Home / Work / Marketing operations
Multi agent content pipeline with a QA gate and human approval
Topics in, four agents per topic (research, outline, draft, QA), a deterministic quality gate, human approval, then publish as drafts with the token cost accounted per run.
The problem it solves
Content automation fails in two ways: it publishes rubbish, or it needs so much editing that nobody saves any time. The fix is a gate in code, not in a prompt, and a human who approves rather than rewrites.
What it does
- Webhook or weekly schedule with a list of topics.
- Per topic: research (search API), outline, draft, then a QA gate in code: minimum length, required sections, a sources section with two or more links, banned marketing claims.
- Passing drafts go to human approval (Slack) and are published as drafts to Google Docs or the CMS; failing drafts go back with the reasons, never published.
- Token usage and cost are accounted per run.
How it was tested
- Three topics researched, drafted, checked and published as drafts in one run; a fourth failing the gate returned with reasons.
Screens
Who it is for
Firms that publish weekly and want the boring checks automated, not the judgement.
See it running on a call → hamza.jamal@finamatik.com
Built and tested on sample data before any live credential is connected. Every system Finamatik ships is adapted to the client's own tools and owned by the client outright.