Resources · Case Studies
What hotels actually got
when they bought the AI.
Named vendors, named properties, the metrics they reported, and the caveats that don't transfer. Anchor cases get a full case page; library entries surface vendor, outcome, and evidence tier. Filter by department, region, outcome type, evidence tier, and operating logic · augmentation vs replacement is the architectural decision worth flagging early.
Evidence tiers travel with each case: independent reporting, named vendor case, counter-pattern, and documented failure. Treat vendor-published metrics as hypotheses until you find an independent source.
Legend · How to read the filters
↓
Region
- Americas· North & South America
- EMEA· Europe, Middle East, Africa
- APAC· Asia-Pacific excl. China
- MENA· Middle East & North Africa
- China· Mainland China deployments
- Global· Multi-region or vendor-wide
Outcome
- Revenue· RevPAR, ADR, conversion, upsell
- Cost· Labour hours, food cost, OPEX
- Time· Cycle time, handle time, latency
- Sustainability· Waste, energy, water, carbon
- Guest satisfaction· GRI, NPS, review scores
Evidence
- Tier 01
Earnings calls, third-party audits, named executives.
- Tier 02 → 01
Vendor case now cross-reported in independent press.
- Tier 02
Vendor-published with named property and metrics.
- Counter-pattern
Architecture credible, outcome metric not yet verified.
- Failure
Discontinued, retracted, or reported as not delivering.
Operating logic
- Augmentation
Human keeps the call. AI surfaces, humans decide.
- Support
Behind-the-scenes routing, measurement, first response.
- Replacement
AI substitutes for staff. High failure risk if brittle.
Filters
Region
Outcome
Evidence
Operating logic
Meal saved every six seconds
vendor-reported
Accor hotels in deployment
No independent audit · methodology not disclosed
Vendor-reported aggregate
Vendor-reported "1 meal every 6 seconds" saved across the Accor deployment of 200+ hotels. No independently audited Accor-published case-level metric ties this back to a specific RevPAR or food-cost ratio. Useful as scale-of-segment context only, never quote as a property-level…
Damages awarded · BC Civil Resolution Tribunal
Operator owns chatbot output
Precedent for hospitality / OTA chatbots
Moffatt v. Air Canada, 2024 BCCRT 149 · February 2024
Moffatt v. Air Canada, 2024 BCCRT 149, the BC Civil Resolution Tribunal awarded CAN$812.02 in damages against Air Canada for chatbot misrepresentation. The legal precedent is the relevant signal for hospitality operators: a single screenshot of an AI-invented room type, refund…
Properties on ALICE · 260+ integrations
#1 Concierge at HTA
incl. 2026
Ticket dispatch logic is rules-based, not ML
8 consecutive years #1 Concierge Software at HotelTechAwards
ALICE is the operations layer of the four-product Actabl stack (ALICE ops + ProfitSword BI + Hotel Effectiveness labour + Transcendent assets). Service-request routing, housekeeping board, concierge tickets, internal messaging, two-way integrations with Mews, Maestro,…
RevPAR uplift within 6 months · new-to-RMS hotels
Hours/week saved on Autopilot
Of users on Autopilot (full auto-pricing) since Aug 2025 GenAI launch
Mews-published outcomes since November 2024 acquisition
Atomize is a cloud RMS that monitors demand, booking pace, and competitor rates and pushes per-room-type prices in real time, up to two years (730 days) forward. ML-driven optimisation with human-set strategy parameters; each price recommendation carries an…
YoY revenue growth · Brambles of Inveraray
Rooms scaled during AI-tooling adoption period
Room independent · same PMS as channel manager
AI Concierge launched May 2023
eviivo Concierge is a Gen-AI virtual concierge embedded on the property's own website, answering prospect and in-house guest enquiries with curated, brand-tone responses. Sits on top of eviivo Suite. Confirmed result at Brambles: +8% YoY revenue growth (eviivo / Expedia Group…
Hotels on Canary's underlying stack
Named enterprise customers
Marriott · Four Seasons · Wyndham · Choice · IHG · BWH
Agent Studio launch · no operational metrics published yet
Launched 3 March 2026
Hospitality-specific agent builder launched 3 March 2026. Hoteliers configure pre-built templates (Front Desk, Concierge, Central Reservations) or custom workflows that integrate with PMS, messaging, and voice. Sits on top of Canary's existing stack: AI Guest Messaging, AI…
