Skip to content

Haven by HCCGSA

Haven methodology

How Haven handles a guest issue, and how we have tested it so far.

Draft. This page is not final. Items marked NEEDS CONFIRMATION are waiting on information from HCCGSA and must be resolved before publishing.

Haven method: Capture, Classify, Route, Recover, Retain1Capture2Classify3Route4Recover5RetainRetained history informs the next stay

What Haven does with a guest issue

  • Intake: staff log an issue, or a guest submits one through Haven Guest Link™ (a QR code).
  • Routing: the issue is assigned to a department and a person.
  • Follow-through: open issues stay visible across shifts until someone marks them resolved.
  • Memory: preferences and history stay attached to the guest for future stays.

Two layers: rules and AI

Haven combines fixed rules (for example, escalation when an issue stays open too long) with an AI layer that drafts suggested responses and categorizes issues. Staff decide what is sent to guests. [NEEDS CONFIRMATION: which AI provider and model are used, and whether any guest text is sent to that provider]

How Haven has been evaluated

Haven has been evaluated only in simulation. In HCCGSA's internal synthetic evaluation, Haven scored 97.5 out of 100 against a 64.0 manual baseline across 60 simulated runs, with 0 safety failures. These numbers come from simulated complaint scenarios, not real hotels.

[NEEDS CONFIRMATION: public link to the evaluation logs and scoring rubric; who designed the scenarios; how "safety failure" is defined]

Pilot results from real properties will be published here once available, with the property's permission. No pilot results are claimed today.

What this page does not claim

Patents in preparation and research identifiers describe what HCCGSA is developing and has written. They are not evidence that Haven improves guest outcomes.

Book a 20-minute guest-memory audit