“Can a customer text me three photos and get a real quote back before I'm off the ladder?” When window cleaning owners talk about AI for window cleaning businesses, this is almost always the first question. The honest answer is mostly yes, for standard houses, as long as the system is set up to flag the jobs it shouldn't touch. Window cleaning is a route-and-season business. The calendar overflows in April and October, goes quiet in January, and fills up or not depending on how fast you answer and how reliably you call people back. That is exactly where a little automation pays off.
Below are the questions owners of one- to four-truck window cleaning companies actually ask, answered roughly in the order they come up: quoting from photos, what to ask customers for, which jobs to keep off autopilot, spring and fall rebooking, missed calls, texting rules, and what it all realistically saves.
Can AI actually quote a window cleaning job from a few photos?
Yes, if you are clear about what the AI is doing. It counts and categorizes. It does not price. A vision model looks at the photos a customer sends and returns a structured draft: number of stories, approximate pane count, window types (double-hung, casement, sliders, picture windows, skylights, storm windows), whether grids are visible, and whether the upper windows look reachable from a ladder or better suited to a water-fed pole. Your own price book, the one that already lives in your field service software, turns that draft into a quote. You glance at it and send it.
This is a faster version of what many window cleaners already do by hand. Plenty of owners quote at night by pulling up the address on Google Street View, flipping to the satellite view, and hunting down old real-estate listing photos to count windows. It works, but it eats evenings, and Street View almost never shows the back of the house, where the big sliders and the sunroom usually are. Photo intake closes that gap because the customer sends you the rear elevation.
Customers can do this without much friction. According to the Pew Research Center, 91% of U.S. adults own a smartphone. The camera is already in their hand when they're thinking about their windows.
Where photo counting goes wrong
- Grids. A double-hung window with a six-over-six grid is twelve panes, not one opening. If you price per pane, tell the model to count divided lites. Otherwise it quietly under-quotes every colonial in the neighborhood.
- Inside counts. Exterior photos say nothing about interior-only glass like shower doors, interior French doors, or a mirror wall the customer expects you to do.
- Screens. Casement screens sit on the inside and rarely show up in exterior photos. Ask the customer how many screens there are instead of trusting the count.
- Ground conditions. A photo can make a second-story window look easy even when there's a steep slope, a deck, or a boxwood hedge where the ladder needs to go.
Wondering which AI tool pays for itself first? Request a free AI audit and find out.
What should a window cleaning quote form ask the customer for?
Ask for specific shots, not "a few pictures." Vague requests get you one blurry photo of the front door. A form that works for residential window cleaning asks for these, in order:
- Street address. You can cross-check against Street View and satellite, and your software can plot the job on a route.
- Front of the house, whole elevation, taken from the street.
- Back of the house, whole elevation. This is the photo remote quoting always lacks.
- A close-up of one typical window, so you can see grids, screens, and the window style.
- Their worst window. This is where sprinkler staining, hard water spots, and paint overspray show up.
- Skylights, a sunroom, or a glass railing, if they have them.
- Checkboxes: outside only or inside and outside; screens cleaned; tracks and sills; number of stories; any steep or landscaped ground under the windows.
- Timing preference, as a window like "any weekday in the next two weeks" rather than a single date.
Most field service platforms can host a form like this. If you run Jobber or Housecall Pro, check what their request forms and integrations already handle before you buy anything new. We walked through that in detail in which AI tools actually connect to Jobber, Housecall Pro, and ServiceTitan. Some exterior cleaning companies use dedicated self-quoting tools such as ResponsiBid instead. Either route works as long as the quote ends up on the customer record in the system that runs your schedule.
Which window cleaning jobs should never go out on an AI quote alone?
Some jobs look routine in a photo and aren't. Set the intake to flag these for your eyes instead of producing a quote:
- Hard water and sprinkler staining. Mineral deposits are a restoration job. They need a test spot and their own line item. Any photo that shows white haze or spotting on the lower third of the glass goes to the owner.
