Skip to content

Knowledge base

Adrià Castany 3 min

What is a knowledge base?

A knowledge base is the organized set of articles documenting how a product works and how to resolve what gets asked often. It serves two audiences at once. Customers looking up the answer themselves, and the team using it as the source for their replies.

Also called

support documentation, docs

Topic
Support concepts
On this page
  1. Knowledge base and help center
  2. Why it matters
  3. Types of knowledge base
  4. What articles it includes
  5. How to build one
  6. How it is measured
  7. Title it the way the customer would
  8. What a knowledge base cannot answer
  9. Maintaining it
  10. What to document first
  11. How Intake handles it
  12. Frequently asked questions

Knowledge base and help center

The knowledge base is the content; the help center is the site where it is published, with its search and its navigation.

They get used interchangeably and it rarely matters, except when comparing tools. Some sell the first and some the second, and the price is not the same.

Why it matters

It is the only support investment that pays for itself over time, because a well-written article answers an unlimited number of times.

It is also the foundation any automated agent stands on. An AI system cannot answer better than what is written, so documentation quality sets the ceiling.

Types of knowledge base

There are two, and since AI agents started answering both have a third reader, the system that searches them to write the reply.

That is why the internal one also has to make sense without context.

  • External. The one customers read, published in a help center.
  • Internal. The team’s own, with procedures, special cases and what should not be published.

What articles it includes

Four kinds of article cover almost everything.

For example, “How to change the account email”, “The verification email is not arriving” or “What happens to my data if I cancel”.

  • Guides on how something is done, step by step.
  • Fixes for known problems, with the error message as the customer sees it.
  • Frequently asked questions about billing, account and policies.
  • Reference, meaning the limits of each plan, the settings and the integrations.

How to build one

Start from the conversations, not from the table of contents.

  • Pull the questions that repeated most over the last few months.
  • Write one article per question, in the customer’s words.
  • Group them into a few categories.
  • Publish it with search.
  • Review each article when the part of the product it covers changes.

How it is measured

With searches that return nothing, which say what is missing.

With the rating on each article, which says which one does not help even though it exists. And with the conversations that arrive right after reading an article, which point to the one that gets read and does not resolve. The number of articles measures nothing.

Title it the way the customer would

If you call it "retention policy" and the customer types "how long do you keep my data", the article has to be called the second thing.

It is the most common mistake and the most expensive. The article exists, it is correct, and nobody finds it.

What a knowledge base cannot answer

Anything that depends on the specific account.

"How many seats do I have left?" cannot be answered with an article, because it changes depending on who is asking. Writing more articles does not reduce those questions, and confusing the two types is what makes a documentation project disappoint.

Maintaining it

An out-of-date article is worse than none, because the customer acts on what it says.

Tying review to product changes is the only thing that works over the medium term; calendar reminders do not.

What to document first

Ordered by frequency times cost, not by frequency alone.

A question that comes up five times a month and ends at engineering costs more than one that comes up fifty times and is answered in two minutes.

How Intake handles it

Frequently asked questions

What types of knowledge base are there?

Two. The external one, which customers read in the help center, and the internal one, which the team uses for procedures and cases that are not published. An AI agent can read both to answer.

How do you know whether a knowledge base works?

By the searches with no results, the article ratings and the conversations that arrive after reading one. If the questions that repeat most still reach the team, the article is missing or cannot be found.

Are a knowledge base and a help center the same thing?

Almost. The knowledge base is the content and the help center is the public site where it is published. In day-to-day conversation the two are used interchangeably.

How many articles do you need?

Enough to cover the questions that repeat most. Twenty articles that answer most of the volume are worth more than two hundred nobody can find.

Can a knowledge base be used to train an AI?

It is not trained on. It is consulted. The system retrieves the relevant passages at answer time, which is the difference between a fact you can update today and one frozen into a training run.

Glossary

Keep reading

Glossary

Related terms

See the full glossary

Add it without dropping what you use

15 minutes, no commitment