Back to Marketing

    Content Generation

    AI-assisted, human-edited. Volume without slop.

    Every channel needs more content than your team can produce by hand. AI fixes that — but only if the output sounds like you, holds up to your edit standard, and ships on a schedule. Built right, it does all three.

    What's inside

    The engine: briefs in, on-brand drafts out.

    1. 01

      Voice + style training

      Your existing best content becomes the voice anchor. I build the prompt + retrieval setup so every draft sounds like the team, not the model.

    2. 02

      Brief-to-draft pipeline

      Topic in, fully drafted post or article out — with the outline, the hook, the data points, and the internal links. Reviewer edits, doesn't rewrite.

    3. 03

      Channel-fit variants

      Same idea, shaped for LinkedIn, the blog, the newsletter, the X thread. Auto-generated, human-approved at the cut.

    4. 04

      Anti-slop quality gate

      Every draft passes a checklist: banned phrases, source verification, voice match, factual accuracy. Catches what 'just paste it' won't.

    How an engagement starts

    Train the engine. Ship for two weeks. Decide.

    1. 01

      Voice + topic intake

      I read your best 30 pieces and 30 worst. We agree on the wedge, the banned list, and the cadence.

    2. 02

      Two-week pilot

      I run the engine for two weeks, shipping the agreed cadence. Your team reviews; we tune the prompts every other day.

    3. 03

      Hand over or hosted

      If it works, keep it running as a hosted retainer, or take over with the prompts and the docs. Either ends with a working pipeline.

    30 minutes. We figure out what's worth running on AI vs by hand.

    Common questions

    Will my readers tell it's AI?
    If it's done right, no. The voice anchor + retrieval + edit gate is what kills the AI tells — most writers can't tell when the engine is humming.
    Which models?
    Claude for long-form drafting, GPT for variants and ideation, Gemini for research-grounded work. Pick the right tool, don't pretend one model wins everything.
    Does this work in Swedish?
    Yes, with care. SV content needs more voice anchoring than EN — the models are less native — but I run several Swedish content engines.

    Cookies. Privacy