Automate Everything, Starting With Us

We tell property managers we automate the work that repeats. Here is the audit of our own company, because we did it to ourselves first and we keep doing it.

A person approves everything that goes out. The drafting, the research, and the sending around that approval are automated.

Email

Neither of us writes an email by hand. A rough draft, a few lines of what needs saying, goes to Baraq, our own AI employee, or to Claude. The agent writes the real thing, brings it back, and once one of us says yes, it sends.

Cold outreach runs the same way with one addition. Before a word is drafted, an agent researches the company and finds one specific, verifiable flaw in their public material: a 24/7 maintenance line promised on a site whose office hours are 9 to 5, Monday to Friday, or template placeholder names still live on the About page. The email leads with that, and the subject line is the flaw itself. If no reachable email can be found, the agent says so instead of guessing one.

Prospecting

Our prospecting agent works through eight Texas metros in rotation. It reads public directories and company team pages, then publishes a digest as a GitHub issue: who to talk to, why they fit, and a one-line opener for each. It checks every name against everyone it has already reported, so nobody comes up twice.

It never touches LinkedIn. No login, no scraping, no automated connecting or messaging. The agent writes the list. One of us sends every connection request by hand.

Client work

For Build or Break, a new episode page is built from the transcript and cover the team drops into a shared tracker, the episode's YouTube, Apple, and Spotify links are fetched automatically, and the page is staged through Webflow's API. The tools dry-run by default, never write to the live endpoints, and the publish step refuses if a placeholder is left in the page.

For Danly Properties, a push to the repository is live in about 30 seconds, and an audit script checks a property page for its structured data, breadcrumbs, description length, address, heading, and analytics tag, and logs what it finds.

Our own site

Before 6aiatx.com deploys, a script measures 58 text-and-background pairs for contrast, including type sitting over photographs, and fails any pair below the accessibility line. A sweep loads all 23 routes and fails on a single console error. A panel of AI judges reviews the screenshots and the code and reports what they would send back. Deploying also pre-warms every image size and submits the new sitemap to Bing and the other IndexNow engines. Nobody types any of that. An agent runs it and I read the results.

The back office

Our internal dashboard, the one with clients, prospects, money, and this month's goals, was built in two days in early September. It fills itself. A hook on my laptop reads my Claude Code sessions and extracts what changed. An email poller reads my inbox every 30 minutes and pulls out the client facts, payments, and todos. Everything lands in a review inbox where a person approves or rejects it, and a manual edit always wins over an automated one. The first real sync cost 16 cents. An email extraction run costs about 19.

Baraq takes the notes and keeps the calendar. He drafts the follow-ups. When we're weighing a new idea, we ask him first. Our LinkedIn posts and Medium articles go through the same draft, approve, send loop, and the posting itself is done by an agent driving the browser.

What stays human

The approvals, and the first call with a client. Everything else is a candidate.

The list of things we still do by hand is shorter than it was in August. Read the mission, or tell us what your team still does by hand.