McKinsey’s Technology Trends Outlook 2025 underlines a near-term shift that hospitality owner-operators must plan for: agentic AI investment hit $1.1 billion in 2024, agentic-AI job postings rose 985 percent in two years, and 78 percent of organizations use AI in at least one function while only 1 percent describe their deployment as fully mature. That gap between adoption and maturity is where the next eighteen months of hospitality technology strategy will be decided.
Agentic AI: The decision layer keeps moving
McKinsey frames agentic AI as systems built to act, not merely to converse—booking, filling forms, executing multi-step workflows and coordinating with other agents through emerging standards such as Anthropic’s Model Context Protocol and Google’s Agent2Agent protocol. The report’s case studies describe general-purpose and research agents that shortlist and synthesize before a human opens a tab, quantifying momentum behind a shift that moves the guest decision layer out of OTA interfaces and into agent reasoning traces.
McKinsey scores agentic-AI adoption at 2 out of 5—small-scale experimentation rather than scaled deployment—meaning the strategic window is now while standards are still being written. Concepts such as agent engine optimization (AEO) and GEO are presented as the SEO-equivalents of the near future: properties that delay may find agents have already encoded third-party preferences into the discovery layer.
“Agents won’t just automate tasks — they’ll reshape how work gets done,” paraphrased from Lareina Yee.
The agent books it — who keeps the guest?
Airline deployments in production illustrate the sequence McKinsey describes. Amadeus deploys agent teams across network planning, marketing, turnaround and disruption management; Southwest uses agentic tooling for crew scheduling; Alaska Airlines and Volantio run agents that identify likely-oversold flights, select passengers by business rules, construct compensation offers and propose changes without a human step. That observe→evaluate→offer→act sequence maps closely to hotel oversell and upsell workflows.
The Viewpoint sets out a practical definition of Agentic Decision Rights to be defined ahead of vendor requests: OBSERVE (which conditions an agent may monitor), DECIDE (which decisions it may make unassisted), OFFER (what commercial propositions it may construct), COMMIT (what it may transact without approval) and ESCALATE (when a human must take over).
Consumer appetite for delegation is rising: Amadeus APAC research finds 40 percent of travelers open to letting AI book on their behalf. An Amadeus Hospitality study reports 69 percent of travelers consider an AI summary sufficient to decide without further checking, rising to 87 percent in India and 86 percent in China. That shortlisting-before-search dynamic revives the guest-ownership issue hospitality has faced with OTAs: agents can hold intent before a property’s own channels see the guest.
Distribution is shifting to infrastructure. Eviivo’s distribution partnership with Dida exposes independent-hotel inventory to roughly 40,000 APAC distributors and pairs that reach with an MCP-native AI booking gateway designed to sit inside third-party agent applications. The recommended operational response is to track distribution provenance for every reservation—guest interface, demand source, distributor, inventory source and commission—and to measure a First-Party Conversion Rate across a funnel of discovered, selected, booked, known and retained.
AI investments, costs and the Token Cost Per Guest
McKinsey finds 92 percent of executives plan to increase AI investment over the next three years while only 1 percent call their deployments fully mature. The report attributes the gap to organizational factors—process adaptation, capability building and workforce reskilling—rather than model capability. That aligns with the Token Cost Per Guest (TCPG) framing: falling inference costs and smaller domain models put AI within reach of mid-market groups, but unit economics must be priced per guest rather than per token to avoid uncontrolled consumption.
The Viewpoint cautions about tokenmaxxing dynamics discussed ahead of HITEC 2026: cheaper inference can still produce higher bills if consumption incentives are misaligned. Multimodal AI—models handling text, images, video and audio together—has moved from research to production within three years of generative AI’s commercial availability, enabling use cases such as AI-assisted property inspections and voice- or camera-enabled concierge tools. Capability is rarely the constraint; organizational adoption is.
McKinsey’s data includes a property-level example: the same AI-enabled guest-engagement platform produced highly engaged properties generating over $60,000 in incremental annual revenue each, while passive properties on identical software produced close to nothing. The difference was adoption not capability. The recommended milestone approach separates go-live (the system works), adoption (staff use it correctly) and value (measurable results), with a 90-day value review—feature utilization, workflow compliance, manual workarounds and revenue or cost outcomes—as basic insurance.
Digital trust and cybersecurity as adoption gates
Trust is central to McKinsey’s 2025 cross-cutting themes: as systems become more powerful and personal, trust becomes the gatekeeper to adoption. McKinsey reports cybersecurity and digital-trust investment at $77.8 billion in 2024, up 7 percent year on year, and notes that agentic systems—which can take real-world actions—raise governance, legal and reputational stakes. An autonomous agent that can book, cancel or reprice can also be socially engineered or prompt-injected; balancing autonomy and human oversight belongs on the same governance agenda as PCI compliance and data residency.
Robotics and immersive reality: operational tools
McKinsey documents robotics expanding into services, estimating an addressable robotics opportunity of close to $900 billion by 2040. Purpose-built service robots—for housekeeping runners, room-service delivery and back-of-house logistics—have moved past novelty. Immersive reality’s hospitality use-cases are narrower but targeted: AR-guided maintenance and training translate into engineering and housekeeping onboarding where experienced trainers are scarce. Neither trend reshapes guest-facing decision ownership as fundamentally as agentic AI, but both can free staff to be better hosts when integrated into a Human Experience Orchestrator approach.
What hospitality owner-operators should do next
Reading McKinsey’s thirteen trends together surfaces five simultaneous fronts for hotel groups: where you appear in agent-mediated search and booking; who retains the guest relationship once an agent makes an introduction; how you cost and govern AI at the individual-guest interaction level; how you build trust and oversight into systems that can act; and where automation truly frees staff to focus on human-only work.
None of the trends require a hotel group to become a technology company, but the report implies 2026 and 2027 are the years when the agent layer, the cost model and the governance framework will be set. Properties that act now—with a defensible AEO presence, a costed AI operating model and a governance framework that can survive an incident—are positioned to be the agents’ default recommendations in 2028. The remainder will face the harder task of negotiating visibility inside channels others have already built.
