Ten hours in the sun. Then two more at the keyboard.

The Problem

Every working day ends the same way for a landscaper. Hot, tired, done. All you want is to crack a beer or take your kid fishing.

Instead, there's paperwork. The day's job sheet, scribbled on a clipboard between yards, has to get typed into the customer database. Which yards got done. Which ones got skipped. What extra work the crew picked up along the way. For this owner, that was about two more hours at a keyboard, every night, after ten hours in the sun.

The work a computer honestly does better, sitting between a man and his evening.

Why this was hard

A day's job sheet is not clean data. It is the messiest input a business produces.

• It's handwritten. Different crew members, different handwriting, notes crammed in margins, sometimes printed, sometimes not.
• It's bilingual. Part of the crew writes in Spanish. The database is in English. A person had to translate as they typed.
• It takes judgment. "Cut the hedges, pulled some weeds" is an upsell that needs invoicing. "Sprinklers were running, mower would leave ruts" is a skipped yard that needs rescheduling. The typing was never the hard part. Knowing what each note means was.

Off-the-shelf form scanners choke on this. They read characters. They don't read situations. This is exactly what a language model is good at: read, understand, write.