Brands on a single agent runtime
Hotels · ~650K rooms
Standardisation milestone · deployment announcement, not results
April 2026 standardisation milestone
Choice standardised on AWS AgentCore as the shared runtime for intelligent agents across all 22 brands. Three named domains: guest-facing (search, booking recommendations, on-property experiences), franchisee operations (revenue management, maintenance, guest comms), and…
Vendor benchmark · autonomous handling
Portfolio-average automation
Named hotel-brand customers
2025
Voice + SMS + WhatsApp + email + OTA messaging in one AI inbox; human-in-the-loop escalation for the residual 20%. Vendor's own benchmark: 80% autonomous handling, 66% portfolio-average. Two named customers in their material: a "35-property manager" (96% automation) and…
Products: Mobile Check-In · AI Guest Messaging · Digital Compendium · Dynamic Upsells · Smart Checkout
No app download
No deployment metrics published yet
January 2026 announcement
Five-product rollout, Mobile Check-In, AI Guest Messaging, Digital Compendium, Dynamic Upsells, Smart Checkout, all served via mobile web (no app download). The press references "Crown Resorts' Australian properties" generically, with Crown Perth in illustrative photography.…
RevPAR uplift · Jannah Hotels & Resorts (UAE) · +34% ADR
Of Grupo Posadas (199 properties) at top of RGI index
GOPPAR lift in 2025 · Duetto + HotStats co-cohort
Founder-team ex-Wynn · first customer El Cortez Las Vegas
"Open Pricing" is Duetto's pitch against the legacy BAR-based RMS lineage. In BAR, room types and channels are priced as fixed modifiers off a single Best Available Rate; when demand mismatches, the operator must close the channel. Open Pricing decouples them: every room type,…
Group-average food-waste reduction
year-1
Group savings year-1
Meals/year saved · Armani Hotel Dubai
AED 148K
Pilot Feb 2018 → POC March 2019 → official launch May 2019
AI-vision food-waste tracking deployed across 12 Emaar commercial kitchens in the UAE, pilot Feb 2018 → POC March 2019 → official launch May 2019, partnered with the UAE Ministry of Climate Change and Environment (MOCCAE). Each property's chef de cuisine owns the daily review;…
Companies abandoning most AI initiatives in 2025
S&P/451, up from 17% in 2024
Average proportion of GenAI PoCs scrapped before production
S&P/451
GenAI pilots delivering no measurable P&L impact
MIT NANDA, Aug 2025
Gartner July 2024 → MIT & S&P 2025 · Gartner agentic-AI follow-up June 2025
Gartner's original July 2024 prediction (Rita Sallam, Distinguished VP Analyst, on a survey of 822 business leaders): "At least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls,…
- S&P Global / 451 Research (1,006 mid-level and senior IT/LOB respondents, NA + EU): share of companies abandoning most AI initiatives jumped from 17% in 2024 to 42% in 2025; average organisation scrapped 46% of GenAI PoCs before production
- MIT NANDA Initiative (Aug 2025; 150 leader interviews + 350 employee survey + 300 deployments): only ~5% of GenAI pilots achieved rapid revenue acceleration; 95% stalled
- BCG (2025; 1,250 respondents): 60% generate no material value; only 5% generate substantial value at scale
Price the coin-flip into the budget, the timeline, and the political cost of saying "we tried it and stopped." Per-pilot governance discipline (clear KPI targets, named decision owner, forced kill-criteria, post-mortem…
No hospitality-specific GenAI abandonment-rate data has been published by Gartner, Forrester, IDC, S&P/451, MIT, or BCG as of April 2026. The cross-industry data is the floor; whether hospitality is above or below is…
Robots peak
2018, up from ~80 at 2015 launch
Removed during 2018–early 2019
Henn-na branded properties still operating in 2026
narrow-task only
Opened 17 July 2015 · cull reported January 2019 (WSJ)
Henn-na ("Strange") opened on 17 July 2015 at the Huis Ten Bosch theme park in Sasebo as the world's first robot-staffed hotel, a vision underwritten by H.I.S. Co. founder Hideo Sawada, who told reporters at launch that the chain would scale to ~90% robot staffing across 100…
- >50% of robots removed during 2018 and early 2019
- Sawada's verdict (verbatim, WSJ January 2019): 'When you actually use robots you realize there are places where they aren't needed, or just annoy people.'
