Knowledge base vs help center
A knowledge base is a structured store of articles and facts. A help center is a public-facing website built on top of a knowledge base. The knowledge base is the data; the help center is one of several user interfaces over that data. AI grounding is another.
Knowledge base
A knowledge base (KB) is the corpus. It is the underlying set of documents, FAQs, policies, product specs, and how-to guides. It can be expressed as Markdown files in a repo, articles in a CMS, pages on a website, or chunks in a vector database. The exact storage format is a detail. The shape matters.
In message.com the KB is a Postgres table holding crawled HTML chunks, each with a vector embedding and metadata about the source URL, the page title, and the section heading. See AI knowledge base.
Help center
A help center is a public website that lets customers search and read articles before contacting support. It has navigation, a search bar, a featured-articles section, and usually a contact form at the bottom. Zendesk Guide, Intercom Articles, and Help Scout Docs are all help-center products.
Help centers serve a different audience than knowledge bases. The KB is consumed by software (AI agents, search indexes). The help center is consumed by humans. Both can be backed by the same corpus.
The AI grounding angle
In a modern stack, the AI agent grounding pipeline is another consumer of the same KB. When a customer asks a question in chat, the AI retrieves relevant chunks from the KB and generates a grounded reply. See LLM-grounded RAG.
This means a single high-quality knowledge base powers three deflection paths: humans searching the help center, AI answering in chat, and AI answering on the phone. message.com is engineered for all three to share the same corpus.
message.com's approach
Most customers do not have a formal KB when they sign up. They have a marketing site, a blog, and maybe a few FAQ pages. message.com bootstraps a KB by crawling those sources automatically. The customer never writes "KB articles". The pages they already have become the AI's reference material.
Once a customer has scale, they can author dedicated articles inside message.com and (when the help-center product ships) publish them as a customer-facing site without re-entering the content.