Knowledge base (for AI email)

A knowledge base for AI email is a curated store of an organization's reference material — pricing, product facts, policies, FAQs — that an AI email assistant retrieves from when drafting replies. It grounds generated answers in accurate, organization-specific information rather than the model's general training alone.

What a knowledge base contains

A knowledge base for email typically holds the reference facts a person or team relies on to answer messages: price lists, product specifications, delivery and return policies, opening hours, onboarding steps, and answers to frequently asked questions. The material is usually broken into small passages and indexed, so that the pieces relevant to a specific incoming message can be found and pulled in on demand, rather than the whole document being read every time.

Why it matters for AI-drafted email

A general-purpose language model knows how to write, but it does not know your prices, your policies, or your product details. Left to its training alone, it can only guess at those specifics — and a confident guess in a customer reply is worse than none. A knowledge base closes that gap through retrieval-augmented generation (RAG): the assistant looks up the relevant facts first, then drafts an answer grounded in them. The result is replies that are accurate, specific, and consistent no matter who — or which inbox — sends them.

How it works in DraftKite

DraftKite includes an optional knowledge base. You add documents and notes — pricing, FAQs, policy text — and DraftKite splits them into passages and stores them as embeddings. When a new email arrives, the pipeline retrieves the most relevant passages and attaches them as context before it drafts a reply, so the draft reflects your own facts instead of a generic answer. As always, nothing is sent automatically: the grounded draft waits in your drafts folder until you review it and click send. The knowledge base is optional and, like the rest of DraftKite, hosted in the EU.

Frequently asked questions

It is a curated collection of your reference material — pricing, product facts, policies, and FAQs — that the assistant retrieves from when drafting a reply. It grounds the generated answer in your real, organization-specific information instead of the model's general training.

A knowledge base supplies the facts a reply should contain; tone of voice governs how those facts are phrased. An AI assistant can use both at once — retrieving the right information from the knowledge base and expressing it in your learned writing style.