Partly. AI can group houses into days, pick which to move after a washout and word skip messages, but it cannot hold the round: round software or a tidy spreadsheet does that. Single-user round apps cost about the price of two cleaned houses a month; ServiceM8 and Jobber start at $29 and climb with job count or features.
The useful split is this. The software remembers who is on which cycle, when they were last cleaned and who has paid. The AI sits on top for the fiddly judgement jobs: grouping streets, reshuffling after a washout, and wording messages that do not sound like a form letter. Nothing the AI suggests should change the round until you have checked it against the last-cleaned dates in the software.
Why a round is not the same as a diary
A window cleaning round is a repeating list rather than a set of appointments. Each house has a frequency (every four or eight weeks, sometimes six, twelve or only twice a year), a price, an access note and a last-cleaned date. The calendar is generated from that list, and it has to survive bad weather, holidays and customers who are out.
General scheduling tools are built around the appointment: one customer, one job, one slot. Round tools are built around the cycle. That difference decides most of the choice, and it is why a cleaner of 400 houses can be miserable in a tool that suits a plumber perfectly. The mistake to avoid is letting an AI assistant plan the next cycle from scratch, because it will happily produce a tidy week that ignores the fact that half the houses were cleaned three weeks ago.
A 420-house round, worked through
Take a solo cleaner, as an illustration, with 420 houses on a four-weekly cycle at an average of $18 a visit. Working 20 days in a four-week cycle, that is 21 houses a day. Gross per cycle: 420 × $18 = $7,560, or about $378 a day.
Now add the things that go wrong. Say 6% of visits are skipped because nobody was in or access was blocked, which is 25 houses a cycle. Rain stops work for two days in a typical cycle, which is 42 houses to move. And about 8% of customers are late to pay, so 34 houses carry an unpaid balance at any time.
The cost of each, in the cleaner's own time: skips are rebooked by hand at about four minutes each, which is 100 minutes. Rain reshuffles take about 45 minutes per washout day to move houses, text customers and check the route, so 90 minutes. Chasing 34 balances at three minutes each is another 100 minutes. In total roughly five hours of admin per cycle, or an evening a week, before any planning. That is the target. A reasonable aim is to cut it by half, mostly from the rain reshuffles and the skip and payment messages, where drafting help saves the most time.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowWhat can AI actually do on a round?
Four jobs are worth handing over, in this order of usefulness:
- Sorting a new batch of houses into daily groups. Useful when you buy a round or add a new street.
- Reshuffling after a washout. Moving a day's houses without breaking the cycle.
- Rebooking skipped houses. Deciding when to return and what to tell the customer.
- Drafting payment and price-rise messages. The wording, not the sending.
It is poor at anything that needs live data or memory. It will not know a road is closed, that a customer has a dog in the garden on Tuesdays, or that you cleaned that street 11 days ago. You can supply some of that in the prompt, but the round app remains the authority.
Grouping streets into days
If you are loading 80 new houses, paste the addresses, without names, and ask for groups of 20 to 22 that sit close together. Illustrative prompt:
Here are 80 house addresses for a four-weekly window cleaning round.
Group them into four days of about 20 houses each, keeping streets
together and listing each day's streets in a sensible walking or
driving order. Do not split a street across days unless it has more
than 12 houses. Flag any address you cannot place.
[paste street names and house numbers]A realistic output might come back with four groups, each with a short list of streets and a note such as "Day 3 has 23 houses; moving numbers 40 to 46 of Mill Lane to Day 2 would balance it". What you check: open each day in a map app and look at the route yourself. An assistant that cannot see a map can place two streets next to each other in a list that are actually separated by a river or a rail line. Treat its grouping as a first draft and correct it in ten minutes, which is still faster than doing it blind.
Reshuffling after rain
Rain is the main source of chaos. The rule that keeps the cycle intact is simple: a washed-out day is moved to the next dry day, and everything after it shifts by one, unless you have a spare day. Some round apps do this in one click. Round Monster, for example, lists a one-click rain-day reschedule among its features. If your tool does not, an assistant can work out the shift from a list.
