Skip to content
References · From live operation

Real systems.
With numbers, not marketing.

Few, but real. One paying customer plus what runs on our own brands. Names stay anonymous, the numbers are genuine.

A content operation we ran

Two channels,
one brand voice.

Posts that reliably go out, without starting from zero each week.

Before, posts were made entirely by hand: text via ChatGPT, visuals built manually in Canva, everything copy-pasted between tools. From May to August 2026 a system we built and operated ran inside their Microsoft Teams: drafts were produced against a fixed rule set, each lawyer approved with one click, and publishing to Facebook and Instagram happened on its own. Nine of the eleven lawyers with access authored and approved posts of their own, 24 in total.

9Lawyers with own posts
2Channels · Facebook & Instagram
May–Aug 2026Operated by us throughout

Before

  • Posts made entirely by hand, every week anew.
  • Text via ChatGPT, visuals built by hand in Canva.
  • Everything copy-pasted between tools.
  • No clean approval trail.

After

  • Drafts produced against a fixed rule set.
  • Lawyers approve with one click.
  • Auto-publish to Facebook and Instagram.
  • Operated and monitored by us.
Also running

What we run on our
own systems.

DogfoodOwn D2C brand · E-commerceR&D + proofStack: Shop · Automatisierung · LLM

The same engine for product content.

We run the same engine on our own brands: raw product data in, brand-consistent descriptions per market language out, approval before anything goes live. We test what we sell on our own skin first.

Product contentMultilingualApproval gateDogfood
1Brand voice, all languages
Market languages
100%Approved before live
DogfoodOn our own brands
Next step

Which of these patterns fits
your company?

Describe in two sentences where it's stuck today. We'll tell you if we've seen a similar case, and how long a typical entry takes.