- Post-construction cleans. Paint, stucco, labels, and adhesive mean blade work. On tempered glass, fabricating debris can scratch under a razor. That's a liability conversation, not an automated number.
- Fog between the panes. Cloudiness inside a double-pane unit means a failed seal. No amount of cleaning fixes it. If the intake catches the word "foggy" or "cloudy," the reply should say that up front, before the customer expects a miracle.
- Anything above a comfortable ladder or pole reach. Three-story homes, tall entry walls, and all rope-access work need a site look. For commercial high-rise, OSHA's standard for rope descent systems (29 CFR 1910.27) requires written certification of anchorages from the building owner before anyone goes over the edge. A photo form can't confirm that.
- Commercial storefront route bids. Frequency, access hours, and whether you need to be done before the store opens all matter more than the pane count.
- Historic, leaded, or wavy glass, and add-ons like solar panel cleaning where you'll want to check the manufacturer's care guidance.
The rule is simple: AI drafts the quote, the owner approves it, and flagged jobs get a human first. The same principle shows up across the trades in how contractors use AI to quote jobs faster without losing accuracy. Window cleaning just has an unusually clean split between the routine homes and the handful of jobs that need a real look.
How do I set up spring and fall rebooking texts for window cleaning customers?
This is where most window cleaning businesses leave the most work on the table. Residential customers who liked you in April often mean to book again in October, and then the neighbor's kid with a door hanger gets there first. Your job history already tells you who to contact and when. The automation just has to act on it.
Step 1: Build the list from what you already have
Export every residential customer from your field software with the date of their last clean, the services on that ticket (inside and out, screens, tracks), and their zip code or neighborhood. Leave out storefront route customers. They're already on a recurring schedule and need a different kind of message (see the FAQ below).
Step 2: Pick the trigger
Two approaches work:
- Anniversary-based: text each customer about five months after their last clean. The schedule fills evenly, but you lose the "it's spring" hook.
- Season-based: send in waves starting a few weeks before your first reliable good-weather stretch in spring, and again in early fall. In colder regions, fall rebooking has to land well before the first hard freeze, because water-fed pole work gets unreliable once rinse water starts freezing on the glass. In warmer markets, fall demand tends to bunch up before the holidays, when people want clean glass for hosting.
Step 3: Send in route-dense waves, not one blast
If you text 400 past customers on the same Tuesday and a third reply, you've just overbooked three weekends and crammed your driving into one week. Send in batches of 40 to 60, grouped by neighborhood. Have the system offer each batch dates when a truck is already on that side of town. That's where the AI earns its keep: it reads the reply, matches it to open slots near other booked jobs, and proposes something specific.
Step 4: Write it like you'd text them yourself
Pull in details from the last job so the message doesn't read like a mass blast:
“Hi Dana, it's Mike from Clearview Windows. Last April we did your windows inside and out plus the screens. We're in Oak Hollow the week of the 14th. Want us to put you down for Tuesday or Thursday? Reply STOP to opt out.”
(The names are placeholders. Use your own voice.) Follow up once, about four days later, if there's no reply. Then stop. Replies like "outside only this time" or "can you add the gutters?" should update the draft ticket and come to you for a quick look. If you want recurring customers locked in instead of re-asked every season, the tradeoffs are covered in AI for maintenance plans and service agreements. A twice-yearly window plan is one of the easiest service agreements to sell.
What does missed-call text-back do for a window cleaner who is up a ladder?
It answers for you when you physically can't. You shouldn't be taking calls from the top of a 28-foot extension ladder or with a pole in both hands. So the phone rings out, and the caller, who is usually shopping around in spring, dials the next company on the list. Missed-call text-back sends a text within seconds of the unanswered call:
“Sorry we missed you. We're on a job and can't pick up safely. Are you looking for a quote or to reschedule? If it's a quote, this link takes about two minutes and you can add photos: [link]”
For window cleaning in particular, a simple text-back often covers most of what a full voice receptionist would. Most inbound calls in this trade are quote requests or rescheduling, not emergencies, and a text thread that ends in the photo form is exactly the outcome you want. Weather reschedules fit here too. When a storm cancels a day, the same text channel can offer the affected customers their next open slot. There's more on how this works at missed-call text-back for service businesses. If you're a one- or two-person operation, these rules for running automation from the truck are a good match for a window cleaning setup.
