AI for house cleaning businesses: quotes, recurring visits and team changes.

A house cleaning company does not keep a regular customer with the first clean. It keeps them by repeating it: the same cleaner, working from the same list of priorities, on the same day of the week. Much of what can go wrong in the office is a break in one of those three threads: a quote that did not match the house, a visit moved without warning, a new face at the door with no idea which room is off limits. The office also holds a fourth thing, which is how the cleaner gets in: a key, a lockbox, a door code, sometimes an alarm code. That record would do the most harm in the wrong place.

Supervised AI assistants can carry much of the paperwork around that continuity: quotes drafted from the owner's rate card, a notes file for each home, the owner's written rules applied when a visit moves, a layout of who could take over a home when the team changes, and a list of the products on the carts beside their safety data sheets. The owner keeps the rest: any home the card does not cover, every access code, who cleans where, complaints, chemical training and whether a cleaner is an employee, some of it with advisers.

Name the threads before deciding what to hand off, because each lives in a different place and breaks in a different way. The first three suit an assistant; the fourth is kept away from it, for the reasons in section 03. The table is Aevrion's framing, not an industry standard.

Cleaning inquiries arrive with most of the quote already inside them: bedrooms and bathrooms, rough size, pets, how long since the last thorough clean, how often the customer wants a visit, and add-ons such as the inside of the oven, the fridge or the windows. A move-in or move-out clean, a first deep clean before a regular plan, or a house just out of a renovation is priced differently, and a home like that is worth seeing, in person or through photos, before any figure is promised.

Under Aevrion's pattern the owner creates a Quote Bot for this work. In Grok Bot a Bot is a named, durable teammate with one job and its own conversation. It works on a persistent cloud computer shared by every Bot on the account, with one browser, one set of sign-ins and a shared workspace folder. The owner keeps the rate card as a file in that folder: base prices by bedroom and bathroom count, the rate for each visit frequency, the add-ons, and the conditions that take a home off the card. The Quote Bot opens the card every time rather than trusting recall, since the documentation advises keeping changing facts in the source system rather than in a Bot's memory.

Take an inquiry, invented for this illustration: a three-bedroom, two-bathroom house with two dogs, asking for a first deep clean and then a visit every other week, with the oven added to the first visit. The Quote Bot drafts both prices from the card, the initial deep clean and the every-other-week visit, shows which line each figure came from, and lists what it could not price. The card has no line for heavy pet hair on upholstery, so that becomes a question for the owner, not an estimate. The quote reaches the customer as an editable email draft in the Bot's conversation; the owner reads it, changes what needs changing and chooses Send or Discard.

Off-card homes are where under-quoting happens. The Bot's job there is to say so early and list what the owner will want to see on a walkthrough or in photos: kitchen and bathrooms, floor types, anything stored that must be moved. The price comes from the owner. Reading inquiries straight out of a booking or quoting product would be a custom integration, or supervised browser use where the vendor permits it, subject to testing; an owner who pastes the inquiry into the conversation, or attaches the form email, needs neither.

Based on the Grok Bot documentation at docs.x.ai as checked on 15 October 2026. The trade examples are Aevrion's illustrations, not documented Grok Bot use cases. Aevrion Ops is an independent consultancy. It is not affiliated with, endorsed by or a partner of xAI, SpaceXSI or Cursor. Grok Bot is named here only to describe it.

The notes that make a home's clean repeatable are the ones a covering cleaner needs most: rooms to leave alone, surfaces that take only a particular product, where the vacuum lives, the cat that must not get out, the order of priorities when time runs short. Aevrion's pattern keeps one notes file per home in the computer's shared workspace, updated from the walkthrough, from the customer's requests and from what cleaners report to the owner. Every Bot the owner creates can read it, which is exactly what makes it useful.

That is also why access details must stay out of it. The documentation is plain about the computer: every Bot on an account uses the same one, files on it are visible to every Bot, browser sign-ins are shared, and separate Bots should not be treated as a security boundary. It advises against putting a credential or file on the computer if another Bot should not be able to use it. A door code, lockbox combination or alarm code written into a home's notes would be open to every Bot the owner runs and every routine those Bots own.

So the owner decides where access details live, and in this design that place is under human control rather than on the Bots' computer: the booking system, a password manager the owner runs, or a locked key register. The notes file says only that access is arranged and who briefs the cleaner. One check matters. If the Bots' browser is signed in to the booking system and that system shows the code beside the address, the code is on the shared computer after all. Whether a booking product can hide that field from a particular login is a question for its vendor; where it cannot, that sign-in stays off the Bots' computer. For keeping a set of credentials apart, the documentation points to a separate user account with its own computer, not a separate Bot.

Recurring work produces a steady trickle of exceptions: a customer away for two weeks, a holiday week, a cleaner off sick before the first visit. Aevrion's pattern for them is a Schedule Bot that applies rules the owner has written, such as a notice period for skips, drafts the replies and proposes swaps for the owner to approve. Two companion articles hold the detail: one on writing those rules down so an assistant can apply them, and one that follows an owner who also cleans through a sick day.

