The Bleeding Edge

// travel

AI in Travel & Hospitality

Hotels have moved from AI-as-feature to AI-as-operator, while the distribution layer argues over who pays for agent traffic. Still no conversion number from anyone.

standing summary updated September 25, 2026

Where AI in Travel & Hospitality actually stands

Travel's AI story began at discovery, and the chains have stopped trying to hold that ground on their own websites. IHG runs a Google-powered conversational search beta on IHG.com; Hilton runs an AI Planner on Hilton.com but has also extended planning, property search and Honors Points redemption into Google AI Mode, ChatGPT and Claude, where the session starts in someone else's window. Behind the chat layer, Google's agentic booking test has reached the GDS, with Amadeus named as the first infrastructure partner and Sabre offering the same capability, and Skift has reported a general-purpose assistant called Instinct booking and rebooking trips from a text thread — the transaction step, not just the search step. What has now surfaced underneath all of it is a bill. Amadeus and Spotnana both say the infrastructure behind AI-powered travel costs more to run than it used to, and the industry has not settled who absorbs that cost or who captures the value. Amadeus is on both sides of that sentence — the named partner making agentic booking possible, and one of the vendors saying it costs more than the old way. Inference

The second front has moved from the marketing surface into the back office, and it has changed shape: AI is now being sold as an operator rather than a tool. Sloan Dean, formerly CEO of Remington Hospitality, has launched AI Hospitality Group, billed as the first AI-native hotel operating company, with more than 60 proprietary agents running back-office work across a hotel's existing PMS, POS, revenue, finance and sales systems, targeting more than 500 basis points of GOP. That is a management-company margin rather than a software feature, which means an owner would hire it rather than approve it as a line item. Inference It also lands as a direct answer to the sharpest critique on this page — the Hospitality Net op-ed, drawing on Google Cloud research covering more than 2,400 executives, arguing that hotels have put AI at the edges in chatbots and upsell widgets rather than in core processes like pricing, labour and procurement, and that this is why returns are thin. Finance, revenue and sales systems are the core, not the edge. The operator-owned precedent was Aimbridge's LIFT labour forecasting; AI Hospitality Group goes further by making the agents the operating company rather than one of its tools. Inference

The numbers here have been cost-side only, and that is still nearly true. Cogwheel Marketing's beta put ChatGPT Ads for hotels at $3–$3.50 per click with click-through of 1.6–1.9%, with no conversion or booking value attached. A LodgIQ executive puts the industry-wide override rate on AI pricing recommendations above 50%, which measures trust and effort rather than return. Against that sits live legal exposure: a federal appeals court let the Atlantic City algorithmic-pricing case proceed, which makes a 50%-plus override rate look less like an ROI failure than like the cheapest available defence while the litigation runs. Inference The 500-basis-point GOP figure is the first yield number this page has carried, and it is a target rather than a result — a launch claim from a company with no portfolio track record behind it yet. The Cendyn beta, combining live ARI feeds and first-party guest data with Google's AI in a closed Search Campaigns for Travel test, still raises the other half of the data question: what a hotel hands Google to stay competitive once the beta opens. Inference

This was a thin week — two stories, one corroborated and one unverified, and between them a single number that describes a future rather than a result. Skift's cost story carries no figure at all: no indication of how much more agentic infrastructure costs, no proposed split, no supplier or agent that has moved first. The standing gap is unchanged. No operator has published evidence that AI discovery shifts actual reservations: not IHG's beta, not Hilton's launch, which came with no direct-booking mix or conversion figure, not Cendyn, which announced participation rather than a result, and not Aimbridge, which published no savings. Skift's Instinct report still names no hotel partner, no booking volume and no commercial terms, so what an agent channel charges a hotel and how it prices a room remains unanswered — as does the earlier unverified claim that booking agents read guest urgency and personal data to mark rooms up while the hotel collects its listed rate, which no platform has responded to. Expedia's AI-framed reorg is still a claim about speed to be judged on what ships. One conversion-level number from anyone at all would move this page further than another integration announcement.

// Big moves

2026-W39
  • Former Remington Hospitality CEO Sloan Dean has launched AI Hospitality Group, billed as the first AI-native hotel operating company: 60+ proprietary agents running back-office work across a hotel's existing PMS, POS, revenue, finance and sales systems, targeting 500+ basis points of GOP. The pitch is a management-company margin, not a software licence.

