“Agentic booking” has become one of those phrases everyone uses and few define. It sounds futuristic and inevitable — but when you listen closely, people are often talking about completely different models.

If we don’t distinguish them, we’ll build the wrong infrastructure for the wrong future.

Based on Ira Vouk, MBA’s original framework, there are three very different interpretations of “agentic hotel booking,” and they are not minor variations:

AI-Assisted Booking – Inside a Consumer AI Platform

Here the traveler starts in a consumer AI interface (ChatGPT, Gemini, etc.) and asks for a hotel that matches their needs.

The AI handles search, comparison, and recommendations.
Booking is completed either inside the AI’s UI or via redirect to OTA.
Payment rails are mostly traditional.
Identity is explicit and controlled by the human user.

The key: this is still human-led execution with AI-enhanced discovery. Power shifts to consumer AI platforms that own the discovery layer.

AI-Mediated Booking – Inside the Hotel’s Own Ecosystem

Here the traveler interacts with an AI on the hotel’s own channels (website, app, or direct messaging).

The guest talks to an embedded AI agent that understands dates, preferences, loyalty, and constraints.
The agent connects to the hotel’s CRS / PMS / booking engine.
Payment, identity verification, and data live inside the brand ecosystem.
The hotel fully controls merchandising logic, offers, and rules.

This is not the same as asking a generic AI to find a room. Power stays inside the hotel (or vacation rental) ecosystem.

This is what I’d call AI-mediated booking: AI is the interface, but the transaction and data remain within the hotel/VR domain.

AI-Executed Booking – Agent-to-Agent (A2A)

This is where the process becomes truly “agentic.”

A user’s personal AI agent detects a travel need (e.g., a meeting appears in the calendar).
That agent talks directly to a hotel’s AI agent (or an intermediary’s agent).
The agents autonomously handle availability, pricing, loyalty, policies, payment, and identity verification.
The human simply receives a confirmation — no search, no clicking, no chat.

To make this work at scale, you need:

Machine-verifiable, portable identity
Delegated payment credentials for agents
Encoded preferences and policies up front
Interoperable APIs and standards for negotiation
Robust trust and authentication frameworks

Power shifts to autonomous digital agents executing commerce on behalf of humans.

Where Una Fits Today – And What We Built For Tomorrow

On our side at Polydom.ai, we’ve already implemented model #2 (AI-mediated booking) in production:

Una operates as a digital employee for hotels and vacation rentals.
She talks to guests across channels (phone, web, WhatsApp, kiosk, etc.).
She doesn’t just “chat” — she executes bookings, modifications, cancellations, and even check-in/check-out flows directly in the PMS and task stack.

In other words, we’ve built the infrastructure where the hotel / VR operator keeps control of:

Data
Payment flow
Inventory & policies
Brand experience

At the same time, our architecture is designed for #3 (AI-executed / A2A) — the moment when:

agent speaks to agent

…instead of guest → agent.

The only missing piece at industry level is the guest-side / intermediary agent layer and the standards around identity, payments, and trust. Once those mature, the same operational rails we use for Una today can support full A2A booking.

Why I Think A2A Has Quietly Already Started

Every one of our clients says some version of this:

Once you experience true end-to-end automation, there’s no going back to humans manually processing routine bookings and changes.

That’s the foundation of A2A:

Agents already execute end-to-end workflows.
The next step is simply who they’re talking to: a human or another agent.

From there, scaling agent-to-agent booking becomes an architecture and standards problem — not a science fiction problem.

And No, the Front Desk Isn’t Going Away

I don’t believe the front desk (or guest-facing roles in vacation rentals) will disappear.

I believe those roles finally become what hospitality was always meant to be:

Welcoming guests
Handling nuances and exceptions
Solving complex situations
Creating human connection

Meanwhile, digital employees like Una handle the repetitive operational work: data entry, PMS operations, confirmations, payment flows, and all the “clicking around” that machines can do faster and more consistently.

Humans stay in the loop — but on the relationship and experience side, while agents quietly run the operational engine.

If we want real progress, we should stop using “agentic booking” as a single buzzword and start asking:

Which model am I actually preparing for —AI-assisted, AI-mediated, or AI-executed (A2A)?

And is my tech stack — and my org design — ready for that specific future?