Knowledge Base

A knowledge base is a structured information store that AI chatbots and support teams use to quickly and accurately answer customer queries. It contains FAQs, product documentation, guides, and internal policies.

What is a Knowledge Base?

A knowledge base (KB) is a centrally maintained collection of information that your team and AI-powered systems use to answer customer queries. Instead of every agent needing to keep the same knowledge in their head, everything is stored in one place – structured, searchable, and always current.

In the context of messaging platforms like SendSeven, the knowledge base serves as the foundation for AI chatbots. The bot searches the KB using RAG (Retrieval-Augmented Generation) and formulates answers based on the stored documents. The result: precise answers instead of generic AI hallucinations.

What Belongs in a Knowledge Base?

FAQ pairs: Frequently asked questions with verified answers – the classic. Example: "How long does shipping take?" → "Standard 3–5 business days, express 1–2 business days."

Product documentation: Technical details, price lists, feature descriptions. Especially important for B2B and SaaS – customers often ask about specific functions.

Process guides: Step-by-step instructions ("How do I cancel an order?", "How do I update my payment details?"). Structured guides help both the AI and human agents.

Policies: Returns policy, terms summaries, warranty conditions. The AI can quote directly from the policy – no room for interpretation.

Internal playbooks: Escalation rules ("When to escalate to team lead?"), goodwill policies, tone recommendations. These documents control how the AI responds.

Knowledge Base and AI Chatbots

Modern AI chatbots use a technique called RAG (Retrieval-Augmented Generation): with every customer query, the system searches the knowledge base, finds relevant passages, and generates a natural-language response from them. Without a KB, the AI "hallucinates" – it invents answers that sound plausible but are wrong.

The difference:

Without KB: "Our return policy is 30 days" (invented – it's actually 14)
With KB: "According to our return conditions, you have 14 days from receipt of the goods" (from the stored document)

The quality of your knowledge base directly determines the quality of your AI chatbot. Rule: garbage in, garbage out.

Best Practices for Building a Knowledge Base

1. Clear structure: Categories and tags instead of one large text file. The bot finds relevant information faster when it is thematically grouped.

2. Currency: Outdated information is worse than no information. Set a review cycle (e.g., monthly) and mark the date of the last update.

3. Consistent style: Consistent phrasing and tone across all entries. The AI adopts the style of the KB – if that is inconsistent, the answer will be too.

4. Feedback loop: Analyze which questions the bot could not answer. These are gaps in your KB – fill them systematically.

The knowledge base itself typically contains no personal data – it stores general company knowledge. However: no customer data in the KB! Example: "Ms. Smith has special pricing" belongs in the CRM, not the KB.

Ensure the KB is hosted on EU servers. If a US-based LLM searches the KB, a data transfer occurs that must be covered by GDPR safeguards.

Knowledge Base in SendSeven

SendSeven offers an integrated knowledge base that is directly connected to the AI chatbot. Upload documents (PDF, web page URLs, FAQ lists), and the bot uses these as the basis for its answers – across all chat channels simultaneously – directly within the Unified Inbox.

Features:

Document upload: PDF, DOCX, web pages, FAQ lists
Automatic chunking and vectorization for RAG
Versioning: see when each document was last updated
Confidence score: bot indicates how certain it is about an answer
Hosted on EU servers – GDPR-compliant

Setup guide: Knowledge Base Setup Guide. Also learn how to set up an AI support bot.