AI Chatbot with Knowledge Base — Your Bot Knows Your Business
Generic AI chatbots give generic answers. Zentor App connects the AI chatbot to your own knowledge base: the bot answers exclusively based on your company's content — precise, consistent, and without made-up information.
The Problem with Generic Chatbots
An AI chatbot that relies solely on the general training knowledge of a language model cannot reliably answer your customers' questions. The model doesn't know your company, your pricing structure, your service terms, or your current products. The result is vague, generic, or in the worst case, incorrect answers — which damages your customers' trust.
First-generation rule-based chatbots solve this problem differently: they are hard-coded and only respond to exact keywords. This leads to rigid, inflexible conversations — as soon as a customer phrases a question differently than expected, the system fails.
Zentor App combines the best of both approaches: answers are based on your own, controlled content (like the rule-based approach) — but the AI model understands the customer's natural language and can respond flexibly (like a modern AI chatbot). The result is a chatbot that truly knows who you are. We explain how this principle differs concretely from a generic bot in Custom AI Chatbot Instead of a Generic Bot.
What Is a Knowledge Base — and What Is RAG?
A knowledge base in the context of AI chatbots is a structured collection of all the content on which the bot is meant to answer questions. You upload your FAQ documents, product texts, service information, and process guides — the chatbot accesses this collection for every request.
The technical mechanism behind this is called Retrieval-Augmented Generation, or RAG for short. It works in two steps: first, the system searches your knowledge base for the text passages relevant to the customer's request (retrieval). It then passes these found passages along with the customer's request to the AI model, which formulates an understandable, context-aware answer from them (generation).
The result: the bot doesn't make anything up. It only answers what's in your knowledge base — in natural language, without a rigid script, and with an understanding of how your customers phrase things. You can read a detailed technical explanation of retrieval and generation in our article RAG Technology Explained.
Your Content — Your Answers
The Zentor App knowledge base is suitable for all text-based content relevant to customers. Everything you want to teach your chatbot, you upload via the dashboard — no programming, no technical setup. Our page Chatbot with Your Own Data.
Frequently asked questions and answers — the classic of every knowledge base. Structured as a Q&A document or as running text.
Features, benefits, variants, and use cases of your products. The bot can address individual products specifically.
Office hours (in the sense of the chatbot's service hours), delivery terms, return processes, and other service-relevant information.
Internal or external step-by-step instructions that help customers use your products or services.
Who you are, what you offer, and why customers work with you. Useful for initial inquiries and general orientation conversations.
Price lists, rate overviews, and terms — without personal data, but with all the information customers typically ask about.
Starter: 1,000 MB of Knowledge Storage
The Zentor App Starter plan includes 1,000 MB of knowledge storage (expandable on request). That may sound like just a technical figure at first — but it's worth putting it in perspective: a text document with 50,000 words typically takes up under 500 KB. With 1,000 MB, you can therefore store a very large amount of company text, FAQ documents, product descriptions, and guides without running into capacity issues.
For most small and medium-sized businesses, this storage is far more than sufficient to cover all relevant customer information. The chatbot draws on this foundation and delivers high-quality answers — regardless of how many conversations are happening at the same time.
You manage the content of the knowledge base at any time via the dashboard: add, update, or remove documents. Changes take effect immediately — the chatbot answers based on the updated content starting with the next conversation.
Vector-RAG in the Custom Package
For companies with extensive knowledge bases or higher quality requirements, Zentor App offers Vector-RAG in the Custom package. The difference from standard search: instead of searching for keywords, Vector-RAG converts both the knowledge content and the customer request into mathematical vectors. These vectors represent the semantic content — that is, the meaning — of a text.
This enables semantic similarity search: if a customer asks "How can I cancel?", the system also finds the section titled "Terminating the contract" — even though there is no shared keyword. This significantly increases retrieval accuracy, which is directly reflected in the quality of the chatbot's answers.
Vector-RAG in the Custom package can be combined with local AI models on-premises: your knowledge base and AI processing remain entirely on your own infrastructure. Maximum data sovereignty combined with semantic search quality.
Frequently Asked Questions About the Knowledge Base
What is RAG technology?+
RAG stands for Retrieval-Augmented Generation. The method combines two steps: first, the system searches your knowledge base for content relevant to the request (retrieval). The AI model then uses this found content to formulate a precise, context-aware answer (generation). The result is answers based on your own company content — not on the AI model's general training knowledge.
What content can I upload to the knowledge base?+
You can upload all text-based content relevant to your customers: FAQ documents, product descriptions, service terms, price lists (without personal data), process guides, product usage instructions, or general company information. The knowledge base can be managed via the Zentor App dashboard — no technical knowledge required.
How much storage is included?+
The Starter plan includes 1,000 MB of knowledge storage (expandable on request). That corresponds to a very large amount of text documents — for most small and medium-sized businesses, this storage is more than sufficient. In the Custom package, additional storage as well as semantic Vector-RAG search can be configured as needed.
Does the chatbot only answer from the knowledge base?+
In the standard configuration, the chatbot answers questions exclusively based on the content in your knowledge base. This prevents the bot from giving generic or incorrect answers from the AI model's general training knowledge. If no matching content is found for a request, the chatbot can be configured to hand the customer over to a team member instead of giving an uncertain answer.
What is the difference between Starter and Custom for the knowledge base?+
In the Starter plan, the chatbot uses classic keyword-based retrieval from your knowledge base: the bot searches for keywords from the customer's request. In the Custom package, Vector-RAG is additionally available: here, requests and knowledge content are converted into semantic vectors, so the bot also finds matching answers when the customer phrases a question differently than it is stored in the knowledge base — significantly improving the hit rate.
Can I automatically populate the knowledge base from external systems?+
Zentor App supports importing documents via the dashboard interface. For automated imports from external systems (CMS, document management, CRM), API access and n8n workflows are available in the Custom package, through which content can be imported and updated in a structured way — without manual uploads via the dashboard.
How quickly do changes to the knowledge base take effect?+
Changes you make in the Zentor App dashboard are processed immediately. Starting with the next conversation, the chatbot answers based on the updated content. There is no delay due to manual retraining cycles — the RAG system reads the current content directly from the knowledge base with every request.
What happens if the chatbot doesn't find a matching answer?+
If the RAG system doesn't find content in the knowledge base matching the request, the chatbot's response is configurable: it can tell the customer that it has no answer and initiate a handover to a team member, or ask a general follow-up question. The exact response for missing matches can be defined in the Zentor App dashboard — no uncertain answers from the AI's general training knowledge.
Turn Your Chatbot into an Expert on Your Business
Try the Zentor App AI chatbot with knowledge base in the demo — or find out more about the right plans for your business.
More in the Help Center