- ~20 Henn-na branded properties still operate in 2026, exclusively with narrow-task robots, not replacement-logic deployment
The negative pattern. Henn-na's value as a case is in what it teaches you not to do: do not adopt AI on a replacement thesis if the human work being replaced contains exceptions, judgment, or regulatory compliance the…
Henn-na's specific architectural choice (purpose-built building physically designed around robot navigation) is a 2015-era mistake that almost no operator will repeat, but the operating logic temptation persists. The…
AI use cases under active management
Chatbot resolution time · 90% positive feedback
LightStay cumulative savings since 2009 · 9,100+ properties
Disclosed Q3 2025 earnings · 22 October 2025
Three buckets: process efficiency, distribution/marketing, customer experience. Nassetta deliberately withheld granular per-use-case detail (citing competitive sensitivity). Named use cases in trade-press coverage: marketing automation with Google (double-digit incremental…
Food-waste reduction · production kitchens
Annual savings at the property
Item-level measurement
was: guesswork
Hilton Dubai kitchen programme
Winnow Vision combines a camera, smart scale, tablet, and recipe-cost attribution so kitchens can see what is being wasted and where the cost is coming from. The AI measures; the humans act. The deck's anchor case is Hilton Dubai · Al Habtoor City: the property reports roughly…
- ~70% food-waste reduction across production kitchens at Hilton Dubai Al Habtoor City
- $65K annual savings at the property
- Daily item-level cost signal that previously did not exist
Computer vision on a single, high-cost waste stream: measure first, then act. The ROI story is concrete (food cost · CO₂ · hours · bin weight). The operational discipline matters as much as the tech: weekly…
The ROI depends on high-volume F&B operations. A 30-cover restaurant will not produce Mandarin-scale numbers, the absolute waste pool is too small. Concentration risk is real: a single-vendor story is an operating…
Plate-waste reduction
102g→64g per cover
Total food-waste reduction
CO₂e avoided · 6,376 meals saved · 340,000+ guests served
Ramadan 2025
Winnow Vision deployed at scale during Ramadan to attack plate waste in a high-volume, time-bounded service (Iftar/Suhoor buffets, the highest-waste service period in EMEA). Daily dashboard fed back into par-level retuning across the month; executive chef owned the response…
Cumulative utility savings since 2009
CO₂ intensity reduction vs 2008
Reduction in resource consumption
Cumulative since 2009
Hilton's LightStay is the older energy-and-resource-management programme, pre-dates the GenAI wave, and is the foundation that Hilton's newer 41-AI-use-case portfolio builds on top of. Numbers are independently verified by KEMA and DEKRA. Mandatory across 9,100+ properties.…
Itineraries actually built
despite 'trip planner' branding
Manhattan property surfaced for a Brooklyn search
Still shipping unredesigned · no public retraction
Launched May 2025 · still shipping in 2026
In May 2025 Expedia Group relaunched the Hotels.com app with an AI-powered chatbot trip planner as the headline feature, part of a broader brand revitalisation following loyalty-programme backlash. The chatbot is built on Expedia Group's underlying AI travel assistant Romie,…
- No retention or trust metrics published, Skift's review is qualitative
- No Expedia Group earnings-call mention specifically calling out the Hotels.com chatbot in 2025 calls
- No public redesign or pull-back announced as of April 2026, the chatbot continues to ship
The negative pattern: do not ship a marquee AI feature without testing it against the basic accuracy bar of the non-AI feature it replaces. The launch standard for AI features must be at least as high as the launch…
Expedia Group has serious AI assets, Romie, the EG Labs alpha, the OpenAI ChatGPT Apps partnership announced October 2025. Capability is not the issue. What shipped on Hotels.com was a marquee feature in a…
Properties · 169 countries
Accor hotels · 110+ countries · 45+ brands
phased rollout
