Jon ONeill @HouseHackerJon
I’m automating as much office work as I can at my Plumbing company using Grok @bot
Here are the agents I’ve “hired” and a little bit of how I used them all in our company
Thread 🧵
Atlas, My right hand man. I basically try and treat him like I would our real life office manager. I will turn on voice typing and I will talk a lot.
After a full day of usage I asked for feedback on how I could be better at giving instructions and was surprised by his response:
The first thing he actually encouraged me in was to continue to give him long responses because he found that the context was important and helpful
The primary change he wanted from me was to end my visionary messages with concrete acceptance criteria or a definition of done.
Dana Dispatcher - the first bot I ever fired (my fault)
We initially hired Dana to be our end-to-end dispatcher.
I had well written instructions and job scope.
Unfortunately, it was just too much for her. She was slow, very slow, and would make mistakes and could not handle it
Marc: Plumbing Manager:
Marc was our answer for overloading and overburdening Dana.
Atlas and I decided we needed a manager who would specialize in overseeing the plumbing companies agent team.
Assuming I put other projects on Atlas’s plate marc is a bit overkill right now.
The other five agents are really nothing more than domain experts.
We looked at breaking out Dana‘s old job into channels of expertise, separated by software.
- Gary listens for new work orders in Gmail
- Steve check our board for capacity in ServiceTitan and Books jobs there
- Quinn text our customers appointment offerings and if they accept it, then she sends that information back to Steve to confirm the job booking
- Sarah makes a Slack post in a dedicated channel and keeps the real life Office team updated in Slack, she raises issues too
All in all having them separated out this way has been fairly efficient.
They still make mistakes from time to time, but what I’m finding is they’re not any worse than mistakes we’ve had real life humans make when we’re training or onboarding them
So we teach, train, correct.
Our goal this Sunday is to have a full team of agents working when the rest of the staff is off.
We don’t work Sundays.
Usually, Monday mornings are a scramble and I hope Atlas and Marc can change that!
What you need
- Connected apps
- Gmail, ServiceTitan, customer text messages and Slack.
What prompted it
The author wanted to automate as much office work as possible at a plumbing company, where Monday mornings were usually a scramble.
What Grok Bot did
- Atlas works like an office manager. The author voice-types long messages to it, and after a full day asked it for feedback on giving better instructions.
- Dana was first hired as an end-to-end dispatcher but was too slow and made mistakes, so the author fired it.
- Marc was added as a manager to oversee the plumbing company’s team of bots.
- Dana’s old job was split up by software: Gary listens for new work orders in Gmail, and Steve checks capacity in ServiceTitan and books jobs there.
- Quinn texts customers appointment options and, if they accept, passes that back to Steve to confirm the booking.
- Sarah posts in a dedicated Slack channel to keep the office staff updated, and raises issues.
Evidence and limits
- Original post checked on X (Oct 17, 2026)
- Outcome reported by the author
- Not reproduced by anyone else yet
- One bot handling all of dispatch was too much: it was slow and made mistakes. Splitting the job by software worked better.
- Atlas asked the author to keep giving long context but to end each message with concrete acceptance criteria or a definition of done.
