Where AI is used
- The B2C bot is an AI assistant. It understands what you write or say, decides which travel tools to use, and composes answers. Its landing page and its replies say so. Its search form is the exception for flights: a flight search entered there, including a flexible-date search, is sent straight to the travel systems and the reply is assembled from the results by fixed rules, not by a model. Hotel searches from the form are passed to the assistant like a typed request. The form reads what you have already typed to fill itself in; that happens in your browser, and you confirm every field before searching.
- The email desks use AI to classify incoming mail by intent, to draft replies, and to pick out rates and dates from supplier emails as suggestions.
- Voice: speech is transcribed by a speech-to-text model and spoken answers are produced by a text-to-speech model.
- “Ask JetNova” on this website is an AI sales assistant for agencies evaluating the products. It answers from the same facts as these pages, can open a WhatsApp chat with our team with a summary of what you asked, and, when you give your name and a way to reach you and agree to it, records a call-back request for a person. It does not book, quote a binding price, or sign anything; every plan is confirmed in writing by a person.
- The page overview videos on this website are generated by an AI tool from the published text of the page they sit on, with a synthetic narrator. They are explanatory only; where a video and these pages disagree, the pages govern.
- Trip suggestions such as the travel-style quiz, inspiration cards and smart suggestions are produced by the same assistant.
Buraq B2B's search, pricing, wallet, ledger, bills and reports, and flight searches from the B2C bot's search form, are not AI: they are deterministic rules and records. The price check in the console runs the exact engine used at search time.
What the assistant may do
Only what the agency has switched on: search and quote; explain fare rules; show seat maps; check visa requirements; look up hotels, packages and services; keep your trips; and, when booking is enabled, take you through the booking steps. Sensitive actions ask you to confirm. If it cannot help, it offers a person.
What it never does
- Issue a ticket, take a payment, cancel or refund by email, in any desk.
- Book or ticket when the agency has not enabled it; a reservation made in that state is clearly marked as a demo.
- Contact anyone else, request your contact details or search the web when answering an email, so an instruction hidden in an email cannot make it act on it.
- Answer automated senders, mailing lists or messages that fail sender checks.
- Pretend to be a person when asked whether it is one.
People in the loop
Agency staff can read conversations, take over a chat, review and edit drafted email replies before they go out, approve or reject an agent's hold request, and see every tool the assistant used. A review queue with a service-level timer holds anything the agency's policy marks for approval. Handoff to a person is always available in the B2C bot.
Limits
The assistant can misunderstand, and it can be confidently wrong about things outside the live data it retrieved. Fares, availability and rules it quotes come from live systems at that moment and can change; the booking review step shows the final figures. Visa information is guidance, not legal advice; check with the consulate. Check every confirmation you receive.
Your data and the models
To answer you, the relevant parts of your conversation and the results of tool calls are sent to a model through a routing provider (a flight search from the search form is not sent to a model); the website assistant sends your messages to its model provider directly; voice is sent to speech providers. These are listed on the sub-processors page. We do not train models on your conversations. We select providers whose terms exclude training on data sent through their APIs.
Transparency to agencies
Agencies can see which capabilities are on, which tools are assigned, the persona and instructions in force, and the chat audit of what was said and done, without the model's internal reasoning.
Tell us when it goes wrong
Use the handoff in the chat, or write to contact@jetnova-ai.com with the conversation and what you expected. Reports change how the assistant is instructed.