Stay Request
Expected arrival 23:30; the request starts pending.
How Tamaga works / Host Desk
Turn unresolved guest requests into work someone owns.
A small hotel does not need another screen full of activity. It needs to know what still requires a person. Host Desk turns unresolved guest conditions into visible, accountable work before arrival.
Governed promise
Later arrival after 20:00 requires written confirmation.
Request23:30 · Pending
AttentionOpen · needs action
OwnerResponsible host
Request state ≠ attention state
The physical scene is unresolved work before arrival. It does not show a completed check-in or a confirmed guest outcome.
Late arrival
Confirmation required
Open
Dietary request
Awaiting decision
Example attention type
Transfer provider
Response overdue
Example attention type
Attention before arrival. Consequential decisions stay human-owned.
Attention
Routine information should stay quiet until it needs judgment.
Host Desk is a focused attention layer around the promises that still need human action. It is not a full PMS dashboard and does not ask a seasonal employee to understand every system underneath.
The defining mechanism
When the guest relies on a condition
Promise Card
Later arrival after 20:00 requires written confirmation.
Attention Item
The responsible staff task starts open.
Human resolution
Confirm
Two state machines
What happened to the guest request is not the same as what happened to the staff task.
- pending
- confirmed
- declined
- expired
- cancelled
- open
- claimed
- resolved
- cancelled
One concrete handoff
The 23:30 request stays visible until someone resolves it.
The responsible person can confirm, decline, or otherwise resolve the arrangement. If the request is declined, the guest sees the correct outcome; a completed internal task does not hide it.
The arrival brief
Attention should arrive with context.
The host needs to see what to do, which system owns the fact, whether an AI answer may simplify it, and when the policy was last reviewed.
Later capability
Guest Memory is a later capability.
A one-time request should not quietly become a permanent preference. Reusable memory needs provenance, scope, consent where required, privacy controls, review, and visibility rules.
AI boundary
Assistance can scale. Judgment still needs an owner.
AI may later help classify, retrieve, draft, summarize, or compare. It should not initially confirm a conditional service, modify reservations, issue refunds, decide accessibility fit from incomplete facts, or infer permanent preferences.
What Host Desk is not
It is not a replacement front-desk operating system.
It is not a full PMS dashboard, housekeeping software, POS, accounting, revenue management, complete internal messaging suite, autonomous service recovery, or automatic approval of consequential requests.
Representative attention types
A few examples can stay visible without pretending to be the same demonstrated flow.
Dietary request
Awaiting decision
Example attention type
Transfer provider
Response overdue
Example attention type
Questions
The attention layer stays bounded.
Does every guest request create an attention item?
Not necessarily. The workflow should be driven by the promise and operating rule; already-established outcomes do not need unnecessary staff work.
Can a staff member resolve an Attention Item without resolving the guest request?
The states remain distinct so internal work does not accidentally imply a guest-facing outcome.
Is Host Desk a CRM?
No. Its public role is narrower: show what needs human attention.
Does AI confirm requests?
Not in the current product direction. Consequential confirmations remain human-owned.
Is Guest Memory part of Host Desk today?
No. Guest Memory is later
Keep going
Once the host knows what to do, the hotel still needs to know what every version is saying.
Continue from accountable host work into representation integrity.