What is live now

The news cycle is steady. In late August, Google added agent style steps to AI Mode that help users complete tasks like dinner reservations through partners such as OpenTable, Resy and Tock. You get real time slots, then a handoff to finish on the partner site. This is small, but it proves the pattern. The agent gathers intent, checks live data and moves the user forward. 

OpenTable brought that spirit inside its own product. Its AI Concierge sits on a restaurant page, answers questions in seconds, then points you to book. It draws from the venue’s own menus, reviews and descriptions. If those are clean, the answers are useful. If they are vague, they are not. 

On the travel side, Expedia has an agent entry point on the Hotels dot com app that blends discovery, shopping and servicing. Booking continues to add AI features like review summaries and property Q and A to make decisions faster. Trip dot com calls AI a growth engine and is scaling its trip planner. Airbnb’s founder says the app will become agent driven over the next cycle. None of this is hype. It is the next user interface. 

Microsoft is also pushing the boundary. Copilot Actions can navigate partner sites on your behalf, including Booking, Expedia and OpenTable. That means a mainstream assistant can already move through the last mile of a booking flow. 

What this means for operations

Discovery compresses. The short list forms before a guest lands on your site. Agents favour clear facts over glossy copy. If your description does not state the essentials, you will not be suggested. 

Inventory becomes an API story. Agents read what they can reach. If your availability sits only in one closed channel, you may be invisible when an agent queries another. Coverage across the platforms your market uses increases your odds of being shown. Google’s first agent steps already rely on partner integrations. 

Policies and fees must be machine readable. Cancellation, deposits, child rules, pet rules, spa access, late check out and resort fees. If these live as a PDF or a wall of text, agents cannot summarise them well. Make them simple and structured.

Your tone still matters. People ask for quiet rooms, soft lighting, gluten safe menus and no loud music after ten. OpenTable’s assistant repeats what you say about noise level, diet and seating. Use precise language a model can quote. 

Service moves earlier. Agents will answer many first questions. That shifts your team to exceptions and higher value moments. Design a clean handover when a human is needed. Microsoft’s approach shows this blend very clearly. 

How to get started?

1) Rewrite the one-paragraph description on your Google Business Profile and on each booking platform. State location, style, two or three signatures, and the top practical details. No filler. These fields seed both AI Mode and partner assistants. 

2) Clean your menus and facilities lists. For restaurants, publish a current menu with clear sections and diet markers. For hotels and clinics, list facilities, access rules and hours as simple fields. These are the facts agents quote. 

3) Make rules legible. Publish a short page with deposits, cancellation, child and pet policy in plain language. Link it from every profile. Agents will lift lines from here.

4) Expand inventory reach where it makes sense. If you only expose availability through one partner, consider a second where demand is strong. Google’s early agent tasks use multiple partners. Do not let your slots go unseen. 

5) Test prompts yourself. Ask for a stay like your offer in your city. Try an agent that exists today, for example Google’s AI Mode for dinner, OpenTable Concierge for questions, or Expedia’s agent entry point. Note what appears, what is missing and how your listing reads. Fix gaps and repeat monthly. 

6) Prepare handovers. Write a short script for when the agent hands a case to a person. Confirm what the agent captured and what is still needed. Guests must feel continuity.

7) Measure agent-era metrics. Track how many bookings start from agent surfaces. Watch reduction in first line questions. Sample answers from assistants to check accuracy and tone.

Risks and how to manage them

Intermediation risk. Replacing OTA dependence with agent dependence would be a step sideways. Protect your direct path by keeping your profiles current and your brand voice distinct. Move repeat guests to your own channels with clear reasons to book direct. The major players acknowledge this tension. They are investing to keep the user inside their ecosystem. You need to earn your place in the conversation. 

Answer quality. If an assistant says a rule that is not true, guests will arrive annoyed. Reduce the chance of bad summaries by keeping your facts consistent everywhere. OpenTable stresses that its assistant pulls from your data. So does Google. Feed it well. 

Privacy and consent. When agents pre fill details, guests should see what is shared and why. Keep consent language simple and easy to find. People forgive very little here.

The bottom line

AI travel agents are not replacing people. They are removing friction and moving decisions earlier in the journey. The winners will be the operators who publish clear facts, keep availability visible where agents look and write like a real host. You will then appear in more short lists and receive better prepared guests. This is brand and UX work, not only IT work, and the time to do it is now.

Photos:

https://blog.google/products/search/ai-mode-agentic-personalized/

https://www.opentable.com/blog/concierge-ai-dining-assistant

https://www.expediagroup.com/investors/news-and-events/financial-releases/news/news-details/2025/EXPEDIA-GROUP-EXPANDS-B2B-PLATFORM-AND-LAUNCHES-GENAI-PARTNERSHIPS-TO-ENHANCE-TRAVEL-DISCOVERY/

https://www.theverge.com/news/643276/microsoft-copilot-ai-actions-feature-availability

https://news.booking.com/bookingcom-enhances-travel-planning-with-new-ai-powered-features--for-easier-smarter-decisions/

https://www.opentable.com/blog/visa-dining-collection-opentable-icons/