From paper receipts in trucks to per-job costs in seconds

The problem

Expenses lived in paper receipts collected in trucks and pockets. Job costs were reconstructed in spreadsheets long after the fact, which made tax time painful, and the owner had no per-job cost visibility while jobs were actually running. Field and office ran on phone calls and photos in chat, so nothing arrived in a form anyone could add up.

Why the usual answers fail

Spreadsheets and paper only work backwards: someone has to gather receipts, decipher them, and retype them, so the numbers are always months behind the jobs. Heavyweight construction ERPs solve the accounting but assume an office team that lives in the software, and a crew in the field will not stop to fill out forms. Either way, the moment of purchase, when the information is freshest, is exactly where both approaches lose the data.

What we built

Cazullo built DWC Tracker, a construction operations system where the job is the center of gravity. Workers photograph receipts from their phones, and AI (Google Gemini) reads each receipt and turns it into a categorized expense attached to the right job. Each job carries its expenses, documents, and financial picture in one place.

• Receipt capture from the phone: one photo, and the expense exists in the system, categorized and attached to the job
• AI extraction with Google Gemini, so the office reviews and confirms instead of retyping
• Jobs carry their own expenses, documents, and financial picture
• Invoices created and tracked per job
• Dashboards showing costs per job and overall financial health for the owner

How it works in practice

A worker buys materials, photographs the receipt on the spot, and seconds later the expense exists in the system, categorized and attached to the job. The office reviews what the AI extracted instead of retyping receipts from a pile. At tax time, the records already exist, because they were created at the moment of purchase all year long.

What we learned

• Field adoption requires the flow to be one photo: anything longer dies in the field
• AI extraction is a reviewer's tool: the office confirms instead of typing, which is where the time savings actually come from
• Per-job cost visibility changes decisions mid-job, not just the reporting after it ends