might look like this:
| Weekday | Orders |
|---|---|
| Monday | 1 |
| Tuesday | 2 |
| Wednesday | 1 |
| Thursday | 2 |
| Friday | 4 |
| Saturday | 2 |
| Sunday | 2 |
The orders add up to 14. A Friday spike is a pattern worth a question, not a judgement: perhaps that is the evening you get home tired. The statement shows the total charged, so delivery and platform fees are inside each amount, not listed separately.
How do you set a gentle limit?
Ask: 'Suggest three realistic monthly limits based on my current count, each a small step down.' Say the first step is from 14 to 11 orders; 11 × 420 = ₹4,620, which is 5,880 − 4,620 = ₹1,260 less. Pick a number you can keep, then review after a month. A limit you ignore teaches nothing.
What did the convenience replace?
This is the question that makes the numbers useful. Ask the assistant to pose it for each cluster: Was the order instead of cooking, instead of a late dinner out, or instead of a trip to the shop? Some orders buy back real time and some are habit. Only you can sort them, and the answer changes what you decide.
What if the totals look larger than you expected?
Check what the assistant counted. Quick-commerce grocery orders, restaurant pickups paid through the same app and a delivery membership fee can all land in the delivery list. Ask: 'List the five largest rows in the delivery list so I can confirm each.' Remove anything that is not an order and ask for the numbers again. Then decide what to count: a monthly membership is a cost of ordering, but you may want it as its own line.
Compare three months, not one. A single month with a festival, guests or travel can mislead. If the monthly total sits near the same figure, that is your baseline; if it swings, look at what happened in the high months.
Weekdays and weekends tell different stories. Weekday orders often replace a packed lunch or a late dinner after work; weekend orders often replace a family outing. The question about what the convenience replaced applies to both, but your honest answers will differ.
Tracking it without repeating the prompts
In SpendixAI, a rule can send a delivery merchant to its own category, and the Report builder lets you filter and group so the monthly count and total stay visible. Run the prompts once for insight, and let the category do the monthly work.
Close the exercise with a plan for next month: your chosen limit, the weekday you will watch, and the date you will review.
Useful ideas start with a clearer picture of your own spending.
Explore your reports
