Every unanswered question becomes a drafted doc, waiting for your review
BeforeQuery watches which questions the assistant can't answer well, ranks them by volume, drafts the missing page in your docs style, and hands the draft to your team. AI proposes; humans review and publish. It's the loop that turns a chatbot from a support tool into an editorial system for your docs.
How Doc Proposals work
Gap analytics identify the missing knowledge; a scan drafts the missing page; your team accepts or dismisses each draft
Ranked by real question volume
Gap Discovery
- Every unanswered or low-confidence question is logged
- Gap report ranks missing knowledge by volume and confidence delta
- Groups similar questions so you don't chase 40 versions of the same topic
- Per-knowledge-base view so DevRel and Ops can each own their gap list
- Public API for gaps means you can pipe them to Linear / Jira / Notion
A starting point, not the final answer
AI-Drafted Pages
- POST to /doc-proposals/scan and the top gaps become drafted pages
- Drafts follow your knowledge base's persona and tone settings
- Each draft carries the questions it would answer and the sources it drew from
- Markdown output — drop into Mintlify, Docusaurus, GitBook, or a Notion page as-is
- Confidence-scored so the team reviews the safest drafts first
AI proposes, humans publish
Human Review
- Accept or dismiss each draft from the dashboard
- Edit in-place before accepting — the AI is a first draft, not final copy
- Accepted drafts export to markdown for your docs pipeline
- Dismissed drafts don't come back — the AI learns which topics you own manually
- Re-run scans as your gap list evolves after each docs release
Close the loop the industry doesn't close
The docs backlog writes itself
Most AI chatbots surface unanswered questions and leave it there — 'here's a report, good luck writing the docs'. Doc Proposals goes further: it drafts the page. Every ticket the AI couldn't resolve becomes evidence of a missing doc, evidence becomes a draft, and the draft goes to a human editor for review. Over a quarter, the flywheel is what turns 'we shipped an AI' into 'our docs are measurably better'.
- Unanswered questions become the docs backlog — ranked by volume
- AI drafts the page in your existing tone and style
- Human review keeps final judgement with the editor, not the model
- Accepted drafts export as markdown for any docs platform
- Re-runs after each docs update track whether the gap actually closed
What teams say
Deployed in production, cited by the buyers who chose it
“Deployed on our docs site in an afternoon. Every answer has citations, and the abstention gate means we've never had a customer complain about a made-up answer.”
PN
Priya Nair
Head of Customer Support · Supabase
“The knowledge base connected to our Slack, Confluence, and helpdesk in one setup. On-call teams get the same cited answer whether they ask in chat, in the widget, or from Cursor.”
TR
Tom Richter
IT Operations Manager · Grafana Labs
“The gap analytics turned into a real docs backlog. Deflection went up because we finally knew which pages were missing — the AI told us.”
AC
Ana Castillo
VP of Customer Experience · Clerk
Frequently Asked Questions
Common questions about Doc Proposals
Volume × confidence delta. High-volume topics where the assistant repeatedly abstains or gives low-confidence answers rank first. You can also trigger a scan for specific gaps or specific knowledge bases; the ranking is a default, not a rule.
Good starting points, not finished pages. Drafts follow your knowledge base's persona and tone settings, cite the sources they drew from, and flag their own confidence — the editor accepts, edits, or dismisses. Teams typically publish 40–60% of drafts with light editing, dismiss 20–30% as 'we don't want to document this', and rewrite the rest.
Drafts export as markdown. Drop them into your Mintlify / Docusaurus / GitBook / ReadMe / VitePress source repo, run your normal PR review, and publish. Alternatively pipe accepted drafts to Notion / Confluence / a Google Doc for editorial review before they hit the repo.
Once you publish the page and BeforeQuery re-indexes your docs source, yes — the new page becomes retrievable and citable like any other. The proposal → draft → publish → re-index loop typically closes the gap within one sync cycle.
Deliberately no. The whole point is a human in the loop for editorial judgement. What you can automate is the export — accepted drafts can flow via webhook to your docs pipeline, to a Notion database, or to a Linear ticket for the docs team to pick up.
Two signals. First, the abstention rate on previously-gapped topics — does the assistant now answer the questions it couldn't? Second, ticket-deflection rate on the topics that got new pages. Both trend in the analytics dashboard.
Turn unanswered questions into a docs backlog
Run your first Doc Proposals scan and see which gaps come back as drafts ready for your editor.
Get Started Free