Agentic Hotel Booking: Three Models, Three Different Futures

A traveler asks ChatGPT to find a hotel in Chicago. Another guest books a room by speaking to an AI assistant on a hotel's website. A third delegates the entire trip to a personal AI agent that could eventually negotiate and complete the reservation with another agent.

All three scenarios are increasingly discussed under the same label: agentic booking.

But they're not the same thing.

They involve different technology, different commercial relationships, and very different implications for who controls the guest, the booking, and the revenue.

For hotel operators, the question isn't simply whether AI will change hotel bookings. It's which part of the booking process AI will control — and who benefits from that control.

Based on Ira Vouk's original framework for agentic hotel booking, there are three distinct models emerging in hospitality:

• AI-assisted booking: Consumer AI platforms help travelers discover, compare, and select hotels.
• AI-mediated booking: AI agents operating within a hotel's own ecosystem help guests complete direct reservations.
• AI-executed booking: Autonomous agents communicate with other agents to arrange and potentially complete transactions on behalf of travelers.

These are not simply three stages of the same technology.

They represent different ways hospitality distribution and booking execution could operate — and they may coexist for years.

Understanding the difference matters before hotels start investing in the wrong infrastructure for the future they're trying to prepare for.

1. AI-Assisted Booking: When Consumer AI Becomes the Discovery Layer

Imagine a traveler opening ChatGPT, Gemini, or another AI-powered search interface and asking:

"Find me a four-star hotel in Chicago next Tuesday, under $300, within walking distance of the convention center."

Instead of opening multiple hotel websites or filtering through traditional search results, the traveler receives a conversational comparison of relevant properties.

The AI may evaluate location, rates, amenities, policies, and publicly available hotel information before presenting a shortlist.

The traveler then selects a property and completes the booking through a supported transaction flow, an online travel agency, or the hotel's own website.

In this model, AI changes how guests discover accommodation, but the traveler still actively participates in the booking decision.

What Changes for Hotels?

The most important shift is in distribution.

Traditionally, hotels compete for visibility through search engines, OTAs, metasearch platforms, and direct marketing.

AI-assisted discovery introduces another layer between the traveler and the property.

The guest may never review a traditional list of search results.

Instead, an AI system interprets the request and determines which hotels to present.

For independent hotels and smaller hotel groups, this creates a new commercial challenge.

It may no longer be enough to have an attractive website and competitive rates.

Hotel information also needs to be accurate, accessible, and sufficiently structured for AI-powered discovery systems to understand.

This includes details such as:

• Property location and accommodation type.
• Room categories and amenities.
• Booking restrictions and cancellation policies.
• Accurate rate and availability information where supported.
• Distinctive features relevant to traveler preferences.
• Consistent property information across distribution channels.

This is one reason AI search visibility and Generative Engine Optimization (GEO) have become relevant topics for hospitality marketers.

However, GEO alone does not guarantee that a hotel will appear in an AI recommendation.

Discovery depends on the platform, available data, retrieval mechanisms, partnerships, and how the AI system interprets a particular traveler request.

Who Controls the Guest Relationship?

In AI-assisted booking, the consumer AI platform influences discovery and comparison.

The hotel may still complete the transaction directly, or the booking may pass through an OTA or other intermediary.

That distinction matters.

If a guest discovers a property through an AI platform but books through an OTA, the hotel may still face the commercial conditions associated with intermediary distribution.

If the AI directs the guest to the hotel's own booking environment, the property has a different opportunity to manage the transaction and subsequent relationship.

The strategic risk is not AI-assisted discovery itself. It's allowing a new discovery layer to become another expensive intermediary without understanding who owns the transaction.

For hotel operators, the priority is to make the property discoverable while protecting the commercial value of direct booking channels.

2. AI-Mediated Booking: When the Hotel's Own AI Handles the Reservation

Now consider a different scenario.

A guest visits a hotel's website or calls the property directly.

Instead of navigating a booking form or waiting for a reservations employee, the guest begins a conversation with an AI agent.

They explain their travel dates, room preferences, and any additional requirements.

The AI retrieves relevant information from the hotel's connected systems, presents available options, and helps the guest complete the reservation.

This is AI-mediated booking.

The difference from AI-assisted booking is fundamental.

