Recurring cleaning schedules without the chaos: skips, holidays and swaps.

Regular cleaning is easy to schedule on the first day and hard by the third month. The base pattern seldom causes trouble; the exceptions do: skips, one-off reschedules, holiday weeks, pauses while a customer is away, a cleaner's planned leave. Each is settled by a rule that, in a small company, may live only in the owner's head and gets applied a little differently every time.

An AI assistant cannot apply a rule that nobody has written. Once the rules are on paper, a Schedule Bot can match each request to the right rule, draft the reply and propose any swap for the owner to approve. This article is about getting the rules into that state. The rules and scenarios in it are illustrations, not a record of any company's practice.

The table is Aevrion's starting template, not a description of what cleaning companies do. The middle column is the owner's to write.

Compare two versions of a notice rule. ‘Customers should give reasonable notice’ can be applied only by its author. ‘A skip needs notice of two working days; a skip asked for later is treated as a visit under the late-change terms; the owner may waive that once a year for any customer’ can be applied by a new office hire or a Bot, and checked afterwards by anyone. Both rules here are invented for the illustration.

Holiday rules work the same way. In the illustration the owner lists the closure dates and writes that a visit falling on one moves to the nearest open day that week with the same cleaner, or else the customer is offered a skip or the following week. The pause rule says the slot is held for four weeks, and that a customer who asks to cancel is offered a pause once and not pressed after that. Offering a pause first is a retention choice, not a requirement, and the owner's commercial decision.

Swap rules are the most delicate, because they touch continuity. In the illustration a cleaner's planned leave is entered as soon as it is agreed; each affected customer hears at least a week ahead and is offered either a named stand-in or a move to the cleaner's return; a customer who has asked for no stand-ins is offered the move only. Every swap is proposed to the owner and approved before any customer is told.

Grok Bot calls a saved, repeatable method a skill, and its documentation suggests six things a skill should cover: the occasion for using it, the inputs and access required, the order of the work, a way to validate the output, the form of the result and the points needing approval. Schedule rules fit that shape closely. Under Aevrion's pattern the owner asks the Schedule Bot, which the owner creates for this job, to save the written rules as a skill, with the notice period, holiday list, pause terms and swap terms as its decision rules and the owner's sign-off on every reply and swap as its approval line.

Every Bot on the account draws on the same private skill library, so the rules are written once. A rule a Bot is told to remember guides the Bot without enforcing anything; the real stop is that replies reach customers only when the owner sends them.

Before any customer sees a reply, the skill is tried on last month's actual requests, and the owner compares each draft with what was really done. Every difference is either a gap in the rule or an exception the owner had never written down.

Based on the Grok Bot documentation at docs.x.ai as checked on 15 October 2026. Its statements on Bot rules, Bot-to-Bot messaging, routine triggers, connected apps, approval expiry and payments were rechecked against docs.x.ai and Cursor's help pages on 26 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.

Consider, as an illustration, an every-other-week customer who emails on a Wednesday to skip next Monday's visit. It falls inside the notice rule; what matters is who holds each step.

Entering the change in the booking system is the one step that touches a live record. Doing it from the Bots' computer is a custom integration or supervised browser use where the vendor permits it, subject to testing. Until that has been tested, the owner makes the entry and the Bot checks the next export for it.

1. 1

   Received

   Owner

   The owner

   Advances when

   The owner pastes the customer's message into the Schedule Bot's conversation, or attaches the email file

   Ends when

   A request never forwarded stays in the inbox, so the owner clears it daily

2. 2

   Matched

   Owner

   The Schedule Bot

   Advances when

   It reads the latest booking export and the skill, and names the rule that applies and the dates involved

   Ends when

   If no rule fits, it says so and the request returns to the owner

3. 3

   Drafted

   Owner

   The Schedule Bot

   Advances when

   An editable email draft appears in its conversation, citing the rule it applied

   Ends when

   Discarding the draft ends it

4. 4

   Decided

   Owner

   The owner

   Advances when

   The owner edits and sends the reply, or rewrites it as a deliberate exception

5. 5

   Recorded

   Owner

   The owner, until an integration is tested

   Advances when

   The change goes into the booking system, and the next export shows it

   Ends when

   Closed when the Bot finds the change in that export

A Monday holiday moves every Monday visit at once, and improvising that from replies is how double bookings happen. A monthly routine on the Schedule Bot can look four to six weeks ahead for any date on the owner's closure list and draft a holiday plan: each affected visit, where the rule moves it, which cleaner's day becomes overloaded, and which customers' no-stand-in preferences the move would break.

The owner approves the plan as a whole, adjusts the overloaded days, then works through one draft notice per affected customer. Because the routine runs on its own, it stops at the plan.

A holiday-week deep clean can call for products the team does not normally carry, such as a stronger oven cleaner or a descaler. Under OSHA's Hazard Communication Standard, 29 CFR 1910.1200 (as checked on 16 October 2026), an employer trains workers whenever a new chemical hazard they have not been trained on is introduced into their work area, and keeps a safety data sheet for each hazardous chemical. The holiday plan can flag any add-on that calls for a product missing from the company's list, so the owner can deal with it before the week begins.

When a cleaner books a week off, the Schedule Bot applies the swap rule: it lists that cleaner's visits, suggests a stand-in for each from colleagues with room that day nearby, and proposes moves for customers who refuse stand-ins. The owner approves, changes or rejects each line. Only then are the notices drafted, one per customer, naming the stand-in or the new date.

Access is the part of a swap that stays outside the design. A stand-in needs a door code, a lockbox combination or a key, and those stay in the booking system or a password manager under the owner's control; the owner passes them on in person. They cannot sit in the skill, the notes or the export, because every Bot on the account works on the same computer, each file on it is visible to every Bot, and the documentation says separate Bots are not a security boundary. That includes the export: if the booking system's report carries an access-notes column, the owner removes it before the file reaches the shared folder.

Write the notice rule, the holiday list and the swap rule on a single page, then put the skip-and-reschedule workflow through the workflow automation readiness assessment, which is free and ungated. It marks that one workflow against twelve fixed dimensions, among them rule clarity, exception frequency and ownership clarity, and gives an immediate result, from not ready yet to high-leverage candidate, with the blockers and a next step. Rule clarity is the score to watch: a low one caps the result, and the remedy is the one-page rules, not software.

The guide to documenting a business process is a good companion for that page, and the reliable-workflows pillar covers routine design. Once the rules hold, Aevrion Ops can help turn them into a skill and a Schedule Bot, and test whether the booking software's vendor allows a custom integration under its API or terms.

1. Recurring cleaning schedules break on exceptions such as skips, holiday weeks, pauses and planned leave, not on the base pattern.
2. A rule an assistant can apply names a number, an exception and who grants the exception; a vague rule can only be applied by its author.
3. Saved as a skill, the written rules are read the same way by every Bot, and every reply is an email draft the owner sends.
4. Holiday weeks are drafted as one plan a month ahead, and planned swaps are proposed line by line for the owner to approve.
5. Access codes stay in the booking system or a password manager the owner controls, and never in the export the Bots read.