A Content Engine for LinkedIn and Short-Form Video

A content and UGC pipeline covering research, scripted and written generation in a defined voice, creative variants for short-form video, scheduled publishing, and engagement collected back into the CRM.

These examples describe workflow and implementation experience. They do not imply independent product ownership or identify a client.

The challenge

Consistency, not creativity, is what content programmes die of. The work is spread across tools that do not know about each other, so a gap in one stage silently stops the whole thing, and nobody notices until the feed has been quiet for three weeks.

The approach

01
Separated research from writing, so a post starts from a specific input rather than from a blank prompt.
02
Conditioned generation on a voice profile built from existing posts, because the failure mode of AI content is sounding like AI content.
03
Produced variants rather than one output: scripts, hooks, and creative options for short-form video, so a campaign has something to test.
04
Put publishing on a queue with retries, since a platform API refusing a post at 9am should not cost a day.
05
Fed engagement back, so the people interacting become records rather than notifications nobody opens.

Architecture

Primary pathConditional

n8nOpenAIVoice and video generationScheduling APIsCRM API

What was hard

Voice is the part that fails quietly. Generated copy that is technically fine but does not sound like the person is worse than nothing, because it costs credibility rather than time.
Publishing APIs refuse posts for reasons that are not always visible. The queue needs retries and a way to see what did not go out, or a silent failure becomes a three-week gap.
Attribution across platforms is approximate. Engagement collected back into the CRM is useful as a signal, and pretending it is precise would be the wrong claim to make.

Curious how this was built?

The full experience and skill set behind it are on the resume.

View resume