Ask for the whole unit, the data plate, the surrounding space and the failure, check what arrives, and chase what is missing before an estimator leaves the yard.
An AI agent for repair photo intake is a 24/7 digital assistant that asks each customer for the specific photos a quote needs, checks what arrives, chases what is missing, and files the set against the job. It stops estimators driving across town to look at something a photo would have shown while the customer was still standing in front of it. Most customers who are asked for "a picture" send one dark close-up of the part they think is broken, which answers nothing. The agent asks for four framed shots and says what each is for: the whole unit in its space, the data plate, the clearance and connections around it, and the failure itself. It counts what came back, names what is still missing, and prompts the customer to type the model and serial from the plate while they are standing at it. A quote request either arrives complete or it does not become an estimate visit.
Requests the right four shots, checks them, and files them on the job.
Asks for four shots: the whole unit, the data plate, the surrounding space, and the fault
Explains what each photo is for, so customers do not send one blurry close-up
Counts what arrived against what it asked for and chases the gaps by name
Prompts the customer to type the model and serial off the plate while they are there
Flags jobs whose photos show work your shop does not take on
Attaches the complete set to the job in your field-service software
A free estimate visit costs a small shop ninety minutes of a technician's day plus fuel, and a fair share of them end in a job the shop was never going to take: a twenty-year-old machine, a discontinued part, or a unit wedged into a space that needs two people and a dolly to reach. Photos answer all three of those questions before the van moves. They also settle the argument that usually happens on site, because the customer described the job and the pictures show it.
A simple, three-step flow.
When a quote request comes in the agent replies with a short list of the shots it needs and one plain line on what each one has to show.
It counts what arrives against what it asked for and follows up on the gaps by name — "still need the data plate, it is usually inside the door frame".
The complete set is attached to the job, the model and serial are written to the record, and the request moves to your estimator or your booking queue.
A realistic use case with concrete timing and output.
Scenario: A three-van appliance repair shop books around 40 quote requests a week and writes off five wasted visits in a normal week. A customer messages at 10pm about a front-loader that will not drain. The agent asks for four photos: the machine in its space, the data plate inside the door, the floor and drain hose behind it, and the error code on the display. Three arrive; the agent chases the data plate and gets it eleven minutes later. The model is a 2009 unit whose pump assembly was discontinued years ago. The shop declines the job by text before 8am and keeps a 90-minute slot it would otherwise have spent driving.
Roles that gain practical value from this AI agent.
You stop losing half a morning to a machine that was never economic to repair.
Every wasted estimate visit comes straight out of a day you cannot buy back.
You arrive knowing the access, the model and the failure instead of finding out on the driveway.
The model number lands with the request, so availability is checked before anyone commits to a date.
Fewer jobs come back as "could not complete" because the photos already ruled them out.
Claims and landlord approvals both need photo evidence, and it is collected the same way every time.
Key tools and what the AI agent does inside each.
Carries the request and receives the images, so the customer needs nothing but their phone.
Stores each job's photo set in its own folder, named by customer and date.
Attaches the photos and the model number to the job record before it reaches the schedule.
Creates the quote request with the images attached so the estimator prices from what they can see.
Logs which requests came back complete, which were chased, and which were declined on the photos.
Six practical scenarios that this AI agent excels in.
Common questions about using the AI agent in workflows.
An AI agent for repair photo intake is a 24/7 digital assistant that asks each customer for the specific photos a quote needs, checks what arrives, chases what is missing, and files the set against the job. It stops estimators driving across town to look at something a photo would have shown while the customer was still standing in front of it. Unlike a form on your website or an answering service taking a message, it holds a conversation, notices the shot that never came, and asks again.
Four things, and the agent asks for them by name. The whole unit from about two metres back, so you can see what it is and how it sits. The data plate, close and in focus, for model, serial and electrical detail. The surrounding space, showing clearance, connections and how you would get the thing out. And the failure itself — the leak, the crack, the error code. One close-up of a broken part on its own tells an estimator nothing.
No. Photos make a quote faster and more accurate; they do not authorize the agent to price anything. It can read back a published range for the job type from your price book, but every reply says the same thing: the figure is confirmed once a technician has assessed the job. Nothing in the photo thread changes that, and no amount of detail from the customer unlocks a firm number.
The agent restates the published range once, explains that photographs show the fault but not the condition of what is behind it, and offers the paid diagnostic. It will not narrow the range, will not commit to the low end, and does not treat persistence as new information. Contractors consistently tell us this is a feature: the pressure that gets a firm number out of a tired human simply has no effect here.
Some will not, and the agent does not badger them. After two prompts it stops asking and offers the paid diagnostic visit instead, which is the honest fallback — if nobody can see it, someone has to come and look. The refusal is logged, so over a quarter you can see how much of your estimate schedule is made of customers who would not photograph the job.
It asks the customer to read the model and serial off the plate and type them in while they are standing at the machine, and stores both on the job alongside the image. That is more reliable than interpreting a phone photo taken at an angle in a dark laundry cupboard, and it costs the customer about ten seconds.
Attached to the job in Housecall Pro or Jobber, and mirrored into a dated folder in Google Drive so nothing depends on a technician's phone. You set how long they are kept. For warranty and property-manager work most shops keep the set for the life of the claim plus a year.
Ask for the whole unit, the data plate, the surrounding space and the failure, check what arrives, and chase what is missing before an estimator leaves the yard.