Decision authority, advisory to fully automated
IDeaS (a SAS company) is the incumbent enterprise RMS, pricing science, demand forecasting, group-revenue optimisation, and total-revenue/profit modules. Decision authority is configurable from advisory to fully automated. Confirmed enterprise customers visible on the homepage…
Direct cost savings
single-source CIO India
Hours of human effort optimised
Processes automated
Programme led by Vishal Anand (Senior Director of IT)
Two-layer stack: Microsoft PowerApps as the low-code UI/workflow front-end, UiPath as the headless RPA engine for screen-scraping legacy PMS, finance, and HR systems. Named guest-facing workflows: automated F&B recommendations, room-upgrade routing, online check-in. Reported:…
ROAS lift
vendor-published, baseline undisclosed
CPC reduction
Hours of manual effort saved
Always-on PBA
Always-on Predictive Budget Allocation (PBA) is an ML model that runs 1,000+ automated optimisations across the campaign portfolio, reallocating budget between Meta/Google ad sets in near-real-time against ROAS and booking-window KPIs. Smartly's automation feeds + Workspaces…
Food-waste reduction
Annualised group savings
Waste avoided · 289t CO₂e
Mandarin Oriental group press release, 5 June 2024
AI-vision food-waste tracking (Winnow Vision: camera + smart scale + tablet) deployed across the four-property pilot, paired with kitchen-team redesign of menus and prep. The pilot ran 6 months before Mandarin extended Winnow group-wide. The property-level lever is daily…
Response time · 5–10 min → ≤2 min
Roomers Hotel
Response rate · Little BIG Hotels
Response rate · Gekko House
Per-property pricing from €60/mo
MARA centralises feedback from Booking.com, Google, and TripAdvisor into a single inbox; AI drafts each reply tuned to sentiment, tone, and topic, deliberately varying phrasing to avoid template tells. Named features: Review Inbox, Smart Snippets, Brand Voice, Review Analytics,…
RENAI properties globally
March 2024
ACU upgrades · day-before 3pm local
Associates still confirm at check-in
RENAI pilot Dec 2023 · ACU live 14 July 2025
RENAI (Renaissance Hotels): QR code → SMS or WhatsApp. AI blends "Renaissance Navigator" human picks (compass-emoji tagged) with ChatGPT + curated open sources. Pilot December 2023 at Charleston, Plano, Nashville Downtown; expanded to 20+ properties globally by March 2024. ACU…
Test restaurants in the pilot
Pilot terminated
Cited failure modes
Pilot ended 26 July 2024
McDonald's ended the IBM AI drive-thru pilot across 100+ test restaurants on 26 July 2024, citing accent-recognition and order-accuracy failures. The closest hospitality-adjacent rollback referenced in the GenAI abandonment landscape (Section 001-2): voice AI at the…
Packages sold
vendor-verified on Thaiger case page
Property · single mid-market Novotel pilot in Thailand
Vendor case only · no Accor press · no Thai trade-press
Single-source vendor case study
Stack: an AI-built landing page, paid social campaigns on TikTok / YouTube / Meta, and ThaigerGo CRM running a multi-touch email + SMS drip. The funnel runs until the lead books, opts out, or times out, i.e. classic marketing automation with AI in the chatbot/copy layer rather…
RevPAR uplift
ADR · average daily rate
RGI 2019–2023
since 2022 · 44 properties
OBVIO replaced spreadsheet-driven forecasting across 44 Accor-branded properties with IDeaS G3 starting in 2022. The deployment sits inside the broader Accor + IDeaS global revenue-management partnership (Feb 2024), which is being rolled out across 5,000+ Accor properties in…
- ~20% RevPAR uplift since 2022 across the OBVIO programme
- +14% ADR year-on-year
- +9.5% RGI (Revenue Generating Index) for 2019–2023
The clean separation of AI forecast vs human decision. Staff retain override authority; the scorecard measures the system as a whole, model + manager, not the model in isolation. The transferable architectural pattern…
Accor's scale gives IDeaS data advantages specific to the Accor network. A single boutique property will not replicate Accor's performance curve on day one; the RMS takes months to learn a property's demand…
Group-wide ROI across 8 hotels
