Glossary
Knowledge base
In short
A knowledge base is the collection of information an AI agent answers from: your services, prices, hours, policies, FAQs and anything else customers ask about.
For an AI receptionist or chatbot, the knowledge base is what turns a general AI model into an assistant for your business in particular. It is usually built from your website, your documents and the questions your team answers every day.
Its quality sets the ceiling for every answer. Out-of-date prices or vague policies produce out-of-date, vague replies, so keeping it current, ideally whenever something in the business changes, is the most important upkeep an AI agent needs.
A useful habit: check which questions the agent could not answer, and add those answers.
A worked example
A cleaning company's chatbot keeps telling customers the old price for an end-of-tenancy clean, because nobody updated the price list it answers from. One change to the knowledge base fixes every future answer at once — which is exactly why the knowledge base, not the AI model, is where most of the upkeep goes.
When it is worth the effort, and when it isn't
It is always worth the effort for the questions customers actually ask. It is not worth writing pages about things nobody asks; start with your top 20 questions and grow the list from real conversations.
Common questions
Is my website enough?
It is a good start. Most websites leave out the details customers actually ask about, such as exceptions, edge cases and "what if" questions, so expect to add those.
How often should it be updated?
Whenever something changes — a price, an opening hour, a new service — and once a month from the questions the agent could not answer.
What format does it need to be in?
Plain, clear sentences work best. Long PDFs and marketing copy work less well than short, direct answers.