The heavier version is a team change. When a cleaner leaves, or cuts back to fewer days, every home that cleaner covered breaks the first thread at once, and for good. How a Bot proposes cover from the roster and the home notes is shown step by step in the sick-day article; a departure adds two things a sick day does not need.

The first is the knowledge that leaves with the cleaner. In this illustration a cleaner with a dozen regular homes gives notice, and the Schedule Bot reads each of those homes' notes files for gaps: no products to avoid recorded, no order of priorities, nothing on pets or rooms left alone. It turns the gaps into a short question list per home, which the owner works through with the departing cleaner in the final weeks and records in the notes, because much of what that cleaner knows exists nowhere else.

The second is the introductions. Who takes which home is the owner's decision, and so is the order in which customers hear about it. For each home the Bot drafts an introduction email naming the new cleaner and the first visit; the owner edits and sends each one. A customer who replies with a concern goes back to the owner, not into the Bot's next draft.

Cleaning products are chemicals, and a company whose employees use hazardous ones has duties under OSHA's Hazard Communication Standard. OSHA's hazard communication page, as checked on 16 October 2026, says that every employer with hazardous chemicals in the workplace must have labels and safety data sheets for exposed workers and train them to handle the chemicals appropriately. The standard itself, 29 CFR 1910.1200, asks for a written program listing the hazardous chemicals present, and for training when a worker is first assigned and whenever a new chemical hazard is introduced into their work.

Several provisions read differently for a team that works in other people's homes. Where employees travel between workplaces during a shift, the standard allows the sheets to be kept at the primary workplace, provided the employer ensures workers can obtain the information immediately in an emergency. A spray bottle filled from a labeled concentrate needs its own label unless it is only for the immediate use of the person who filled it. And the standard exempts consumer products used as the manufacturer intended, where the employer can show the duration and frequency of exposure are no greater than a consumer's. Whether that exemption fits a cleaner using a household product in several homes a day is a judgment for the owner, with a safety adviser where needed; this article does not make it.

These are federal OSHA rules; OSHA's state plan page notes that states running their own approved plans must be at least as effective, so an owner in one of those states checks the state plan too. The assistant's part is a register: a product list with each product's name, the date of its sheet and a copy or link, and the training the owner has recorded for each cleaner. When a supply receipt the owner forwards shows a product that is not on the list, the Bot flags it: no sheet on file, nobody recorded as trained. What goes on the carts, and the training itself, are the owner's.

The label on a working relationship has legal weight. The US Department of Labor's Wage and Hour Division, as checked on 16 October 2026, states that employers are responsible for determining whether a worker is an employee under the Fair Labor Standards Act, that misclassified workers may not receive the minimum wage and overtime they are entitled to, and that misclassification is a serious problem. That determination belongs to the owner and a qualified adviser. This article offers no view on it, and neither does any assistant described here.

In this design the Bots stay on the operating side of that line. They draft rosters and reassignment sheets; they do not decide hours, pay or status, and payroll is left out of scope entirely.

Start small, as the documentation advises: one Bot owning an outcome, and a second only when a specialist role has settled. In a cleaning office the Quote Bot comes first, because quoting is self-contained and every draft passes through the owner before it leaves. The Schedule Bot follows once the rules for skips, holidays and swaps have been written down. The product register is light enough to sit with the Schedule Bot rather than earn a Bot of its own.

Both Bots share one computer, which is why the access question is settled before either is created. A routine, such as an evening sweep for schedule requests still awaiting a reply, is scheduled at least five minutes from the next, keeps its twenty most recent run records and has no documented automatic retry or failure alert, so the owner reads its run history.

The workflow automation readiness assessment costs nothing and sits behind no form. The roster above starts with the Quote Bot, so rate-card quoting goes through it first. It rates that workflow on twelve fixed dimensions, rule clarity, ownership clarity and security and compliance readiness among them, and answers straight away with one of four verdicts, the obstacles it detected and what to do next. If the conditions that take a home off the card have never been written down, expect rule clarity to score low; if inquiries and home notes still sit in the same spreadsheet as the door codes, score security and compliance readiness honestly too. A weak mark on either caps the verdict, and both are fixed on paper, not in software.

With a favorable verdict, the scoping questions for Aevrion Ops are concrete: what the Quote Bot and the Schedule Bot each read, where each hands back to the owner, where the access details will live, and whether a custom integration with the booking software is allowed by that vendor's API or terms and holds up in testing.

1. A regular cleaning customer pays for continuity: the same cleaner, the same home checklist and the same day, and the office protects all three.
2. A Quote Bot can draft quotes from the owner's rate card as email drafts the owner sends, and it stops at any home the card does not cover.
3. Home notes can sit in the Bots' shared workspace; door, lockbox and alarm codes stay somewhere the owner controls, off the shared computer.
4. Hazard communication duties for cleaning products, and every decision about a worker's employment status, stay with the owner and their advisers.
5. The readiness assessment starts with rate-card quoting; unwritten off-card conditions, or door codes kept beside the inquiries, can cap its result.