Blog ·
Hospitality and Events: AI trends to focus on - AI agents transacting real bookings
AI voice agents now handle real guest calls and bookings in many languages, but voice scams are also rising. Inventory is being watched and bought by automated agents, not just humans. Ensure your pricing and availability are machine-readable or you’ll lose bookings.
Jeroen Erne
This week hospitality AI stopped being a demo and started completing real transactions—phone calls, reservations, food orders, and checkouts. The shift isn't about better chatbots. It's about agents that can book, buy, and occasionally mess up in ways that need a human to fix fast.
What changed this week
Voice agents reached a new level of capability and risk. ElevenLabs released v4 with expression control across 90 languages, while Inception launched Mercury Voice for low-latency agent conversations. These tools can now handle multilingual front-desk calls and reservation lines. But the same week, a deepfake voice scam targeting a grandparent spurred a new identity-verification startup, and Truecaller extended its scam detection beyond caller ID to the open web. For hospitality, the message is clear: voice agents are ready for guest-facing work, but so are the impersonators.
Transactional AI agents became the default, not the exception. Instinct, an AI agent platform, raised $1 billion at a $10 billion valuation, with travel accounting for more than half of its transactions. Shopify opened its checkout to browser-based AI agents, meaning a guest's personal AI can now complete a hotel booking or event ticket purchase without human involvement. DoorDash launched a text-based ordering agent. Airbnb added AI search and social trip-planning features. The common thread: hospitality inventory is being discovered and purchased by machines, not just people browsing websites.
Agent persistence changed the guest relationship. Manus 2.0 introduced persistent computers and event-triggered agents that can monitor availability and act on a guest's behalf over days or weeks. OpenAI launched Dots, always-on agentic avatars that maintain context across interactions. xAI released Team Bots for shared enterprise workflows. For hotels and venues, this means a booking inquiry might arrive from an agent that has been watching your inventory for a month and will complete the transaction the moment a room opens up.
Content production tools became faster and cheaper in ways that directly affect hospitality marketing. fal's Seedance 2.0 added 1080p and 4K video output, plus Pixelcut looping video for product animations. HeyGen launched a low-cost prompt-to-video API for business content. Tavus introduced Griffin for real-time face-to-face AI interaction. A venue can now generate walkthrough videos, menu animations, and personalized host greetings with a few prompts—but brand consistency and licensing need attention before publishing.
What it means for you
Your inventory is being watched by machines you didn't invite. Travel-focused AI agents are polling booking systems, checking availability, and completing transactions without a human ever seeing your website. If your pricing, room descriptions, or event capacity aren't cleanly exposed in a machine-readable format, those agents will either ignore you or, worse, book incorrectly and create a service recovery situation.
Voice is now a frontline channel you need to manage. When an ElevenLabs-powered agent calls your front desk in perfect Japanese or German to book a table, your team needs to know how to verify identity, confirm payment intent, and escalate to a human if the agent mishears a date or overbooks a section. Consent matters too—guests should know when they're talking to your AI and when they're talking to a person.
Personalization is colliding with privacy. Instinct's proactive product recommendations triggered user discomfort this week—the "ick" factor of an agent that knows too much. For hospitality, the line between helpful ("I noticed you prefer corner rooms") and invasive ("I see you searched for anniversary venues") is thin. Keep personalization tied to explicit transaction approval, not inferred intent.
Content creation is now fast enough to match the pace of events. You can generate a venue video, a menu animation, or a personalized booking confirmation in minutes. But the speed creates a new risk: publishing un-reviewed AI output that misrepresents your property or uses unlicensed music. Build a review step into every AI content workflow, and keep channel attribution clear so you know which content drives bookings.
What to focus on next week
- Audit one booking channel for machine readability. Check whether your room types, rates, and availability are exposed in a way an AI agent can parse without errors. Fix missing fields or ambiguous descriptions that could cause misbookings.
- Draft a voice-agent escalation protocol. Write down the steps your front-desk team should follow when an AI agent calls to book: what to confirm, when to ask for a human, and how to log the interaction for service recovery.
- Test one AI content tool against your brand standards. Generate a short video or menu image using HeyGen, fal, or a similar tool, then run it through your usual approval process. Note where the tool saves time and where it needs human correction.
- Review your privacy language around personalization. If you use guest data to make recommendations, check that your terms are clear about what's inferred versus what's explicitly shared—and give guests an easy way to opt out.
- Watch your booking logs for agent traffic patterns. Look for repeated queries from the same source, rapid availability checks, or transactions completed outside normal guest browsing hours. These may signal AI agents polling your inventory.
These stories came from daily monitoring across the week. For the full set of hospitality and events AI developments, see all Hospitality and Events AI news.