Nine languages, one answer
English, Turkish, German, Russian, French, Dutch, Spanish, Arabic, Persian — detected automatically, and the same hotel fact in each.
Whispr AI
Whispr AI answers from what your hotel actually knows, in the guest's language, at any hour — the same correct answer on the tenth night of the season as on the first.

Requests
When a message is a service request rather than a question, Whispr opens a case, routes it to the right department and alerts the team — then tells the guest it's in hand, and tells them again when it's done.
Illustrative interface representation with sample data.
Guest · WhatsApp · English
Understood: request · maintenance · room 412Case opened and routed to the maintenance team
Whispr · instant reply
Case closed by staffThe guest is informed once the work is done
When it isn't sure
Every decision — intent, department, urgency, tone — has a threshold. Below it, Whispr asks a clarifying question or hands the conversation to your team with context. Reliability comes from grounding and human control, not from a percentage.
Guest message · WhatsApp · English
Understood — each decision carries its own confidence
Tick = the decision's threshold. Below it, Whispr does not act on that decision. Bars are illustrative — they show the mechanism, not a measured accuracy.
Confidence gate
Act(path taken)
Case opened · routed to Housekeeping · staff alerted on Telegram · guest told it's in hand
Answer
Reply grounded in hotel knowledge — or say it doesn't know, and log the gap
Below threshold
Ask a clarifying question or hand the conversation to staff — never guess
Always on
Information handling
Grounded answers only, thresholds on every decision, staff takeover that pauses the model, and a freshness cutoff that sends late messages to people — built into the product, not written into a policy.
What Whispr AI works from
The guest's message
Text, or a transcribed WhatsApp voice note
Your hotel's knowledge
Documents you upload + synced restaurants, menus, spa, room types, events
Stay context
Who is writing and their stay, where the PMS is connected — kept inside Whispr except during a reservation
Built-in controls
Grounded answers only
An answer must come from retrieved hotel knowledge; otherwise Whispr says it doesn't know and logs the gap
Confidence thresholds
Each decision has an abstention threshold; below it, Whispr asks or hands over
Human ownership wins
Staff takeover pauses the model; safety topics and "human please" hand over immediately
Freshness cutoff
Late messages are never answered automatically
What can happen
A reply
In the guest's language
A tracked case
Routed to a department, visible to staff
A reservation
Only after the PMS accepts it
A handover
To your team, with context
Classification, embeddings and ranking run on models inside Whispr. Answering, reservation details, translation and voice transcription use OpenAI's API: the guest's messages as written, recent conversation, retrieved hotel passages, hotel name and local time — and, during a reservation, the guest's name and room from their stay record. Details on the Security & Data page.
Illustrative interface representation — conversations, records and values are sample scenarios, not real guest data.
Plainly
English, Turkish, German, Russian, French, Dutch, Spanish, Arabic, Persian — detected automatically, and the same hotel fact in each.
One click makes a conversation human-owned and the AI waits. Safety, medical, or "get me a person" hand over immediately, with a staff alert.
Classification and knowledge search run on models inside Whispr. Answering, reservation details, translation and voice transcription use OpenAI's API; the application and your data are hosted in Türkiye. Exactly what is sent is on the Security & Data page.
Exactly what leaves Whispr, what stays, and what is still being documented.
Security & Data →Bring the messages your front desk answers every day — we'll show you how Whispr understands, answers, and routes them.