Top-property ROI · Woolloomooloo
cherry-picked
Leads → reservations recovered
group-wide
2-month campaign from April 2022
Ovolo is an 8-property boutique group. Revinate is a hotel CRM and marketing-automation platform. The pilot ran 2 months from April 2022 under group marketing director Stephen Howard, and targeted abandoned-cart recovery, guests who started a booking on the property's own…
- 17× group-wide ROI across 8 properties (the operator-grade headline)
- Range across the 8 properties: 4× (Mamaka) to 52× (Woolloomooloo); the spread is the operator-grade story
- 600 abandoned-cart leads → 92 reservations recovered (group-wide, 2-month campaign)
Lead with the spread, never the cherry-picked headline. Cherry-picking the top-performing property's ROI (52×) is exactly what AI-washing looks like. The transferable pattern is the framing rule: when a vendor shows you…
Abandoned-cart recovery requires a property with enough direct bookings and enough abandoned-cart events to make a campaign statistically meaningful. OTA-dependent properties (where guests book on Booking.com or…
Phased rollout begins · no audited KPIs published
Specialised AI agents named in trade-press summaries
Integration areas: housekeeping · F&B · guest services · concierge · reservations
phased rollout from December 2025 / January 2026
Palazzo Versace Dubai launched Laura, a digital-human AI concierge built on Quantum Neuron's hospitality platform. The deck uses Laura as the guest-operations anchor because it shows how far guest AI has moved since single-channel voice assistants. The pattern is multi-channel,…
- No audited KPI has been published yet; the rollout is live across selected touchpoints, with further capabilities being phased in
- Relevant evidence is therefore deployment shape, not ROI: guest-facing AI connected to internal service systems, operating as a support layer
The support-layer framing. Laura is useful as a pattern when it strengthens operational execution: multilingual first response, routine request handling, service-task routing, and staff visibility into what the system…
Luxury and ultra-luxury operators can justify integration work and supervision protocols for VIP-grade interactions. A smaller property should not copy the digital-human wrapper unless the back-end workflow is equally…
Named hotel deployment of a branded predictive-maintenance vendor in trade press
Otis ONE · KONE 24/7 · Schindler Ahead in elevator/escalator service contracts
Augury vibration + ultrasonic on chillers/pumps
manufacturing-anchored
Cross-industry · 2026 maturity
Two distinct architectures dominate: (a) OEM-embedded, Otis ONE connects elevators to Otis cloud and dispatches technicians before failure; KONE 24/7 Connected Services and Schindler Ahead are direct equivalents. (b) Retrofit sensors, Augury uses retrofit vibration + ultrasonic…
Room-turnaround reduction · single-source trade-press
Hotelspeak, Apr 2026
Optii is at sister Ritz-Carlton Cleveland; SF attribution is inference
Cleanliness uplift at the Cleveland comparator
Optii / HotelTechReport
Hotelspeak coverage, April 2026
Predictive model layers historical check-out times by room type, length-of-stay patterns, stay-over vs departure mix, and staff availability. Suppresses housekeeping dispatch to rooms that won't free up until late, re-routing attendants to confirmed departures. Reported result:…
Flights launch · hotels phased after
Airlines (incl. 150 LCCs) in Sabre's inventory base
No hotel deployments live · no measured conversion or servicing data
Launches Q2 2026 (flights first); hotels phased after
Three-party stack: Mindtrip is the conversational front-end ("describe destination, timing, budget in plain language"); Sabre Mosaic APIs supply shopping, booking and servicing across the inventory; PayPal supplies identity, checkout, and BNPL. Architecture details beyond…
Distinct robot models in the fleet
VLA foundation model + KEENON ProS hospitality fine-tune
Outcome-unverified · no KPIs published
Announced 28–31 October 2025 · launch coverage only
