PROPOSED PROJECT
Guest Request Dispatcher
Give hotel front desks one queue for guest requests from connected messaging channels. Classify and translate requests, suggest the right department from hotel procedures, and track staff acceptance and completion; front-desk staff approve guest replies.
Received is a start. Who has the job?
Three requests from connected channels
Original messages stay attached
Could we have two extra towels?
The air conditioning is not cooling.
Can we check out at 3 pm?
One request record.
From message to handover.
At 12:20, the fictional front desk has three requests. R-201 asks for towels and has a staff completion note at 12:16. R-202 reports in Spanish that the air conditioning is not cooling; maintenance accepted it at 12:10 and the job remains open. R-203 asks for a 15:00 checkout and is waiting for a front-desk decision. Reply drafts stay unsent until reviewed. No real guest messages or hotel operations are used.
WHAT IS USUALLY DONE
Guest requests arrive through several channels and get forwarded to department chats. The front desk may know a message was sent but not whether someone accepted the job. Translation, room checks, and replies happen separately, making shift handover harder.
WHAT SOVRUN CAN DO FOR YOU
Sovrun would build a shared request desk that keeps the original message, translation, suggested department, staff acceptance, and completion record together. AI helps interpret multilingual messages and suggest routing from the hotel’s procedures. Staff own the actual job status, and the front desk reviews replies before sending them. Requests that need a decision, such as a late checkout, stay visibly unresolved until an authorized person approves them.
Keep the original words.
Reply from the actual job status.
Translation and routing help the desk understand the request. Staff acceptance and completion are separate events. The guest reply should reflect those events, without inventing a repair time or an approval.
El aire acondicionado no enfría.The air conditioning is not cooling.
Room 318 is linked through the hotel’s verified guest record in this example. The source remains attached if staff need to correct the translation.
Suggested department:
Maintenance.
- Reason
- Cooling fault in the room
- Staff acceptance
- 12:10 / Maintenance team B
- Completion
- Not recorded
- Promised arrival
- None
“Our maintenance team has accepted your request. We’ll update you when we have more information.”
Front-desk approval requiredA completion needs evidence.
- ReceivedTwo extra towels requested
- AcceptedHousekeeping team A
- CompletedTwo towels delivered; staff note attached
“Can we check out at 3 pm?”
STAY-03 requires the front desk to check room availability and any applicable charge. The request stays open. A model-generated reply cannot approve the extension or invent a fee.
No extension approvedA desk that can see
what still needs doing.
The first build connects an agreed set of guest channels to the hotel’s department procedures. The useful outputs are a shared queue, replies ready for review, and an accurate handover.
Accepted. Still open.
- Received
- 12:07
- Accepted
- 12:10 / team B
- Completed
- No completion event
- Next step
- Staff progress update
Say what is known.
“Our maintenance team has accepted your request.”
No invented arrival time or repair promise.
Approval pending · unsentCarry the work forward.
Completion still pending.
Front-desk decision needed.
How we’d make it happen.
Connect messages to the hotel’s rules
Agree on channel access, identity checks, request categories, and staff permissions. Use permitted sample messages to test translation and department suggestions.
Keep the job with its owner
Create a request record with the original message and suggested route. Staff accept work and update progress; unresolved assignments remain visible to the front desk.
Review replies and hand over clearly
Prepare guest replies from the actual job status for front-desk approval. Carry unfinished work, owners, and pending decisions into the next shift without marking it complete automatically.
Where AI helps
We evaluate multilingual models on the hotel’s approved languages and common request types, including ambiguous wording and messages containing several tasks. Models draft translations, split requests when needed, and suggest departments with the relevant procedure attached. Connected channel events and staff actions determine delivery, acceptance, and completion; a generated reply never counts as evidence that a job is done.
What people approve
Front-desk staff verify the guest and room through the hotel’s existing process, correct translations, and approve guest replies. Department staff accept and complete jobs with notes, and supervisors handle exceptions or requests outside the rules. The first build uses only authorized channel connections and agreed staff roles; it does not invent channel access, promise arrival times, or approve charges and extensions.
What we’d need to start
- Authorized guest messaging channels, example requests, and supported language requirements
- The hotel’s department procedures, escalation contacts, room-verification process, and reply templates
- Staff roles, status definitions, acceptance rules, and the shift-handover format
How we’d test it
Test mistranslations, duplicate channel events, mixed requests, unverified room references, unavailable staff, and premature completion messages. Confirm that forwarded, accepted, and completed remain distinct states. Verify that no reply is sent without the required approval and that an unaccepted job cannot disappear from handover. Compare missed requests and staff follow-up effort in a measured hotel trial.
Prepared for Travel & hospitality. We would agree on the work after reviewing your brief, materials, and requirements.
This is a proposed project, not completed client work. The visuals and numbers show example outputs, not measured results.