The AI is operating within the hotel's own booking environment rather than primarily serving as an external discovery platform.

The hotel or accommodation operator maintains responsibility for its inventory, commercial rules, booking processes, and guest service standards.

Why This Model Matters Commercially

Hotels already invest in attracting guests through direct channels.

But attracting someone to the property website or telephone number doesn't guarantee that the reservation will be completed.

The guest may have questions about room categories, booking conditions, availability, parking, or a special request.

If the process becomes inconvenient, the booking may be delayed, abandoned, or completed elsewhere.

AI-mediated booking addresses the execution side of that challenge.

With appropriate PMS or booking system integrations, an AI agent can support the steps that turn a guest inquiry into a reservation.

Depending on the integration and operating rules, those steps may include:

• Understanding requested dates and accommodation preferences.
• Checking availability and applicable rates.
• Explaining property-specific booking conditions.
• Creating a reservation in a connected system.
• Supporting permitted changes or cancellations.
• Providing the appropriate confirmation or next step.
• Escalating exceptions requiring human approval.

The commercial opportunity is to reduce friction in the hotel's own booking process.

Unlike external AI discovery, this model doesn't necessarily introduce another distribution intermediary.

Instead, it can improve how the hotel handles demand arriving through channels it already operates.

Where Una Fits Today

This is the model Polydom has implemented with Una.

Una by Polydom is a voice-first AI Digital Employee designed to work within existing hospitality technology systems.

She can communicate with guests through supported channels, including telephone, website interactions, and messaging platforms.

Through compatible integrations and configured workflows, Una can retrieve reservation information and execute authorized booking actions, including supported reservation creation, modification, and cancellation.

The key distinction is that Una is not simply providing information about how to book.

She can participate in the booking workflow itself.

The scope of execution depends on the property's PMS integration, available functionality, configured policies, and the permissions granted to Una.

Payment handling and verification requirements also depend on the connected systems and specific transaction flow.

That matters because an AI conversation is not automatically a completed reservation.

AI-mediated booking becomes commercially valuable when the conversation connects to a real booking action inside the operator's existing systems.

Learn how Una connects to hotel PMS platforms and operational workflows.

AI-Mediated Booking Is Already Producing Measurable Results

Unlike fully autonomous agent-to-agent booking, AI-mediated reservations are not merely a future concept.

They are already being executed in real hospitality operations.

At Lee Gardens Extended Stay in Orlando, Florida, Una was deployed to support guest communication and direct booking workflows.

Over a ten-month reporting period, the property recorded:

• 232 direct reservations made by Una.
• $80,968 in reported direct booking revenue from those reservations.
• 4,104 guest conversations handled.
• 80 conversations requiring human involvement, approximately 1.95% of the total.

These figures represent one property's deployment and should not be interpreted as a guaranteed result for other hotels.

They also do not establish how much of the booking revenue was incremental.

But the case demonstrates something important about the current state of AI-mediated booking.

An AI Digital Employee can participate in real reservation execution rather than simply recommending accommodation or directing guests to another booking channel.

Read the Lee Gardens case study: $81K in Direct Bookings.

For hotel operators, this makes AI-mediated booking the most immediately actionable of the three models discussed here.

It offers a practical way to improve direct booking execution using existing hospitality systems.

3. AI-Executed Booking: When Agents Book With Other Agents

The third model is the most ambitious.

Imagine a business traveler whose personal AI assistant has access to their calendar, approved travel preferences, and booking budget.

A meeting is scheduled in another city.

Instead of asking the traveler to search for a hotel, the personal agent identifies the travel requirement.

It contacts a hotel's AI agent, or an intermediary representing available properties.

The agents exchange information about availability, rates, accommodation preferences, loyalty eligibility, booking conditions, and payment requirements.

If the traveler has provided appropriate authorization and the necessary transaction infrastructure is available, the agents could complete the reservation.

The traveler receives confirmation.

No manual hotel search. No traditional booking form. Potentially no direct conversation between the guest and the hotel's reservations team.

This is AI-executed, or agent-to-agent (A2A), booking.

Here, AI doesn't simply assist the traveler or mediate a conversation. It acts on delegated authority to complete the transaction.

What Needs to Exist for Agent-to-Agent Booking?

Agent-to-agent commerce requires more than two language models exchanging messages.

