Model Context Protocol

Hello, and welcome to today's presentation on the Model Context Protocol. My name is Janette Roush. I'm the Chief AI Officer for Brand USA, and today we're going to be talking about something that I think is going to fundamentally change not just how destination marketing organizations operate, but how all of travel and tourism operates in the next 5 to 10 years.

So let me start by giving you a little bit of context. There was a report that came out from Booking.com where they surveyed travelers from around the world, and they asked them about their use of AI for travel planning. And what they found is that 91% of travelers are excited about the idea of using AI to help them plan their trips. So there's a huge appetite, huge excitement around this technology.

But then they asked a follow-up question, which is: How much do you trust AI to plan your trips? And only 6% of respondents said that they fully trust AI to plan their travel. So you've got this massive gap between the excitement and the actual trust. And the reason for that gap is because large language models—so tools like ChatGPT, Claude, Gemini—all of these tools have a fundamental problem, which is that they hallucinate. They make things up. They sound very confident when they're giving you information, but that information is often wrong, often outdated, and that creates a trust problem.

And so what I want to talk about today is a technology that is designed to solve that trust problem. And that technology is called the Model Context Protocol, or MCP.

So let me give you a little bit of background. When you're using a large language model like ChatGPT, what you're doing is you're interacting with a model that was trained on a massive corpus of text from the internet. And that training happened at a specific point in time. So for example, ChatGPT's knowledge cutoff is April 2023. That means that anything that happened after April 2023, ChatGPT doesn't know about it because it wasn't part of the training data.

And so if you ask ChatGPT a question about something that happened recently, it's going to have to either tell you "I don't know" or it's going to make something up. And oftentimes, it makes something up because it's designed to be helpful and to give you an answer even when it doesn't actually have the information.

And so the solution to this problem is to give the large language model access to a source of truth—a database, a real-time feed of information that it can query in order to give you accurate, up-to-date answers. And that's what the Model Context Protocol does.

So let me define what MCP is. The Model Context Protocol is a new open standard for connecting AI systems to the data and tools they need to produce accurate, relevant outputs. It's essentially a way for an AI to say, "I don't know the answer to this question, but I know where to find the answer. Let me go query this database, let me go check this API, and then I'll come back and give you the correct information."

And so MCP is what we call AI-native, meaning it was designed specifically for AI systems to use, unlike traditional APIs which were designed for software developers to use. And the key difference is that MCP is self-describing. The AI can look at an MCP server and understand what data is available, what tools are available, and how to use them without a human having to write custom code to integrate that API.

So let me show you a diagram that I think really helps to explain this. Before MCP, if you wanted to connect ChatGPT to a bunch of different data sources—so let's say you wanted it to be able to access your calendar, your email, your CRM, a hotel booking system, a flight booking system—you would have to build custom integrations for each one of those systems. And that's a lot of work. It requires software developers. It requires ongoing maintenance. And it's just not scalable.

After MCP, what you have is a single protocol that all of these different systems can implement. And then the AI can connect to any MCP server and automatically understand how to use it. So it's like a universal connector. It's like USB-C for AI. Instead of having to have a different cable for every device, you just have one cable that works with everything.

And so that's the power of MCP. It dramatically reduces the amount of work that's required to connect AI systems to the data and tools that they need.

Now, I want to make a really important distinction between an API and an MCP because they sound similar, but they're actually quite different. An API—Application Programming Interface—is a set of rules for how one piece of software can talk to another piece of software. And APIs have been around for decades. They're the backbone of the internet. Every time you use a website or an app, there are APIs running in the background making that possible.

But APIs were designed for software developers. They require you to read documentation, write custom code, and build integrations. And that's fine if you're a software developer, but it's not something that an AI can do on its own.

MCP, on the other hand, is AI-native. It's self-describing, which means that the AI can look at an MCP server and automatically understand what data is available, what actions it can take, and how to use those tools. So it's much more flexible, much more scalable, and it's designed specifically for AI systems.

Now, the first mainstream implementation of MCP that most people are familiar with is something called Apps in ChatGPT. So if you have a paid ChatGPT account, you can go into the settings and you can enable different apps. And these apps are essentially MCP servers that ChatGPT can connect to in order to access external data and tools.

