The knowledge base is what your AI assistant draws on to answer customer questions. Without it, the assistant can still hold a basic conversation and capture leads — but specific answers (prices, services, policies) need your data.
OzcenkLabs supports two types of knowledge:
- Catalog items — structured entries like services, treatments, real-estate listings, vehicles, or merchandise. The sidebar label adapts to your sector (Services, Portfolio, Vehicles, or Products). Best for anything with a name, price, stock, or duration.
- Documents — free-form text like FAQs, policy PDFs, internal procedures, sector-specific information. Best for prose knowledge that doesn't fit a price-list shape.
The AI retrieves the most relevant entries from both sources for every message using semantic search — so a customer asking "Do you offer Saturday appointments?" gets the right answer even if your document says "weekend availability".
Adding catalog items
In the dashboard sidebar you'll see a section named for your sector: Services (dental, beauty), Products (retail, internal medicine), Portfolio (real estate), or Vehicles (car dealership). The form fields adapt too — a dental clinic captures treatment duration, a real estate agency captures square metres, a car dealer captures year and mileage.
You can add items three ways:
- Manual — click the Add button and fill the form. Best for <20 items.
- CSV import — download the template, fill it in a spreadsheet, upload. Best for 20+ items or migrating from another system.
- Edit — update prices and stock anytime. Changes are re-embedded automatically (you don't have to do anything).
Each item is automatically converted into a vector embedding so the AI can find it through semantic similarity. You don't need to do anything special — every create, update, and delete syncs in the background.
Adding documents
Go to AI Knowledge Base in the sidebar. You can paste text or upload files in PDF, DOCX, or TXT format.
What happens behind the scenes:
- The document is split into ~800-character chunks with a 100-character overlap (so context isn't lost at boundaries).
- Each chunk is embedded and stored in the same searchable index as your products.
- When a customer asks a question, the top matches across products and documents are passed to the AI as context.
What to upload
Good candidates:
- FAQ documents — customers will ask the same questions.
- Cancellation, refund, no-show policies.
- Insurance, payment methods, accepted documents.
- Practitioner bios, specialties, languages.
- Office directions, parking, accessibility.
- Sector-specific knowledge (treatments, materials, brands carried).
What NOT to upload
- Patient or customer personal data — the AI doesn't need it to chat and you'd be exposing it to retrieval.
- Login credentials, API keys, or anything sensitive.
- Outdated documents. The AI will quote whatever it finds. Remove old versions before uploading new ones.
How the AI uses the knowledge base
Every customer message triggers a retrieval-augmented generation (RAG) flow:
- The customer message is embedded as a vector and compared against your stored products and documents.
- The top relevant chunks are added to the AI's context window.
- The AI generates a reply using your sector's tone, your knowledge base, and the customer's conversation history.
- If something specific is requested (a price, a service the customer named), the AI quotes it directly from your data — not from general knowledge.
Tips for accurate answers
- Be specific. "Cleaning: £45, takes 30 minutes" beats "We offer cleaning".
- Use your customers' words. If they say "braces", your document should too — even if you'd call it "orthodontic treatment" in a brochure.
- Keep documents short and focused. One topic per document is easier for the AI to retrieve than a 50-page handbook.
- Update when you change. Prices, availability, hours — if your real-world reality shifts, update the knowledge base the same day.
Re-syncing products
If you bulk-imported products via CSV and the AI seems to be missing them, go to AI Knowledge Base and click the Product Sync button. This re-embeds all your products in one shot. Useful after a large import or when migrating from another system.
Trial limits
The 14-day free trial includes a 500-message quota across all your conversations. There's no fixed cap on the number of products or documents you can store — focus on quality over quantity. See Terms for paid plan details.
Next steps
- Test the AI's answers using the embedded preview.
- Iterate. The first knowledge base is rarely the final one — refine based on real conversations in the Live Chats module.