A reliable booking involves financial, contractual, and operational responsibilities.

Several challenges need to be addressed before autonomous hotel reservations can operate broadly across different providers.

1. Verifiable identity

A hotel needs a reliable way to establish who the traveler is, which agent represents them, and whether that agent is authorized to act.

2. Delegated payment authority

An AI agent must have access to an approved payment mechanism with appropriate limits and controls.

The system needs to distinguish between recommending a booking and being authorized to commit funds.

3. Machine-readable hotel policies

Cancellation rules, deposits, occupancy restrictions, rate conditions, and other terms need to be represented in ways participating systems can interpret consistently.

4. Interoperable booking infrastructure

Agents need compatible mechanisms for exchanging availability, prices, reservation details, and transaction status.

5. Authentication, accountability, and dispute handling

Operators need clear processes for validating instructions, recording authorization, handling failed transactions, and resolving errors.

Industry initiatives are developing different parts of this infrastructure.

But the availability of individual protocols or demonstrations should not be confused with universal interoperability across hotels, AI platforms, payment providers, and booking systems.

For a broader explanation of the three models, see Ira Vouk's analysis of agentic hotel bookings.

Why Agent-to-Agent Booking Is a Distribution Question, Not Just a Technology Question

It may be tempting to see agent-to-agent booking as a simple extension of today's AI-assisted reservations.

But the commercial consequences could be considerably larger.

In the AI-assisted model, a consumer platform influences which hotels a traveler discovers.

In the AI-mediated model, the hotel's own AI helps complete a booking within its existing ecosystem.

In an agent-to-agent model, software acting on behalf of the traveler could negotiate directly with software representing hotels or intermediaries.

That raises important questions.

Who determines which hotels the traveler's agent considers?

Will hotels interact directly with personal agents, or will intermediaries continue to control access to inventory?

How will commissions work?

Who is responsible when an agent books the wrong room, applies the wrong conditions, or makes a reservation without sufficient authorization?

What happens when a hotel needs to make an exception after the automated transaction has been completed?

These are not minor technical details.

They determine who controls distribution, who accepts transaction risk, and who owns the commercial relationship with the guest.

For independent hotels and professional vacation rental operators, the danger is preparing for autonomous booking without first understanding which parts of the distribution relationship they want to control.

Where Una Fits in an Agent-to-Agent Future

Una already operates within the second model: AI-mediated booking.

She works with guests directly through supported communication channels and executes authorized actions inside connected hospitality systems.

That operational foundation is relevant to the third model.

An AI system capable of receiving a guest request, checking available inventory, applying booking rules, and performing a reservation action has some of the capabilities that would also be useful when communicating with another AI agent.

The difference is who initiates the request and how authority is established.

Today, Una can interact with a human guest through a supported channel.

In a future agent-to-agent scenario, a suitably authorized external agent might initiate a booking request instead.

However, making that transition would require additional work.

Agent identity, delegated authority, payment controls, secure interoperability, and responsibility for failed or disputed transactions would all need to be addressed.

Una's existing booking workflows provide a foundation for exploring such integrations, but that should not be interpreted as a claim that fully autonomous cross-platform A2A hotel booking is already universally available through Una.

The important architectural idea is that the hotel needs a reliable execution layer regardless of whether the request comes from a guest or another authorized system.

For Polydom, that is the strategic relevance of building an AI Digital Employee around operational execution rather than conversation alone.

What Should Hotel Operators Prepare for Now?

The three models have different implications for investment.

Hotels should avoid treating agentic booking as one technology purchase that automatically solves discovery, direct conversion, and autonomous distribution.

A more practical approach is to evaluate readiness separately.

For AI-Assisted Booking: Make Your Property Discoverable

Start with the quality of the information available about your property.

Review your website, accommodation descriptions, booking policies, structured data where appropriate, and consistency across distribution channels.

Make sure potential guests and AI-powered discovery systems can understand what your property offers and who it serves.

The immediate objective is accurate representation and discoverability, not a promise of guaranteed AI recommendations.

For AI-Mediated Booking: Improve Direct Booking Execution

Examine what happens after a prospective guest contacts your property.

Can someone receive accurate availability information outside staffed hours?

Can the system support a real reservation action?

