The crew finishes a change-order repair Tuesday afternoon. The superintendent signs the ticket on the hood of the truck, the foreman folds it into the visor, and the truck spends the next two days on a different job across town. The office learns the work happened when the folder gets emptied Friday — if it gets emptied Friday.
The gap nobody put on the schedule
Between the moment a job is done and the moment its invoice goes out, there is a lag, and in most field trades that lag is measured in days. Nobody chose it. It is the natural consequence of the fact that the record of the work is physical — a carbonless duplicate job ticket, a time-and-materials sheet with the GC superintendent's signature, a handwritten timesheet, a counter receipt from the supply house — and physical records travel at the speed of the truck they are riding in.
The number worth writing down is days from job-complete to invoice-sent. Every one of those days pushes the payment clock back by the same amount, because the customer's clock starts when the invoice arrives, not when the work ended. On commercial work the lag has a sharper edge: most general contractors process pay applications on a monthly cutoff. If your backup paperwork misses the cutoff by one day, that billing waits an entire additional cycle. A ticket that sat in a visor for seventy-two hours can quietly cost thirty days of cash.
What a job ticket has to survive
The ticket itself is having a hard week. It rides on a dashboard in West Texas sun, which fades the carbonless second copy first. It gets folded, rained on, and set under a coffee cup. And the thing that makes it valuable — the signature — is exactly the thing you cannot reconstruct later. A signed T&M ticket is often the only proof that extra work was authorized at all. Lose it, or let it fade past legibility, and a billable change order becomes an argument.
Timesheets have their own version of this. Start and stop times get written from memory at the end of the day, job codes get guessed, per-diem gets noted in a margin. Payroll runs Wednesday against whatever sheets made it in, and everything that arrives after gets reconstructed and corrected next cycle. On public work, where certified payroll reporting is required, those corrections are not just internal annoyances — the record follows the job.
None of this is negligence. The foreman's job is the job. Asking a crew to do office work at four in the afternoon has been failing for as long as anyone has tried it.
Photograph it at the tailgate
The change that actually holds up in the field is small: the moment the ticket is signed, the foreman photographs it. Tailgate, hood, wherever the light is. The photo goes to one place — a dedicated number, a shared intake, whatever the crew will genuinely use — before the truck rolls.
That is the entire field-side ask. Not an app with seven required fields, because crews will not do data entry standing in a parking lot, and every failed system in this category failed by asking them to. They will take one picture. The design has to fit inside that.
Once the photo lands, the office has the record Tuesday at 4:15 instead of Friday at noon. Everything else in this system exists to make use of that head start.
Reading the photo, and what happens when the machine is not sure
On the office side, software reads each photo and pulls the fields into a draft — job number, date, hours, quantities, whether a signature is present — staged in the billing system for a person to approve. A clean digital document reads almost perfectly. A photographed carbonless copy, shot at an angle in truck-cab light, sometimes does not.
The honest design decision is what happens then. Every extracted field carries a confidence score, and anything below the line is routed to a person, who reads the original photo side by side with the draft and types the number themselves. The system never guesses at a quantity or an hourly total. A wrong invoice costs more than a slow one: it comes back disputed, and now you are both slower and doubted. Machines drafting, people confirming, and the doubtful cases always landing on a human desk — that is the arrangement that survives contact with real paperwork.
Measuring whether it worked
Pick the metric before anything gets built. Days from job-complete to invoice-sent, averaged over a month, before and after. A second one worth tracking is payroll corrections per cycle. Both come from records you already keep, and both are numbers no vendor can spin, which is the point.
Note what this does and does not do for the bank account. It does not add revenue. It moves the same money earlier — which is what covers a Friday payroll in a slow month, and why owners who have lived through a tight February care about it more than any percentage would suggest.
What this does not fix
It cannot recover a ticket that was never written, and it cannot give an unsigned ticket authority it never had — a photo of an unsigned T&M ticket is a photo of a future dispute. It does not catch pricing errors on the ticket itself; if the foreman wrote the wrong rate, the draft will faithfully contain the wrong rate. The review queue is a real job that someone in the office has to own, ten minutes a day, and if nobody owns it the drafts pile up exactly the way the paper did. And if your billing software has no way to accept imports, the final step is a person keying approved drafts — faster than deciphering handwriting from scratch, but still a person.
Where to start
For two weeks, write down two dates per job: the day the work finished and the day the invoice went out. Nothing else. If the average gap is under two days, this page is not your problem. If it is a week, you now have the baseline that everything afterward gets measured against.