Is it legal to send automated rebooking texts to my window cleaning customers?
Generally yes, if you set it up properly. This is a summary, not legal advice:
- Get consent at booking. Put a checkbox on the quote form and the job confirmation that says you'll text about scheduling and seasonal service. Keep a record of it.
- Know the difference between informational and marketing texts. An appointment reminder is informational. A "spring special" is marketing. Under the federal Telephone Consumer Protection Act, automated marketing texts require prior express written consent. The FCC's guidance on unwanted calls and texts is the plain-language starting point.
- Always honor STOP and include opt-out language in the first message of any sequence.
- Register your texting number. Business texting from software platforms runs over A2P 10DLC. Your provider handles the registration, but carriers filter unregistered traffic, and filtered texts never arrive.
- Watch state rules. Some states are stricter than federal law. Florida, for example, limits covered marketing messages to 8 a.m. to 8 p.m. and caps them at three on the same subject within 24 hours. That matters if your spring campaign includes a follow-up.
What does AI realistically save a small window cleaning business?
The savings come in hours and in jobs you don't lose, not in headcount. For a small window cleaning company, the realistic wins look like this:
- Evening quoting time. Street View counting and phone tag turn into reviewing a draft and tapping approve.
- The January text marathon. Many owners spend the slow season scrolling last year's customer list and texting people one at a time. Batched, route-grouped rebooking takes that over.
- Calls that used to go to voicemail. Each one is a spring quote that may never have come back.
- No office hire. A part-time person answering phones and chasing rebooks is the usual next step for a growing window company. These three automations handle much of that work.
Don't take anyone's word for the numbers, including mine. Measure three things for one season before and after: median time from quote request to quote sent, the share of last year's residential customers rebooked by the end of your spring rush, and the share of missed calls that got a reply within five minutes. If those don't move, change the setup or turn it off. For a wider frame on judging the return, see this honest ROI breakdown for service owners.
Where should AI stay out of my window cleaning business?
Keep it away from any decision that depends on standing in front of the building. That covers ladder placement, whether a pitch is safe to walk, whether a pane can take a blade, how you handle a customer who says you cracked a window, and the relationship with the property manager who controls your biggest storefront route. AI is good at counting, sorting, reminding, and following up. Your crew's judgment on the glass is what customers are paying for, and no photo model replaces it.
If you're starting from zero, set things up in this order: missed-call text-back first (fastest to set up, protects spring leads), then the photo quote form, then the fall rebooking waves once you have a season of clean job history. That narrow, owner-approved sequence is how we approach setups at Turnkey AI.
Frequently asked questions
Do I need window-cleaning-specific software to do this?
No. A general field service platform like Jobber or Housecall Pro, a photo-capable request form, and a missed-call text-back tool cover all three automations, as long as quotes, customer history, and texts all land on the same customer record.
Will customers actually send photos?
Most will if you ask for specific shots (front, back, one close-up, worst window) instead of 'some pictures.' For customers who won't, keep the Street View fallback and treat the back of the house as unconfirmed.
Should storefront route customers get the same rebooking texts?
No. Route customers are already on weekly, biweekly, or monthly cycles. Send them service notices and weather reschedules instead of seasonal rebooking offers.
How early should I start spring rebooking texts?
Start the first wave a few weeks before your first reliable good-weather stretch so the calendar is built before demand peaks. Warm markets can start in late winter. Northern shops usually wait until freeze risk has passed.
Can the AI tell whether a window just needs cleaning or has hard water damage?
It can flag likely staining from a close-up photo but can't confirm it. Treat any flag as 'needs a test spot on site' and quote restoration separately after you've seen the glass.