What happens when the guest changes their dates or requests a different rate?

Can the AI recognize when a booking requires human approval?

This is where connected operational AI can provide measurable value today.

For many independent hotels, serviced apartments, and professional STR managers, reducing friction in existing direct channels may be more immediately useful than preparing for fully autonomous consumer-agent transactions.

For AI-Executed Booking: Prepare Your Systems and Governance

Operators interested in future agent-to-agent commerce should begin by reviewing their underlying operational infrastructure.

Consider whether reservation information is accessible through reliable integrations and whether booking policies are consistently represented.

Identify which systems control inventory, pricing, payments, and guest identity.

Understand how booking actions are authorized and audited.

These preparations can improve existing operations even if broad agent-to-agent adoption takes longer than expected.

They also help operators avoid becoming dependent on a future distribution model they do not control.

The Front Desk Isn't Disappearing

All three models change how travelers may interact with hotel technology.

None removes the need for human hospitality.

Guests will continue to arrive with unusual requests, complex situations, personal preferences, and expectations that cannot always be reduced to a predefined workflow.

A hotel employee may need to resolve a complaint, accommodate a special circumstance, or make a judgment that goes beyond established booking rules.

An AI Digital Employee can support routine reservations, communication, and operational execution.

But human employees remain responsible for the relationships, exceptions, and service decisions that define the guest experience.

The purpose of agentic booking should not be to remove people from hospitality.

It should be to reduce the repetitive work required to complete transactions and coordinate routine services.

That creates more room for hotel teams to focus on the parts of hospitality that benefit from human attention.

Frequently Asked Questions About Agentic Hotel Booking

What Is Agentic Booking in Hospitality?

Agentic booking refers to the use of AI agents in hotel discovery, reservation interactions, or booking execution.

The term can describe several different models, from AI-assisted search to AI-mediated direct reservations and agent-to-agent transactions.

These models differ in how much authority the AI has and which organization controls the booking process.

What Is the Difference Between AI-Assisted and AI-Mediated Booking?

In AI-assisted booking, a consumer AI platform helps travelers find and compare accommodation.

In AI-mediated booking, an AI agent operates within the hotel's own channels and can support or execute booking actions through connected systems.

The primary difference is where the AI operates and who manages the transaction.

Is Agent-to-Agent Hotel Booking Available Today?

Elements of agentic commerce and automated booking already exist, but fully autonomous agent-to-agent hotel transactions across unrelated platforms still depend on compatible identity, authorization, payment, and interoperability systems.

Operators should distinguish specific working integrations from broader predictions about industry-wide adoption.

Can AI Agents Make Direct Hotel Bookings?

Yes. AI agents with supported PMS or booking system integrations can execute authorized reservation workflows.

For example, Una made 232 direct reservations during the reported Lee Gardens deployment.

Available actions depend on the integration and configuration.

How Can Independent Hotels Prepare for Agentic Booking?

Independent hotels can begin by improving the accuracy and accessibility of property information, strengthening direct booking workflows, reviewing PMS integrations, and establishing clear controls over automated actions.

They do not need to adopt every emerging agentic commerce technology at once.

Three Futures, Three Different Decisions

Agentic booking isn't one inevitable future arriving all at once.

It's three different models that may develop alongside one another.

AI-assisted booking changes how travelers discover and compare properties.

AI-mediated booking changes how hotels turn their own guest inquiries into completed reservations.

AI-executed booking could eventually change how authorized software agents negotiate and transact with hospitality systems.

Each model shifts control differently.

For hotel operators, that's the real strategic question.

Not whether AI will participate in booking.

But whether the hotel's business is prepared for the specific role AI will play — and whether the property retains appropriate control over its inventory, policies, revenue, and guest relationships.

At Polydom, we're building Una around the part of that future hospitality operators can already put to work: connecting guest requests with actual operational execution.

Because before hotels can transact reliably with someone else's AI agent, they need their own systems to reliably execute the work behind a reservation.

See How AI-Mediated Booking Works with Una

Curious how Una handles direct booking conversations and reservation workflows inside existing hospitality systems?

Talk to our human sales team, explore live customer use cases, and see which booking processes Una can support at your property.

Book a quick call here to discuss AI-mediated booking, PMS integrations, and your property's readiness for the next generation of hospitality automation.