CASE LOG 01

Case study · Autonomous AI system

Content OS

An autonomous content engine that audits six weeks of posting history, finds the gaps, writes five complete scripts, and delivers them every Monday morning — without being asked.

Overview

Content as a system, not a weekly task.

Most content workflows have the same problem: the thinking is manual. You sit down, try to remember what you posted last week, figure out what's missing, decide what to make next, and then write it. It eats hours before a single second has been recorded.

Content OS removes that work entirely. Every Monday it runs a full audit — reading six weeks of posting history, identifying gaps in topic coverage and funnel balance, generating ideas to fill them, selecting the best five for the week, writing a complete script for each, and delivering them as a ready-to-act brief. Two decisions are left to you: approve the brief, record the videos.

It was built in about a day once the architecture was defined, and it runs on its own from there.

Type Autonomous content engine

Build time ~1 day

What it does Audit · Plan · Write · Deliver

Built with

Claude AI Google Sheets iOS Shortcuts

THE DIAGNOSIS

The challenge

Turning content from a task into a system.

"Content isn't hard because of writing — it's hard because of decisions."

The core problem wasn't generating ideas. It was managing the whole lifecycle: ideas scattered across notes, chats and memory; coverage that drifted out of balance across topics and funnel stages; planning that demanded constant manual effort; scripts written inconsistently and often last-minute.

The harder problem was automating judgment — deciding what to make, what's missing, what needs rebalancing. That requires a system that understands context, not one that just executes steps.

What was built

A decision engine that runs the loop end to end.

autorenew

Dual-mode capture

Reactive mode files the ideas you dump in. Proactive mode generates new ones based on historical gaps. Both feed the same planning engine — flexibility without losing consistency.

event_available

Weekly autonomous engine

Every Monday, unprompted: audits six weeks of history, finds gaps in pillars and funnel stages, generates ideas, selects the best five, writes full scripts, and emails a clean brief.

bolt

Frictionless idea capture

An iOS Shortcut on the home screen. Dump ideas by voice or text; each one is extracted, tagged by pillar and funnel stage, and filed in the ideas bank automatically.

calendar_view_week

Dynamic content calendar

A rolling six-week calendar: this week confirmed, the next planned, the rest roughed in. Mark a video posted and the system logs it, retires the idea, and rebalances.

mail

End-to-end delivery

Each cycle delivers five fully written scripts — hook, body, CTA, format notes and hashtags — sent as a single email brief. Open it, approve it, record. Nothing else required.

psychology

Judgment, automated

Memory, logic and generation in one loop: history informs the audit, the audit drives prioritisation, prioritisation produces the scripts. A self-sustaining content cycle.

Results

What changed.

5Scripts every Monday

~6 hrs/wkPlanning → 0

~1 dayTotal build time

0Manual planning required

System insight

Content OS isn't a content tool — it's a decision engine.

scattered idea capture→a structured input system

manual planning→automated prioritisation

inconsistent output→a repeatable workflow

By combining memory (history), logic (audit and rebalancing) and generation (scripts), the system creates a self-sustaining content loop that runs without supervision.

Next case study

AI content evaluation

SalesScore

An AI evaluation tool that scores every submission against a fixed rubric and returns feedback automatically — none of the cost or drift of a human reviewer.

Read case study →

Got a
weekly
grind?

If a repeatable task is quietly eating your week, it might be a custom build. Take the 2-minute diagnostic and get the honest read — sometimes the answer is "not yet," and I'll tell you that too.

Start the diagnostic → Just email me