The AI operating system I run the business on

The problem

This one is not client work. There is no invoice behind it. It is the system I built for myself, to run OtivaxAI.

The business ran out of my head. Client facts, pricing logic, what we decided and why: none of it was written down anywhere a tool could reach.

Before

3 international client calls happened with nothing captured from any of them.

2 teammates had to be briefed by hand every day.

Every LinkedIn post and every outreach message started from a blank page, which meant the same thinking got redone every week.

Now

Built between September 3 and September 17, 2026, across 30 commits. It has been in use every day since.

Four parts, built in that order.

1. Context

Business facts, priorities, positioning and voice, written into files a fresh session reads before it answers anything. Alongside them sit 11 reference guides it can look things up in. No re-briefing.

2. Connections

4 live connections, wired through Python bridges I wrote. Google Sheets for the team's operations tracker. Tavily for research. Apify for lead sourcing. Modal for jobs that need to run with the laptop closed.

3. Capabilities

6 skills that each produce a finished artifact, not advice: content drafts with rendered images, weekly system audits, team reporting, decision capture.

4. Cadence

A morning command builds the day's outreach queue. A night command counts what actually went out against target. The team's numbers land in the same sheet.

What changed

31 decisions are on record with the reasoning attached, so nothing gets re-argued or quietly reversed.

A fresh session answers questions about the business without being re-briefed.

The support layer behind roughly 100 outreach actions a day is built and counted, not tracked by hand. Sending stays manual on purpose, because the platforms ban automated sending.

Content gets drafted from a written source bank instead of from nothing.