
- GUIDELINESBrand guide, positioning, approved claims and banned terms go into Knowledge.
- DRAFTThe assistant writes the draft and cites the document behind each guideline.source cited
- CLAIMSData or claims outside the approved list are marked as data to confirm.nothing made up
- EVALUATIONStudio evaluates every agent version against the brand criteria.blocks a worse version
- REVIEWThe draft goes to the team for review through an n8n flow.awaiting approval
- PUBLISHINGThe team publishes: the agent does not publish to any channel.human decision
How it works
During setup, the brand guide, positioning and the list of approved claims and banned terms go into Knowledge. The assistant writes drafts of posts, emails and page copy from them and says which document each guideline came from. It does not invent numbers, research, testimonials, customers or certifications: a results claim goes in only if it is on the approved claims list, and anything missing is marked as data to confirm. A banned term is swapped for the alternative in the base, with a note. Everything it produces is a draft: sending it for review is an n8n flow that runs only after a person approves, and publishing is the team's decision. Studio evaluates every agent version against the brand criteria, written in plain language, and blocks publishing a worse agent version.



