A two-truck septic company in spring: 140 due tanks and three closings.

Spring puts routine pump-outs and home-sale inspections on the same two trucks. A pump-out that slips a week inconveniences a customer. An inspection that slips past a closing date can hold up a house sale whose contract or local rule requires it. So the week is planned from the immovable jobs outwards, and every route has to leave room for the trips to empty the truck.

One illustrative spring in a two-truck company runs through this article: one hundred and forty customers due, three sale inspections with closing dates, and a set of AI assistants that prepare lists, route drafts, reports and disposal paperwork for the owner to check. The company, the numbers and the weeks below are an illustration built to show the design; none of it describes a real business or reports a result.

In this illustration the owner drives one truck and does the inspections; a hired driver runs the second truck. A family member takes calls on weekday mornings. Records live in a spreadsheet and a box of paper tickets, and the evenings go on reminders, invoices and disposal records.

The pillar article on the One-Man Army sets out the model this company borrows: one decision-maker in the cab, and a named AI assistant for each recurring desk job. Under Aevrion's pattern, this owner sets up three Bots in Grok Bot: a Recall Bot that owns the due list and reminder drafts, a Route Bot that drafts each day's sequence for both trucks, and a Paperwork Bot that drafts service reports and pre-fills disposal records. Inspections for the three closings are tracked by the Paperwork Bot this spring; a dedicated real-estate Bot can come later if sale work grows.

All three work on the one persistent cloud computer the account provides, so a ticket photo the Paperwork Bot saves to the shared workspace is readable by the Route Bot too, and a disposal-facility portal signed in for one is open to all three: separate Bots are not a security boundary. Each keeps its standing rules in memory, which guides the Bot but is not a technical control.

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.

The Recall Bot builds the list from the service records, not from what it remembers of last year; the documentation tells users to keep changing facts in the source system rather than in a Bot's memory. Each customer is due for a stated reason: the owner's interval since the last pump-out, an annual check under a service contract for a system with pumps or float switches, or levels recorded at the last visit that were already close to the owner's threshold. Those rules can start from EPA's SepticSmart advice to homeowners (as checked on 15 October 2026): inspection at least every three years, pumping typically every three to five years, and alternative systems with mechanical parts inspected more often, generally once a year.

The owner reads the list once and makes the calls a list cannot: which contract customers go first, which long-standing customers get a personal note, and which overdue tanks are worth a second reminder. The Recall Bot then drafts the reminder emails in the owner's words. Each sits as an editable draft that the owner sends or discards; the Bot does not ring or text anyone, because the documentation describes no telephony. Customers who reply with a preferred week are added back to the list by a person, and the list goes to the Route Bot.

In the second week of April three inspection orders arrive, each tied to a sale with a closing date: two closing in early May, and one in mid-May whose buyer's contract also sets an earlier inspection deadline the company has only been told about by email. None of those dates is the company's to move.

The owner enters each order, its closing or contract date and the date the report must reach the client. The Paperwork Bot works backwards from that delivery date to the latest inspection day that still leaves time to write the report, and hands those latest days to the Route Bot as pinned stops. How a sale inspection is run, reported and kept apart from any repair quote is covered in the home-sale inspections article in this series; in this week the point is simply that three stops are fixed before anything else is placed.

Each evening the Route Bot drafts the next day for both trucks. It starts with the pinned inspections, because in this company an inspection means opening the tank and, if the owner decides, pumping it, so it is a truck stop as well as an appointment. Then it fills the gaps with due pump-outs grouped by area, using the tank sizes from the records and the capacity the owner has entered for each truck to decide when that truck must break off for the receiving facility. The sequencing is Aevrion's design, not a Grok Bot function, and every number in it is one the owner typed in.

The owner puts the three Bots in a group chat (the documentation permits between two and six Bots in one). The Route Bot acts as coordinator: in the documentation's example the coordinator of a group is able to hand out work, so it posts the next day's draft for both trucks and asks the Recall Bot for any customer who has confirmed a week. A Bot may also message another Bot directly or hand it a task. Cursor's help pages say no switch turns off messaging between Bots; a Bot follows instructions about it.

