Is the property identified?

No. Every case study on this site uses anonymised property profiles only (segment, room count band, market region). Property names and identifying details are never published.

Are the outcome numbers verifiable?

The figures come from operator-anonymised data plus published vendor benchmarks. Each outcome cites the relevant industry source where applicable; un-sourced figures are property-specific and presented with that framing.

Did the property use Guestivo?

Some case studies on this site reflect properties using Guestivo; others reflect properties using competing platforms. The case study text identifies the named tools used. Per editorial policy, no case study is positioned as a Guestivo marketing piece.

Case study · anonymised

Multi-property independent group: AI concierge deflection ramps from 45% to 72%

Independent group, 4 properties (140 rooms total), ADR ~EUR 95, multi-language guest mix (DE, EN, PL, ES)

The challenge

The group received approximately 250 inbound guest messages per day across email, WhatsApp, and the property website chat. Front-desk staff spent 90-120 minutes per day on repetitive questions (breakfast cutoff, late check-in, Wi-Fi password reset, restaurant recommendations) that pulled them away from real guest service. The group wanted to deflect the routine questions but had been burned by an earlier chatbot project that hit 35% deflection and stalled.

The approach

The team deployed an AI-concierge platform with a hospitality-tuned model and four-property setup. The deliberate process discipline was a 20-minute Friday review by the front-desk manager at each property: pull the AI transcript log, identify the top five unanswered or poorly-answered questions from the week, add them to the knowledge base. The platform exposed the unanswered queries as a queue, which made the review tractable.

Measured outcomes

Deflection rate, month 1

Before: n/a

After: ~45% on vendor-template knowledge base

Deflection rate, month 3

Before: ~45%

After: ~72% after weekly KB reviews

Front-desk minutes per day on routine messages

Before: 90-120 minutes

After: ~25-35 minutes (residual queries)

The failure pattern and the fix

The naive earlier project failed because no one was assigned to maintain the knowledge base; the AI answered "I am not sure, let me connect you with our team" to any question outside the vendor template. The fix was the 20-minute weekly discipline. The platform sat on top of the existing PMS (Mews) so guest context (booking dates, room type, preferences) flowed into the conversation; the earlier project had been on a standalone chatbot that did not see PMS context.

What we took away

The 45% to 72% ramp matches the published deflection curve from Hotel Tech Report for AI-concierge platforms with active KB maintenance. The lesson is operational rather than technical: the AI model is good enough; the bottleneck is whether someone reviews the unanswered-query queue every week. Multi-property groups can share the KB-review burden across property managers; single properties need to designate one owner.

Anonymisation note

This case study uses anonymised property data: segment, room-count band, market region, and outcome metrics. The property is not named. Operator-reported figures are presented with that framing; published industry benchmarks are cited inline.