Automation
400 receipts a month. Month-end closing from four days to one.
Construction company with five job sites running at once and two people in the office. Four weeks to set up and adjust.
- Manual work a month
- Before: 32 hoursAfter: 10 hours
- Receipts pending at closing
- Before: 60After: 15
- Time to close the month
- Before: 4 workdaysAfter: 1 workday
The situation
Every site lead sent photos of receipts through text messages. About 400 a month, from five job sites at once.
The two people in the office had to figure out which job each receipt belonged to, type in the amounts and chase the site leads for anything missing. Month-end closing took four working days.
What I found
- Receipts arrived through personal messages, with no standard way to send them.
- Every receipt was read and typed by hand.
- Missing information was discovered only at closing, when it was hardest to get.
What I changed
- One central place to send receipts.
- Automatic reading of the receipt data.
- Each receipt filed under the right job.
- A person reviews only the incomplete or doubtful ones.
The numbers
| Metric | Before | After |
|---|---|---|
| Receipts a month | about 400 | about 400 |
| Manual work a month | 32 hours | 10 hours |
| Receipts pending at closing | 60 | 15 |
| Time to close the month | 4 workdays | 1 workday |
Four weeks to set up and adjust.
Why it worked
The goal was never to remove people from the process. AI reads and files, and the office reviews what matters. The office went from typing to checking, and problems surface when the receipt arrives, not three weeks later at closing.
Real clients. Names and identifying details are withheld, and numbers are rounded. Ad cost figures are media spend only, without the management fee. Results depend on the market and are not a promise of future results. Work done by Leo Agostini.
Sound like your business?
Tell me what is going on in a few lines. I will tell you honestly where the money is leaking, and if I can help.