Illustrative prompt: "Here is Tuesday's list of 21 houses and the next two weeks' plan. Tuesday is washed out. Move the houses to Friday, which has 14 houses planned, and tell me which of Friday's houses to push to the following Monday so no day exceeds 24. Keep each customer within two days of their usual four-week date." The answer you want is a revised list plus the customers to text. Check one thing every time: that no customer ends up cleaned less than 25 days after their last visit or more than 32, since beyond that, a four-weekly customer starts to notice.
Rebooking skipped houses
A skip is a house you could not clean. The sensible rule has three steps: leave a card or send a message the same day, offer a return within the same week if the round passes nearby, and otherwise slot the house into the next cycle with a note. The AI's role is the wording. A sample skip message, as an illustration:
Hi [first name], we called today for your window clean but couldn't get to the back windows because the side gate was locked. If you'd like us to come back this week, reply with a day and we'll fit you in. Otherwise we'll clean them on your next visit in four weeks at no extra charge. [Your name]
What to fix: do not let the assistant promise a free return visit unless that is your policy. It will often add a "we'll be back tomorrow" line from politeness. Strip it. For the broader question of keeping regular customers booked, can AI help a cleaning company rebook more regular clients has the follow-up wording.
What round software costs, tool by tool
Prices below are list prices from each vendor's page in November 2026. They change, so confirm before you sign. Both round apps price only in their home currency, so I give their figures as printed rather than convert them.
| Tool | Type | Published price | Round features to check |
|---|---|---|---|
| CleanerPlanner | Round app for window cleaners | Three tiers at 19, 29 and 49 a month, priced only in the vendor's home currency; unlimited rounds and jobs on all; 30-day free trial | Imports from several other round apps and Excel; the 29 tier adds text messages, bank debit collection and bulk email but stays single-user; the 49 tier includes five users |
| Round Monster | Round app for window cleaners | Solo 24 a month (one user), Teams 59 (two users), Monster 119 (three users), extra seats 20 each, in the vendor's home currency; 14-day free trial, no card; only accepts companies registered in its home country | Frequencies of 4, 6, 8 and 12 weeks, yearly and custom; one-click rain-day reschedule; skip reasons; automated invoice chasing; phone-based worker app |
| ServiceM8 | General field-service tool | $0, $29, $79, $149, $349 a month by new jobs per month (30, 50, 150, 500, 1,500+); unlimited users on paid plans | Recurring jobs on every plan: set a job up once and it re-creates itself on schedule; strong on quoting and invoicing; every new job card uses a job credit |
| Jobber | General field-service tool | Core $29 to $49 a month for one user depending on billing; Connect, which adds route optimisation, $99 to $139; Grow and Plus cost more | Recurring jobs repeat weekly, every two weeks, monthly or on a custom schedule, billed per visit or at a fixed price; route optimisation reorders only visits with no fixed time, up to seven days ahead |
Two points stand out. First, the dedicated round apps do not charge by job count, so a 400-house round costs the same as a 100-house one, while ServiceM8's bill steps up as houses are added. Second, Jobber can hold a repeating visit on a custom schedule, but its route optimisation works over a window of up to seven days, so it helps with the order of houses within a week and does not rebalance a four-week cycle. That is not a flaw in the product; it means a round with 400 houses needs either a round app or a careful trial of Jobber's custom schedule with your real frequencies loaded.
ServiceM8's pricing is built on new jobs per month, and its pricing page says every new job card uses a job credit. Recurring jobs re-create themselves as new jobs, so expect each visit to count: a 420-house four-weekly round is about 455 visits a month (a month is a little longer than four weeks), which needs the 500-job Premium tier at $149 a month. Ask ServiceM8 to confirm how recurring visits are counted before you commit. ServiceM8 does list an AI quote-and-invoice feature on every plan, which helps with one-off jobs such as gutters and conservatories more than with the round itself.
Compare that with the sum earlier. Our cleaner spends about five hours of admin per cycle, which at a conservative $20 an hour is $100; halving it is worth about $50 a cycle. Every single-user round app tier above costs less than that saving, while ServiceM8 at $149 costs about three times it and only pays if it also replaces your quoting and invoicing tools. So the question is less whether to buy and more which tool fits the way you work. For the wider decision between a person on the phone and software, AI receptionist or chatbot for a cleaning company is a useful companion.
