How it works
One pipeline behind every door: author, render, verify — and only then deliver.
The pipeline
Work enters DocMark through whichever door fits you — the web app, the REST API, your own AI assistant over MCP, or a watched folder on your own network. From there, every job follows the same path.
Author. Your people write the content, or — where you allow it — AI does: outlines, drafts, charts, even cited web research. You choose per deployment option how much AI is involved, from none at all to fully agentic authoring.
Render and brand. This step is never AI. Deterministic, tested code turns the authored content into your deliverable, applying your committed brand specification — colors, fonts, logos, layouts. Pixels never come from a model, so the output is reproducible and on-brand by construction.
Verify. Before anything is delivered, the visual-QA gate renders every page and pixel-checks it: logo placement, color compliance, text overflow, layout breakage. This is the step most document generators skip.
When the gate says no
A deliverable that fails verification is withheld — you never receive it. Instead you (or the AI assistant that submitted the job) receive precise findings: what failed, where, and why. An AI caller can fix the content and resubmit in the same session; the loop continues until the deliverable passes or the caller stops.
What you get with every deliverable
- The verified document itself (deck, report, model, web page, or video).
- A confidentiality label — its egress class, written by the pipeline, not by hand.
- A provenance record — exactly which version of your brand specification produced it.