AI-generated sections

An AI-generated section is a document block that an agent writes — and keeps up to date — under your control. Instead of pasting AI output into the doc, you insert a section, let the agent generate content into it, and approve the result before it becomes part of the document.

What they look like

An AI section renders with a visible frame showing its status:

• Generated — the agent produced a draft; review is pending.

• Needs review — content changed or a regeneration completed.

• Approved — a human accepted this content.

• Error — generation failed; retry from the block.

Creating one

1. Slash menu → AI-generated section (or ask an agent to add one).

2. Trigger generation with a prompt — e.g. "Summarize our refund policy from the linked docs."

3. The agent writes a candidate replacement for the section, with sources attached.

Reviewing candidates

• The block shows the candidate side-by-side with current content.

• Approve replaces the section with the candidate content (the document as it was before approval is kept in version history).

• Reject discards the candidate; the section stays as-is.

• Regenerate re-runs generation, optionally with an edited prompt.

• Every candidate records its model, prompt, source references, and the agent run that produced it — provenance you can audit later.

How agents use them

Agents can author or refresh AI sections as part of larger runs — e.g. a docs-maintenance agent that updates "Known issues" in a release doc from recent support conversations. Each candidate lands in the same review flow, so agent output never silently edits a document.

Governance

• Approval is always human: agents cannot approve their own sections.

• All events (generate, approve, reject, regenerate) are logged with actor and agent run.

• Combined with version history, every AI edit is traceable end-to-end.