Case study
Forty platform posts a week, four platforms, effectively one person.
An AI services marketplace brand needed a daily short-form presence for a C-level audience. The answer was a pipeline of agentic skills, not a content team.
225k+
lifetime views on a single post
+63k
views one post gained after its debut week
40+
platform posts per week, run by one person
Engagement: In-house build — offered to clients as a content engine
How do you run a multi-platform content engine with AI?
You encode the editorial judgment, not just the production. This engine runs on a set of agentic skills: one plans the weekly content mix, others produce the videos, carousels, and image posts, and a weekly performance report feeds what worked back into next week's rules. The AI does the volume. The rules — hook forms, platform splits, what counts as a real win — come from the data.
Who they were
An AI services marketplace brand, built and run in my role as AI enablement lead at a digital outsourcing company. Audience: C-level executives, founders, and investor-minded operators. Beat: frontier tech and robotics. Dedicated video team: none.
The problem
A brand posting daily across TikTok, Instagram, YouTube, and Facebook needs 40-plus platform posts a week. Done by hand, that's a full content team. Done with generic AI, it's forgettable slop that the algorithms bury and the audience scrolls past.
The real problem wasn't producing content. It was producing content with consistent editorial judgment — the right hook forms, the right platform behavior, the right topics — at a volume one person could never sustain manually.
What we built
- 01
Weekly mix planner
An agentic skill generates the week's content plan as a structured brief: a balanced mix of short-form videos, deep dives, news, carousels, and image explainers, with industry-vertical coverage enforced so the feed never narrows.
- 02
Editorial rules, encoded
Hook rules live inside the skills, and they're data-backed: curiosity-gap questions beat news statements, mainstream anchors beat spec dumps. When one hook tied to a famous TV show carried roughly 57% of a week's network views, that finding became a written rule the planner applies every week after.
- 03
Asset pipeline
Separate skills produce each format from the plan — sourced and clipped video, avatar segments, carousels, image posts — so every row of the weekly brief becomes a finished, platform-shaped asset.
- 04
Scheduling layer
Posts go out in platform-specific timing windows anchored to where the audience actually is, not the local clock.
- 05
Weekly performance report
An automated report separates new-post views from evergreen tail, flags single-post concentration, and feeds structural insights back into the next week's mix. The engine learns in public, one week at a time.
What changed
The brand's best post has passed 225,000 lifetime views. Another gained more than 63,000 views after its debut week — evergreen compounding the report surfaced, which turned re-posting proven winners into a standing play.
The numbers get read honestly. When one post drove most of a week's growth, the report said so instead of celebrating the total — and that discipline is why the rules keep getting sharper instead of the feed getting lucky.
The engine doesn't replace editorial judgment. It encodes it — and every week's data tightens next week's rules.
Your version of this
Every one of these started as a conversation.
Tell me what you're sitting on — expertise, archives, a content bottleneck — and I'll tell you what I'd build.
Not ready to talk yet? Start with the AI Readiness Diagnostic.