The fleet, six named robot models, each task-specific: - XMAN-R1, humanoid greeter (front-desk welcome, natural-language Q&A, welcome-gift presentation) - W3, in-room delivery (amenity drop-off) - S100, high-payload luggage transport - C40, autonomous cleaning - T10 and T3, F&B…
- No KPIs, no cost figures, no guest-satisfaction data, no supervision-protocol details, and no executive on-record quotes from either Shangri-La or Keenon were published in any of the seven canonical launch sources
- No 2026 follow-up coverage of operational performance has appeared as of April 2026
- Confirmed gap, not a research oversight
The structural choices distinguishing Shangri-La/Keenon from Henn-na are task decomposition and vertical fine-tuning on top of a foundation model, not the humanoid form factor itself. Adopt the pattern; treat the…
Vendor superlatives ("world's first general-purpose + special-purpose hotel collaboration model") do not transfer. The architecture is credible but the outcome is unverified. Foundation-model + vertical-fine-tune…
ADR · per +1 GRI point
Cornell
RevPAR · per +1 GRI point
Cornell
Hospitality concepts · 17 languages
semantic NLP layer
Cornell Anderson study (peer-reviewed)
Two distinct AI layers, often conflated. GRI is a 0–100 weighted aggregate of numeric review scores from 140+ OTAs/review sites in 45+ languages, pure quantitative roll-up, no NLP, recency-weighted. Semantic Analysis is the NLP layer: classifies review text against 700+…
Reduction in time spent crafting responses
ResponseAI
AI Agent Core · per property/month
annual billing
Hospitality businesses · ResponseAI in use
CDP launched 2 July 2024 · AI Agents launched 25 March 2025
Three-pillar story: CXP (review aggregation + ResponseAI auto-replies + SentimentAI), CDP (identity-resolved "Golden Profiles" + consent management + audience segmentation), AI Agents, Booking Agent, Guest Agent (24/7 digital concierge with feedback collection + staff alerting),…
Average HVAC runtime reduction
when rooms unoccupied
Reduction on hotel energy costs
Payback · utility rebates may cover 100% of cost
IHG mandate Nov 2025, required for new-build & renovated rooms
Verdant is a per-room HVAC controller, not a building-wide BMS. Each thermostat has built-in PIR (passive infrared) motion + body-heat sensing; when a room is unoccupied, setpoints drift to a recovery band, then snap back when the guest returns. Optimisation horizon is per-room…
Annual savings
vendor-aggregate, no third-party audit
Food waste avoided · 122,000t CO₂e
Per-customer food-cost saving claim
more defensible than headline
Industry-wide claim · 2026
Aggregate impact across the customer base, Hilton, Accor (200+ hotels), Marriott, Mandarin Oriental, IKEA, Guckenheimer. Methodology is internal Winnow telemetry: camera-logged waste events × food cost × estimated counterfactual ("reductions typically over 50%"). No third-party…
Narrow agents
1 task each
Direct contribution improvement
Q3 2025
Average handle time
Q3–Q4 2025 disclosed outcomes
Wyndham's Connect Plus stack runs ~350 narrow agents, each scoped to one task (booking modification, early check-in, upgrade, recommendation, FAQ in SMS or voice), across 8,300+ hotels under Days Inn, Super 8, Ramada, La Quinta and other brands. Multi-vendor architecture:…
- Q3 2025 transcript: nearly 300 bps direct-contribution improvement; >500K customer interactions; 25% reduction in average handle time
- Q4 2025 transcript: nearly 350 agents handling millions of interactions; hundreds of bps additional direct bookings
- PwC case study (call-centre scope): 30–50% AHT reduction; 28% of incoming calls handled by AI; 94% reduction in brand-standards review time
How to read the directory
Evidence tier
Independent sources outweigh vendor-published metrics. The chip on every card is a shortcut for how much trust to put in the headline number.
Operating logic
Augmentation keeps humans in the loop. Replacement substitutes for staff · Henn-na's discontinuation is the cautionary case. Support is invisible to guests.
What to copy vs what doesn't transfer
Each anchor case names the architectural pattern that's portable and the scale dependencies that aren't. Read both before assuming a vendor's number is yours.