The 6 pieces of the AI-powered direct selling ecosystem

However, amidst the widespread enthusiasm, most hotels continue to grapple with the same question: how to turn AI into a real, sustainable, and business-aligned competitive advantage? And, even more pressing, where to begin?

By now it's clear that AI is not an isolated product or just another plugin. It's a cross-cutting layer of intelligence that only generates value when it's supported by a solid architecture, reliable and structured data, and integrated with all hotel systems.

A realistic and applied view of artificial intelligence
The main risk of accelerated AI adoption is the fragmented approach: disconnected chatbots, inconsistent responses across channels, duplicate data, and assistants that seem intelligent but cannot perform real actions.

AI only adds value when:

It relies on a single source of truth.
It understands the context of the hotel business.
It can act, not just react.
It integrates naturally into existing workflows. 
That's why at Mirai we say that AI adoption is an evolution of the ecosystem, not a replacement of its traditional channels.

The reinvention of the direct sales ecosystem:
The three historical pieces of the direct channel will continue to play a key role in the coming years, although their relative weight will change with the emergence of AI.

1. Web

The main channel for inspiration and bookings. It will remain fundamental, although it will gradually share the spotlight with conversational interfaces like WhatsApp, ChatGPT, or Gemini.

2. Telephone or contact center

Human attention, resolution of complex queries, and closing high-value sales. A critical channel during moments of decision or incident. This channel will undergo a profound transformation thanks to AI, driven by immediate access to reliable and contextual information, as well as the emergence of AI-powered voice assistants.

3. Mobile application

Pre-sales, relationship during stay and loyalty: check-in, services, upgrades and direct communication with the guest.

The arrival of AI does not eliminate these pieces, but expands the ecosystem with three new strategic layers.

4. AI agents serving the direct channel

They allow hotels to offer conversational interfaces to their customers in multiple languages ​​and formats: web chatbot, WhatsApp, Instagram, SMS or even voice.

These agents coexist with the web and the contact center, but they also open new doors of interaction for the customer, especially in channels with high adoption and far removed from the traditional web such as WhatsApp.

At Mirai, this piece of the ecosystem is embodied by Sarai, our conversational agent solution, which answers all of a guest's questions about the hotel, provides quotes, integrates the loyalty program, and even completes bookings without leaving the assistant or being redirected to the website. And it always guarantees that the content, prices, and conditions are consistent with a traditional online booking.

5. MCP Server: The bridge to AI assistants

One piece destined to be strategic in the new ecosystem is the MCP Server -Model Context Protocol-, which acts as a bridge between the hotel and the major AI assistants such as ChatGPT or Gemini, as well as the thousands of native AI applications that will be released in a few years.

Its role is threefold, and that's how we've approached and implemented the MCP server in Mirai.

MCP provides structured and reliable information to attendees as the preferred context for increasing their visibility in the search results they offer to their clients. In practice, MCP could become a great complement to AI SEO/GEO, not in the traditional sense of algorithmic optimization to reach LLMs, but as an alternative to becoming a direct, reliable, and actionable source for AI attendees. It's important to emphasize that MCP doesn't replace SEO or GEO; rather, it introduces a new form of visibility: the battle isn't about competing for links, rankings, or citations, but about being the source attendees choose to consult. The goal, however, remains the same: to influence attendees in what they show their users, which is the ultimate desire of all hotels.

MCP does not guarantee visibility in attendees, but it does prepare the hotel to be searchable and actionable when these channels become consolidated.

MCP enables AI assistants to become truly agentic, meaning they are not limited to answering questions but can perform actual business actions, such as completing hotel bookings.
For an assistant to reliably and consistently offer bookings, the MCP server must have access to the hotel's business rules, rates, availability, offers, and loyalty logic. In practice, this means that AI assistants must operate under the exact same business conditions as the hotel's direct booking engine.

This is a critical point: AI assistants should not be treated as a separate channel with simplified rules or disconnected pricing. If the goal is to allow guests to book through conversational interfaces—whether via ChatGPT, Gemini, or in-house agents—the underlying system must always be the same booking engine, simply exposed through a different interface.

In this context, MCP acts as the orchestration layer that makes agentic commerce possible, securely connecting AI assistants with the hotel's core business logic, within a framework of permissions, control, and traceability defined by the hotel itself.

As new agent commerce frameworks like Google's recently announced Universal Commerce Protocol (UCP) begin to emerge, this separation of responsibilities becomes even more relevant. UCP should not be understood as a new assistant, but rather as a commerce runtime or operating system—a standardized environment where discovery, decision-making, and transaction occur within a single ecosystem.

In this model, the hotel does not control the booking flow or the user experience, as these are defined by the platform. What it does control—and must provide—are its agent capabilities: availability, pricing, business rules, and the ability to reliably execute bookings.

It is precisely in this context that a layer like the MCP Server takes on a critical role. Whether the flow is led by the hotel itself (as with assistants like ChatGPT, Claude, or in-house agents) or by an external ecosystem (such as Google's UCP), hotels need a single, well-defined way to expose their business logic and transactional capabilities. MCP allows this to be done once and remains compatible with multiple assistants and commerce runtimes, without having to rebuild the system core for each new scenario.

Under this approach, hotels don't integrate separately with each new platform or assistant. Instead, they expose their data and transactional capabilities only once, allowing different AI ecosystems to consume and orchestrate them according to their own frameworks, without sacrificing control, consistency, or ownership of the direct channel.

MCP allows you to create apps on marketplaces like the one ChatGPT just launched. Conceptually similar to having your app on the Apple or Android marketplace, hotels will be able to make their app available on various AI assistants.

In that context, an MCP server is the way to expose business capabilities to these apps in order to:

– Not relying on one-to-one agreements.

– To be able to “exist” within the assistants where the user already makes decisions.

– To compete on equal technological terms with OTAs and major “players”.

6. Canonical database: the foundation of the entire system

Neither agents nor the MCP server have any value without access to a master or canonical database, structured and ready for consumption by artificial intelligence. This is the hotel industry's Achilles' heel. In an environment increasingly dominated by AI assistants, those who don't provide structured and actionable data simply become invisible. Without structured, high-quality data, there are no correct answers. Paradoxically, OTAs do have these structures and occupy this space that naturally belongs to hotels. This is no small matter. This ecosystem isn't built all at once; it's deployed in phases, starting with the canonical database. That's the starting point.

To create this database, the website ceases to be the central repository of knowledge, and the challenge lies in compiling all the relevant hotel information, which is currently distributed across multiple tools and, in many cases, simply scattered among individuals and departments. The website will remain essential; what changes is where the knowledge resides to ensure consistency across all channels.

At Mirai, we address this challenge with our Intelligence solution, a text-based database where hotels can store knowledge without prior formatting. From there, Mirai structures, contextualizes, and scalably accesses this information for AI assistants, ensuring a single source of truth.

A unified ecosystem, not isolated tools.
It's a mistake to approach AI by adding more technological layers as a patch. The solution lies in orchestrating a coherent direct sales ecosystem, where each piece reinforces the others and the hotel offers a consistent response regardless of the channel.

In contrast to fragmented models based on multiple providers, complex integrations, and duplicate content, AI allows for the construction of a native, cross-cutting, and intelligent ecosystem.

Artificial intelligence will mark a turning point in the hotel industry. In this new landscape, hotels won't compete for the best chatbot, but rather for the most cohesive ecosystem, seamlessly integrated with the business and ready to be utilized by AI. That, at least, is our vision at Mirai, and how we've designed and implemented it to serve our clients.

Source: Mirai.