    CorroboratedHotel Divesource →

  • Amadeus and Spotnana both say the infrastructure behind AI-powered travel costs more to run than it used to, per Skift, and the industry has not settled who absorbs that cost or who captures the value it creates. No figure, no proposed split, no supplier or agent that has moved first.

    UnverifiedSkiftsource →

  • Amadeus sits on both sides of the agentic bill: it is the named first infrastructure partner for Google's agentic booking test at the GDS layer, and now one of the vendors saying that plumbing costs more than the old way. The cost lands on supplier margin, an agent fee or the traveller's price — nobody has said which.

    InferenceSkiftsource →

// Daily stories

  1. 2026-09-25

    Accor says its ALL Concierge, trained on more than 1 million pilot conversations, now covers 11 languages across five platforms including WhatsApp and iMessage. A parallel AI tool for hotel staff is launching alongside it.

    why it matters

    The first end-to-end concierge claim at this scale from a major chain, with a staff-side tool attached. The metrics to demand are containment and guest satisfaction per conversation, not language count.

  2. 2026-09-24

    Amadeus and Spotnana both say the infrastructure behind AI-powered travel costs more to run than it used to, according to Skift. The industry has not settled who absorbs that cost, or who captures the value it creates.

    why it matters

    If agent traffic raises the unit cost of every lookup and booking, that lands somewhere — supplier margin, an agent fee, or the traveller's price. Worth watching which side of the distribution chain moves first.

  3. 2026-09-23

    Former Remington Hospitality CEO Sloan Dean has launched AI Hospitality Group, billed as the first AI-native hotel operating company. More than 60 proprietary agents run back-office work across a hotel's existing PMS, POS, revenue, finance and sales systems, targeting 500+ basis points of GOP.

    why it matters

    The claim on offer is a management-company margin, not a software feature, which moves AI from a line item owners approve to an operator they hire. Worth watching whether the 500bp target survives contact with a real portfolio.

  4. 2026-09-12

    Hospitality Net's Friday roundup carried Pertlink data showing only 6% of hotels appear consistently in AI recommendations, with 82% of cited content sourced from OTAs. The same roundup covered the argument that hotel websites have shifted from discovery tools to verification tools.

    why it matters

    If assistants cite OTAs for most hotel content, direct-booking spend is buying verification rather than discovery. Worth measuring your own properties' citation rate before the next content or rate-parity budget cycle.

  5. 2026-09-11

    Skift reports that Meta's Muse agent connects directly into flights but shops the rest of travel through the open web — which puts the agent, rather than a travel integration, in control of where the search goes.

    why it matters

    If agents discover supply by crawling rather than through a distribution deal, your public content becomes the distribution channel. Worth checking how your rates and availability read to an agent with no integration.

  6. 2026-09-09

    Cendyn has joined Google's closed beta for Search Campaigns for Travel, combining live ARI feeds and first-party guest data with Google's AI to improve hotel search ad performance ahead of general availability.

    why it matters

    If live availability and rate data move inside Google's ad targeting, hotel paid search stops being a static bidding exercise. Watch what first-party guest data you would have to hand over to compete once this leaves beta.

  7. 2026-09-08

    Writing in Hospitality Net, the author draws on Google Cloud research covering more than 2,400 executives to argue that hotels are deploying AI at the edges rather than embedding it in core value-creating processes, which limits returns.

    why it matters

    If your AI spend sits in chatbots and upsell widgets, this is the argument to test it against. Worth asking which core process — pricing, labour, procurement — you would actually rebuild around AI.

  8. 2026-09-05

    Skift reports that Instinct, a new AI assistant drawing attention in Silicon Valley, can book and rebook trips from a text thread. Skift's takeaway for the industry: talking about frictionless booking is not the same as delivering it.

    why it matters

    A general-purpose assistant that rebooks without a call centre resets the bar for what self-service means. Worth testing your own change-and-cancel flow against a text thread before a guest does it for you.

  9. 2026-09-03

    Aimbridge Hospitality launched its Labor Insight & Forecasting Tool, which per a news release combines labor data from multiple systems into 'one connected view' using data science and AI-assisted forecasting. CIO Keryn McNamara framed it as a common reference for planning labor.

    why it matters

    Labor is the largest controllable cost in a managed hotel, and here the operator, not a vendor, owns the forecast. Worth asking whether the savings reach owner P&Ls before this becomes a management-agreement selling point.

