What does a knowledge base contain?

A support knowledge base usually holds how-to articles, troubleshooting steps, FAQs, reference pages such as settings and limits, and policies such as refunds. An external knowledge base is published for customers, often as a help centre. An internal one is for your team and can include workarounds, known issues and procedures customers shouldn’t see.

Articles work best when each one covers a single task or question and puts the answer first. An article called “Invite a teammate” that lists the steps, the plan it needs and what the invitee receives will match more questions than a long “Team settings” page that mentions invitations halfway down.

Many support teams build their knowledge base as they work. Knowledge-Centered Service (KCS), a method published by the Consortium for Service Innovation, has support staff capture what they learn while solving an issue, structure it in a consistent format, reuse it for the next customer and improve it each time it’s used.

Why does a knowledge base matter for an AI voice agent?

An AI agent answers from what it can find, so the knowledge base sets the limit on its answers. Many agents use retrieval: they search the knowledge base for passages that match the question, then answer from those passages. If an article is missing, out of date or written in words customers don’t use, the agent misses it or repeats the old answer.

Voice adds a constraint. A caller can’t see a screenshot or a table, so write steps in words, in order, with the names of buttons and settings spelled out.

How does Kawl Now use a knowledge base?

In Kawl Now, each agent has its own knowledge. The support agent uses FAQs, guides and notes you write, plus selected .md, .mdx or .txt files imported from GitHub. The sales agent uses only sales notes and never sees support sources. A workspace holds up to 100 sources, each with up to 50,000 characters.

Nothing reaches callers until someone approves it. Adding a source never approves it, and every GitHub re-sync needs approval again. During a call, the agent searches its approved sources by keyword and reads up to four matching excerpts of about 4,500 characters each, so short articles in your customers’ own words are easier for it to find.

There are no file uploads or help desk connectors, and website import is switched off. To use help centre articles, paste their text into sources, or keep them as Markdown in GitHub and import the files.

Common questions

What is the difference between a knowledge base and a help centre?

A knowledge base is the collection of articles. A help centre is the customer-facing website that presents them, usually with search, categories and ways to contact your team. The terms are often used interchangeably because most help centres are built around a public knowledge base.

Should an AI agent use your internal knowledge base?

Only the parts you’re happy for customers to hear. An AI agent can repeat anything in the sources it’s given, so keep internal notes, workarounds and account procedures out of its knowledge. In Kawl Now, nothing is used until your team approves it, which gives you a point to check each source.