Four-, eight- and twelve-weekly houses on one street
Real rounds mix frequencies, and that is where plans quietly break. Suppose one street has 12 houses. Seven are four-weekly, four are eight-weekly, and one is a twelve-weekly customer. Over six cycles the street needs 12, 7, 11, 8, 11 and 7 cleans, and then the pattern repeats. If your planner counts houses per street, you will block out time for 12 houses on a visit when only 7 are due.
The fix is to plan by due date rather than by street. Each house has a "next due" date; the day's list is everything due in that window, grouped by area. A round app does that for you. In a spreadsheet, you need a "next due" column calculated as last cleaned plus frequency, and a filter. An assistant can write the spreadsheet formula, which is worth doing once. Illustrative prompt: "I have columns A address, B frequency in weeks, C last cleaned date. Write a formula for column D that gives the next due date, and a formula for column E that shows OVERDUE if D is more than 3 days before today." A typical answer is =C2+B2*7 for the next due date and =IF(D2<TODAY()-3,"OVERDUE","") for the flag, which is right. What you check is the result on five houses you know, because a formula that is wrong by a week will quietly shift the whole plan; the usual slip is a formula that adds the frequency as days instead of weeks.
A skipped house that was not a skip
A realistic mistake: a cleaner marks a house as skipped because the customer was out and the gate was locked. The software then moves the house to the next cycle, four weeks later. But the customer, who pays by bank debit on the first of the month, is charged anyway for a clean they did not get. They complain.
The lesson is that skips and payments live in different parts of the system, and the planner does not always tell the billing side. Check how your tool treats a skipped visit when collection is automatic. The safe setting is that a skip creates no charge until the next completed clean. Round Monster, for one, says it collects the debit when a job is signed off, which is the behaviour you want; confirm the equivalent in whichever tool you trial. If your tool cannot do that, pause automatic collection for that customer when you mark a skip.
Price rises and the message that goes with them
Once a year the round needs a price review, and AI helps with the wording. A rise of $1 on an $18 visit across 420 houses adds $420 a cycle, about $5,460 a year over 13 cycles. A common mistake is raising everyone at once and losing ten houses: at the new $19 price that is $190 a cycle gone, so the rise nets about $230 a cycle instead of $420. Check that sum before you send. Here is an illustrative message you can adapt:
Hi [first name], from your next clean on [date] the price for your window clean will be [new price], up from [old price]. It's the first change in [length of time]. If you'd like to talk it through or change how often we come, just reply. Thanks, [your name]
Ask the assistant for three tones, from brisk to warm, and pick one. What to remove: any phrase that blames costs or suggests the customer has no choice. And never let the assistant invent a reason, such as fuel prices, that you have not decided to give. Text and leaflet versions are in writing a window cleaning price increase letter with AI.
Card, bank debit or invoice: how a round gets paid
Round apps handle money in three ways: card payment on completion, recurring bank debit, and invoicing with chasers. CleanerPlanner's mid tier lists a link to a bank-debit provider, and Round Monster lists both a direct-debit provider and card payments, plus automated invoice chasing. General tools such as Jobber and ServiceM8 handle card payment and invoicing, but check the bank-debit options for your market and the fee per collection.
An AI assistant can help here only with wording: a payment reminder, a message to a customer whose debit failed, a note explaining a change of date. It should never see card or bank details. If a customer pastes those into a message, delete them from any assistant chat. For the chasing wording itself, see chasing unpaid invoices with AI for trades.
Spreadsheet, round app or field-service tool: three cleaners
Here is how the decision might go for three different cleaners, as an illustration.
A solo cleaner with 130 houses. A spreadsheet with a next-due formula, a payment link in each message and an assistant for wording is enough. If the sheet starts costing more than an hour a week to fix, move to a round app at the lowest tier.
A cleaner with 420 houses and one helper. A dedicated round app. The count is high enough that job-based pricing is expensive, and a helper means two people need the same list. Check the multi-user tier, since the mid tier of one round app is single-user only.
A firm with three vans, a mix of window rounds, gutters and commercial contracts. A general field-service tool for the commercial and one-off work, plus a round app for the recurring domestic list, until the volume justifies something bigger. Two systems cost more attention, so plan a weekly ten-minute check that they agree.
