Marketing content engine

The problem

Marketing a software product is relentless. You need a steady stream of on-brand posts, fresh screenshots, the odd video, new ideas every week, and someone actually watching what performs and adjusting. Done well, that is a full-time content team. Done on the side, it is the first thing that slips. And most “AI for content” tools make the gap worse: they write generic posts in nobody’s voice, and they can’t touch your real product to show what it actually does.

What we built

A custom marketing content engine — really a small content team inside a single dashboard — that runs the whole loop while keeping a human at every approval.

• Writing across skills. Posts for LinkedIn and YouTube are written by a set of Claude Skills, one per kind of post and tone. Every skill is editable, so each round of feedback sharpens how it writes for the brand — it gets more on-voice over time instead of drifting after a week.
• Its own product media. The agent opens the actual software, takes real screenshots, and can assemble a full product video. Posts show the real thing, not stock imagery.
• Draft → approve → publish. Everything lands as a draft. You tweak it, or give feedback and let it revise, then approve — and it either posts straight away or schedules it, e.g. for the next day.
• Weekly idea briefing. Every week it proposes three post ideas to approve. You can add your own and drop them straight onto the schedule.
• Performance that feeds back. It tracks how each post does and feeds that back into what — and how — it writes next, so the output compounds instead of plateauing.

The result

In daily production, running the content for a live software product end to end with a human only at the approval step. The same build — adjustable skills, self-captured product media, draft-and-approve, weekly briefings, and a performance loop — is offered to clients who want a marketing engine without standing up a content team.