  10. 2026-08-31

    Hilton has launched an AI Planner on Hilton.com and extended its technology platform into Google AI Mode, ChatGPT and Claude, enabling conversational trip planning, property search and Honors Points redemption.

    why it matters

    A major chain is placing discovery inside third-party assistants rather than defending its own funnel. Worth watching whether direct-booking economics hold when planning and points redemption start in someone else's chat window.

  11. 2026-08-28

    Writing for Hospitality Net, a LodgIQ executive argues that more than 50% of AI pricing recommendations are overridden industry-wide, attributing it to hotels skipping the trust-building stages needed before full automation rather than to the technology failing.

    why it matters

    If the override rate is anywhere near that, revenue-management ROI is being set by adoption design, not model quality. Note the source sells the tooling — worth checking your own accept-versus-override rate before the next renewal.

  12. 2026-08-27

    A federal appeals court ruled that a lawsuit against several Atlantic City casino hotels can move forward, PYMNTS reports. The case centres on competing hotels feeding their own private pricing data into the same pricing program.

    why it matters

    Revenue-management stacks that pool rivals' private rate data now carry live litigation exposure, not just theoretical risk. Worth asking your pricing vendor exactly whose data trains the recommendation you accept.

  13. 2026-08-25

    Hospitality Net reports Google's agentic hotel booking test now reaches the GDS, with Amadeus named as the first infrastructure partner and Sabre offering the same capability — giving independent hotels a second route into the agent that skips the OTA.

    why it matters

    For independents, distribution into the booking agent now runs through the GDS as well as the OTAs. Worth checking what your GDS contract and content feed actually expose before the test widens.

  14. 2026-08-22

    A Hospitality Net opinion piece describes Shanghai's 100 AI-powered tourist terminals, built on more than 84,000 content entries and 14,000 booking links, and presents them as a reference model for destinations such as the Philippines to build a unified data layer before any interface.

    why it matters

    The argument is that the data layer, not the chatbot, is the hard part. Worth checking whether your own content is structured enough to answer an AI query before funding another guest-facing interface.

  15. 2026-08-21

    Expedia Group is shedding eight senior leaders and shifting others to reshape leadership and work structure, Skift reports, framing the move as a reorganisation around AI. The stated goal is to help product and technology efforts move faster.

    why it matters

    Expedia is a bellwether for travel distribution. The stated goal here is speed, not cost, so watch what actually ships over the next two quarters before reading this as an AI headcount story.

  16. 2026-08-20

    In an interview with Hospitality Net, IHG's SVP of digital direct channels Kim Smith described the Google-powered conversational search beta running on IHG.com, the chain's personalisation roadmap, and the case that hotels must restructure content to stay discoverable in AI-driven booking.

    why it matters

    If a brand of IHG's size is reworking content for conversational search, the near-term job for hotels is structured, machine-readable property data — not another channel. Watch what the beta does to direct-booking mix.

  17. 2026-08-15

    A Hospitality Net opinion piece argues that AI agents are reading guest urgency and personal data to mark up room prices, with hotels receiving only their listed rate while channels pocket the premium. It calls direct booking the only channel that cannot exploit guest data.

    why it matters

    If the markup sits with the agent rather than the hotel, agentic distribution repeats the OTA commission problem in a less visible form. Worth pressure-testing rate parity and direct-booking economics before agent volume matters.

  18. 2026-08-13

    Hospitality Net reports live ChatGPT Ads beta results for hotels from Cogwheel Marketing, showing $3-$3.50 cost per click and click-through rates of 1.6-1.9%. The same briefing covers Tripadvisor Group and Airbnb putting 425,000 tours and activities on Airbnb's platform.

    why it matters

    This is the first public cost benchmark for paid placement inside an AI assistant — one agency's beta, not yet a channel. Enough to justify a small test budget and a question to your digital agency, not a reallocation.

  19. 2026-08-12

    Google Maps is adding spatial context to AI hotel search. Skift notes that adding personal itinerary data would make this more than another AI search update. Booking is not part of it yet.

    why it matters

    If discovery moves into Maps, the economics of direct booking change for every hotel operator. The absence of booking is the thing to watch, not the presence of search.

// Track AI in Travel & Hospitality

One email a week on what actually changed — in travel & hospitality and across the frontier. Verified, labelled, no hype.