I run marketing on an AI operating system. Here's how it works.

What the system took off my desk

Each line is a job I used to do by hand. The public version of every system is on the OS page, with company data removed.

1. Reading every source by hand

   Research agent

   Plans the question, fetches every source, refuses stats it can't quote, and fact-checks before it saves a report.

   Research agent files
2. Re-explaining context in every chat

   Knowledge base

   460+ linked pages, one fact per page, each with a source. The agents read it before they answer.

   Knowledge base files
3. Putting together the Monday report

   Weekly routines

   The marketing report, a Saturday knowledge scan and a monthly reconcile run on a schedule and post a short summary.

   Routine specs
4. Guessing if a draft sounds generic

   Originality check

   Scores how templated a draft reads, so I can tell when it sounds like the default AI answer.

   Originality check files
5. Digging through a competitor's site

   Marketing OSINT

   A dossier on any company from public data: positioning, channels, content, what customers and AI assistants say.

   OSINT files

Confirm each system is your own build before publishing; installed third-party packs must not appear here. "Used to do by hand" is a claim about your own work: keep only the lines that are true.

1. DatePost titleAI in marketing
2. DatePost titleStrategy
3. DatePost titleManagement