j0sh.wtf

i run my own work with 15 ai assistants

Headline "i run my own work with 15 ai assistants" next to a terminal listing assistant roles and the rule "ask first, always."

The problem

ai is easy to start and hard to trust. one chat writes an email, another fixes a spreadsheet, and pretty soon nobody knows what got done, what got sent, or what it cost.

that was me. i was doing everything myself, and the ai side of it was a mess.

What it does

15 assistants. one job each.

there's a chief of staff, sales, an editor, a designer, research, an engineering lead, a builder, a cfo, inbox, calendar, a librarian and social media. a few more cover my own life, like health, plus one that designs new bots.

on oct 26 i retired 8 older bots and renamed the rest, so every job has one owner.

Twelve work roles in three groups: running the business (chief of staff, sales lead, editor, social media), building (engineering lead, builder, designer, researcher) and keeping it running (CFO, inbox manager, calendar, librarian), plus three personal assistants. All of them share one notebook.

How it works

  1. one shared notebook. everything lives in one obsidian vault: a status page where each bot keeps its own section, a goal sheet per bot, a work board, a lessons log and a change log.
  2. written rules. ask first, always. nothing goes out as me unless i asked for that exact message. no buying, no moving money, no passwords. if a source is missing or stale, the bot stops and tells me instead of guessing.
  3. they draft. i approve. sales drafts the emails, the editor makes them sound like me, and nothing sends until i say yes.
    Four steps: the sales lead drafts an email, the editor cleans it up, Josh approves it, and only that exact message goes out. Below them is the rule: ask first, always.
  4. cheapest tool that works. simple checks run as free scripts. cheaper models do the bulk drafting and sorting. the main assistant only gets the judgment calls.
    Three tiers: free scripts handle simple checks, cheaper models do the bulk drafting and sorting, and the main assistant handles judgment calls and talking to Josh.
  5. one place for questions. when a bot needs a decision, it goes on one reminders list on my phone. nothing gets buried in a chat.
  6. a paper trail. every routine logs one line per run: when it ran, how it went, what changed.

no exact number yet. my rough estimate is about 5 hours a week saved on routine work alone: the daily picks runs, sorting my mail and watching for replies.

What it shows

none of this takes a computer science degree. it's clear jobs, written rules and checking the work. same stuff that makes a good crew run, and any owner can learn to set it up in their own business.

Want to run something like this?

that's what my monthly retainer is for. you get voxer access to me and two calls a month, and you learn to put ai to work in your business yourself. i work with owner-run service businesses across south texas.

Book a free 15-min call or email me at josh@mccoybuilt.com