The OnSight Agent
No trigger picker and no rule syntax. An agent that already knows your job types, your price book, your crews and your pricing program writes it for you, and waits for you to turn it on.
Try one
Anything it raises goes through the same pricing engine your staff use, under the permissions of whoever built it. Every run lands in the activity log — what fired it, what it changed, what it sent and to whom.
What people ask it for
Each sentence below is the whole configuration step. There is no second screen.
“Text the customer when the tech is on the way, but only inside working hours.”
Fires on on the way, checks your own opening hours, holds anything outside them until morning.
“Tell me when a crew is running more than forty minutes behind.”
Compares actual starts against routed times, and messages the dispatcher once per crew, not once per stop.
“Ask for a review after a five-star job, and never after a callback.”
Waits a day, skips anyone with a job reopened this month, and never asks the same person twice in a year.
“Take a deposit before anything over two thousand dollars goes on the board.”
Holds the scheduling step until the deposit is taken, priced through your own program, with the pay link on the estimate.
“Flag any card on file that expires before the next scheduled visit.”
Reads the plan’s next run against the card’s expiry, weeks before a charge would decline.
“Put every job that finished without photographs on my desk.”
Checks completed work for attachments, ignores the job types where photos were never the point, reports the rest daily.
“When an estimate has been open a week with no answer, follow it up once.”
Resends their own copy with the original photographs, then files it in the Vault rather than chasing twice.
“Warn me if today’s schedule has a stop with no address on the map.”
Runs before the crews leave, names each unmappable stop with its reason, and offers to locate the fixable ones.
No developer, no integration, no support ticket. Typed in by whoever wanted it, reviewed on screen, turned on.
Where it stops
The interesting part is not what it can do. It is what it cannot.
Thirty days later
Writing the automation takes a sentence. This is the part that pays for it — three of them running, and what each has actually done since it was switched on.
Delgado Landscaping is the example business used throughout this site, and these are its thirty days — not an average, and not somebody’s audited results. Every run above is also a row in the activity log, with what fired it and what it changed.
Book a demo
Bring the thing you keep meaning to automate.
Thirty minutes. Describe it out loud and watch the agent write it against data that looks like yours.