RAG knowledge base
Hotel descriptions, indications, prices, FAQs as a vector index — the AI answers with real knowledge.
Module · AI email
AI where the work happens — not as an external chatbot.
RAG-grounded drafts for every inbound email. Safe intents auto-send; everything else goes into the approval queue.
What it solves
- Repeating answers typed manually — no learning curve
- Information for the reply scattered across multiple tools
- External chatbot doesn’t know the customer — generic replies
How it works
Intent classification
Inquiry → intent (status, cancellation, new booking, complaint) → matching workflow.
Auto-send whitelist
Safe intents (status, FAQ, confirmation) auto-sent; sensitive intents go to approval.
Draft editor
Proposed text with context (customer history, booking, indication) — agent reviews, sends, done.
Learning from edits
Agent edits feed the learning layer — the next draft is closer to the desired style.
Multilingual
DE / EN / RU / ES / FR — reply in the customer’s language, detected automatically.
Inside TourCore
AI runs in an isolated worker (tc-worker-ops). Drafts are drafts, not outbound mails — they go out only on click, with an audit log over every auto-send.
Experience AI email on a demo tenant.
We send a real inquiry in, you see the RAG draft in real time.