In all three cases, trial before you commit. A free trial of 14 or 30 days should include loading 30 real houses and running one washout day, because that is where you find out whether the tool fits.
A seven-day test before you trust any tool
Do not judge round software on its demo. Load 30 houses from your actual round, with real frequencies and last-cleaned dates, and run these five checks:
- Mark one house as skipped and see where it goes, and whether it is charged.
- Declare a washout day and see how many clicks it takes to move the day.
- Add a house with an eight-weekly frequency and check that it appears in the right cycle.
- Send yourself a reminder message and read it on your phone.
- Export the data and open it in a spreadsheet.
If any of these takes more than a few minutes, or can only be done by contacting the vendor, that is a sign about how a normal Tuesday will go. For comparison across a wider range of tools in a related trade, see AI route planning for landscaping crews, which covers the similar problem of planning visits by area.
If you would rather work through your own round with someone, that is what my AI implementation consultation covers; the contact page has the details.
Questions window cleaners ask about round software and AI
Do I need round software if I use a spreadsheet?
A spreadsheet works until you pass roughly 150 to 200 houses or hire a second person. The pain points are skipped houses, rain reshuffles and knowing who has paid. If you spend more than an hour a week fixing the sheet after bad weather, software will probably pay for itself.
Can ChatGPT or Claude plan my round on its own?
It can sort addresses into sensible daily groups and draft customer messages, but it cannot see live traffic, remember what you did last cycle or send texts. Use it for drafting and checking, and keep the master record in a round app or a well-kept sheet.
How should I handle a customer who is never in?
Set a rule: two skips in a row triggers a message asking whether they want to keep the slot. After a third, move them to an on-request list. Never delete their record, because many come back after a house move or a change in the family.
Is round software worth it for a solo cleaner?
If you have more than about 100 regular houses, usually yes. Cheaper round apps cost less than one cleaned house a week and save the evening you would spend sorting rain days and chasing payments. Under 100 houses a spreadsheet and a payment link can be enough.
Can I move my data if I switch software later?
Most round apps import from a spreadsheet and some from rival round apps. Before you commit, ask for an export of your data in CSV format and check it contains addresses, frequency, price and last-cleaned date, because those four fields are what you cannot rebuild from memory.
Further reads
- Photo Checklists and Quality Control for Cleaning Teams Using AI — Photo checks for a second cleaner once the round grows.
- AI for Window Cleaning: What Helps a One-Van Business and What Doesn't — Where else AI helps a one-van window cleaner, and where it does not.
- Landscape Business AI: Work Out Job Profit by Crew From Timesheets — Another round-based trade working out profit per crew.
- Airbnb Cleaning Schedule: Coordinate Cleaners and Changeovers — Scheduling tight same-day jobs with more than one cleaner.
- How to Price Window Cleaning: Per-Window Rates and AI Quotes From Photos — Price new houses from photos before they join the round.
- Mobile Hairdressing Business: Diary, Travel Time and Rebooking With AI — Plan a mobile hairdresser's diary around drive time: patch days, same-street batching and rebooking texts, with two Tuesdays compared in minutes and income.
- Commercial Window Cleaning Contract: Proposal, Access and Safety With AI — Survey notes become a scope table, access and safety clauses, frequency pricing and an insurance section, checked five ways by AI before a lawyer sees them.
- Gutter Cleaning Software: Reminders, Routes and Recurring Bookings With AI — A reminder, route and rebooking workflow for gutter cleaners: recurring jobs, reminder timing, patch days, AI-drafted replies and opt-outs.
- How to Get Window Cleaning Customers: Leaflets, Local Search and AI — A street-by-street leaflet test, a Google profile for a van with no shop, and review and referral wording, with the minutes AI really saves.
- How to Price Commercial Window Cleaning: Per-Pane Quotes From Your Labour — Build commercial window quotes from pane counts and your own timings: a per-pane price list, access costs, frequency discounts and a margin check.
- Can AI Handle Pest Control Route Optimisation for Quarterly Customers? — How to cluster quarterly pest control visits by area with AI, then check the drive-time saving with a real routing tool.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ServiceM8 pricing page, Jobber pricing page and recurring-job help article, CleanerPlanner and Round Monster pricing pages (checked November 2026).