Approval cards and email drafts do not appear inside a group, so the owner approves the route and sends reminders in each Bot's own conversation. The owner's own messages in a group may carry attachments, but ticket and lid photos go to the Paperwork Bot's own conversation, because that Bot owns the paperwork; the group is for status, and a Bot's handoff into the group can currently carry only text. The Chief of Staff role in the documentation's examples produces a digest for a human reader; a supervising Bot that manages the others is not a documented feature. The coordinator here is Aevrion's arrangement of documented parts, and the owner still reads every route.

At the end of each day both drivers hand in their tickets: levels measured, gallons pumped, condition notes and the receiving facility's slip for each load. The owner photographs them and attaches them in the Paperwork Bot's conversation. Because EPA asks homeowners to write down the levels their septic professional found, and expects the service report to record the tank's condition and any repairs done, the Bot drafts each customer's report with the levels and gallons exactly as written on the ticket, and leaves blank anything it cannot read.

It also pre-fills the disposal paperwork: for each load, the date, the truck, the customer addresses it came from and the gallons recorded, carried into the manifest or disposal record the company uses. It flags any load where the gallons on the facility slip and the gallons on the service tickets do not agree. The owner or the driver checks every field against the paper and signs. The Bot never signs a manifest, never chooses where a load goes and never corrects a number to make it agree.

On the Wednesday of the busiest week, a customer calls the office: sewage is coming up through a basement floor drain. The family member who answers rings the owner, and the owner pulls the second truck off its route. None of that touches a Bot: deciding which emergency goes first is a judgment on the phone.

Afterwards the owner tells the Route Bot what changed. It redrafts the afternoon for both trucks, keeping the pinned inspection on the first truck and moving three pump-outs to Friday, and the Recall Bot drafts apologetic reschedule emails for the three customers. If the owner had asked for this overnight through a routine, any approval it raised would have expired after about ten minutes (Cursor's help pages say thirty) and nothing would have been sent. In this company the reschedule emails sit unsent until the owner reads them, however late the backup ran.

The third inspection, the one tied to a buyer's deadline known only by email, is booked on the latest day the Paperwork Bot calculated. The owner inspects, writes the findings and decides what the report says. The Paperwork Bot formats the report from the owner's notes and drafts a short status email to the agent saying the report has gone to the client; it says nothing about the system's condition. The owner sends both.

What makes the month manageable is not that any Bot made a decision. It is that the closing dates were fixed first, the loads were planned with the disposal runs in them, and every figure on a report or a manifest came from a ticket a person had signed.

Take last spring's three sale inspections and, for each, work back from the closing date to the day the order arrived: which day the truck had to be there, and which pump-outs it displaced. Those displaced stops are the Route Bot's first test case, because they show what a pinned stop costs the rest of the week.

Aevrion Ops can work through those walk-backs with an owner and scope the recall rule, the route inputs worth recording, the Bots and where each stops for approval, plus any custom integration to the spreadsheet or septic software in use, which depends on that vendor's API or terms and on testing. Before that conversation, take the workflow automation readiness assessment and score spring scheduling as its one workflow; if nobody has written down who decides when a pinned inspection displaces a pump-out, expect ownership clarity to hold the result down.

1. In spring, home-sale inspections are pinned to closing dates and routine pump-outs are planned around them.
2. The due list is rebuilt from service records, each customer with a reason, and reminders are email drafts the owner sends.
3. A route proposal weighs tank sizes, truck capacity and disposal runs from figures the owner entered; the owner approves it.
4. Approvals and email drafts stay out of the group chat, and ticket photos go to the Paperwork Bot's own conversation, where the paperwork lives.
5. Disposal records and manifests are pre-filled from tickets and checked and signed by a person; a Bot never signs.
6. Emergencies are decided on the phone by people, and reschedule emails wait for the owner because unattended approvals expire.