The docs nobody maintains, written and kept current by the AI.
This replaces the HelpScout or Intercom docs that drift the week after launch. Your Knowledge Base has two faces: public articles your customers and search engines read, and an internal store only the AI reads when it answers questions and drafts content. The AI writes both from real questions and flags what has gone stale.
One base, two faces
The same source of truth the chat widget, support replies, and generated content all draw from.
The public Knowledge Base
Hosted under your domain, full-text and semantic search, clean brand-consistent design, and SEO built in — sitemap, meta tags, and schema. Related articles show in context.
The internal store
Non-public articles only the AI sees: product internals, policies, pricing rationale, support playbooks. Same ingestion as public articles, scoped privately, never exposed.
Articles the AI drafts and keeps honest
Articles carry rich content — images, code blocks, tables — with version history and rollback, authoring metadata, and AI-managed links to related articles. The AI drafts them from common questions or your product docs. Titles stay plain: "How to send your first email," not "A Comprehensive Guide to Email."
Categories, tags, and order
Articles live in hierarchical categories, with tagging, featured articles, and custom ordering. Readers find things through clean navigation; the AI finds things through the same structure when it answers a question or links an article from a blog post.
How a support pattern becomes an article
When product changes ship, the AI scans existing articles for language that has gone stale and proposes updates, one at a time, for you to accept or reject.
The same question, three weeks running
A customer asks the same thing in chat week after week. The AI sees the pattern.
A draft article answering it
The AI drafts an article that answers the question and brings it to you.
It publishes
You read the draft and accept it. The article goes live under your domain.
The next visitor gets a direct answer
Next time the question comes in, the chat widget answers from the new article.
The base every surface reads
The Knowledge Base is the corpus the chat widget answers from, the reading emails link to, and the content that renders inside your site. Write once; used everywhere.
Where the edges are
This is a Knowledge Base, not a helpdesk. There is no ticketing system with SLAs and queues — for that, Zendesk or Intercom Fin fits. There is no in-app tooltip or walkthrough widget in the Appcues or Pendo style; the chat widget is the in-app surface. There are no forums or user-to-user Q&A.
Basic multi-user editing works, but deep editorial workflow with multiple approvers is simpler than an enterprise CMS. And there is no native video hosting — embed YouTube, Vimeo, or Loom.
Let the AI write the article you keep meaning to
Ask for early access. Point the AI at a question your customers keep asking, and read the article it drafts back. Approve it, and the chat widget answers from it next time.