So for example, there's an Expedia app, there's a Booking.com app, there's apps for all sorts of different services. And what these apps do is they allow ChatGPT to go out to those services, query their databases, and bring back real-time, accurate information.

So let me show you an example. If I go to ChatGPT and I enable the Booking.com app, I can now ask ChatGPT a question like, "I'm planning a trip to Paris next month. Can you help me find a hotel that's in the city center, has good reviews, and costs less than $200 a night?" And ChatGPT can go out to Booking.com, query their database, and come back with actual hotel listings that meet those criteria. It's not making anything up. It's pulling real data from Booking.com.

And so that's the power of MCP. It allows AI to be much more useful, much more accurate, and much more trustworthy because it's pulling from verified sources of truth rather than just relying on its training data.

Now, I want to talk about some specific applications for the tourism industry because I think this is where it gets really exciting.

The first application is around accessibility. So one of the big challenges with AI and travel planning is that if you ask an AI, "What's the best wheelchair-accessible route from my hotel to this museum?" the AI might make something up. It might tell you that there's a ramp at a certain intersection when in fact there's no ramp there. And that's a real problem for people who are relying on that information.

But if you connect the AI to an MCP server that has verified, up-to-date accessibility data—so data about where the ramps are, where the elevators are, where the accessible entrances are—then the AI can give you accurate, reliable information. And that's incredibly valuable for travelers who have accessibility needs.

The second application is around B2B, so business-to-business. Let's say you're a meeting planner and you're trying to find a venue for a conference. You might ask the AI, "I need a venue that can hold 500 people, has A/V equipment, is available in June, and is within budget." And the AI can query an MCP server that has a trusted database of meeting venues with all of those details, and it can come back with a list of venues that actually meet your criteria.

And so instead of the AI just guessing or making something up, it's pulling from a verified source of truth. And that makes it much more useful for professionals who are making high-stakes decisions.

The third application is around consumer bookings. So for example, if you're trying to find the best value Broadway tickets, you can ask an AI, "What are the best deals on Broadway tickets this week?" And if the AI is connected to an MCP server that has real-time pricing data from all the different ticketing platforms, it can give you accurate information about what the best deals are right now.

And so those are just three examples, but there are dozens, hundreds of potential applications for MCP in the tourism industry. Anywhere where you need accurate, real-time data, MCP is the solution.

Now, I want to talk about the strategic implications of this for destination marketing organizations. Because I think this is a fundamental shift in how DMOs need to think about their role.

Traditionally, DMOs have been in the business of creating content—creating websites, creating brochures, creating videos—and hoping that travelers will find that content and engage with it. But in a world where travelers are using AI to plan their trips, the content that you're creating on your website might never be seen by a human. Instead, the AI is going to be querying your data, pulling the information it needs, and then presenting it to the traveler in a conversational interface.

And so the role of the DMO is shifting from being a creator of content to being a steward of data. You need to make sure that your data is accurate, up-to-date, well-structured, and accessible to AI systems. Because if your data isn't accessible, then the AI is going to pull from some other source, and you're going to lose control over the narrative about your destination.

And so this is why I think MCP is so important for DMOs. It gives you a way to publish your data in a format that AI systems can easily access and use. And it gives you control over the story that's being told about your destination.

Now, I want to talk about two possible futures for the tourism industry. The first future is what I call the Open Ecosystem. In the Open Ecosystem, DMOs and other tourism organizations lead the way in publishing their data as MCP servers. They make their data freely available to AI systems. And they maintain control over the story that's being told about their destinations.

In this future, travelers benefit because they have access to accurate, trustworthy information from official sources. And DMOs benefit because they maintain their role as the authoritative source of information about their destinations.

The second future is what I call the Closed Ecosystem. In the Closed Ecosystem, third-party platforms like Expedia, Booking.com, TripAdvisor—these platforms become the dominant MCP servers for travel data. And they charge DMOs and other tourism organizations for access to travelers. In this future, DMOs lose control over their data and over the story that's being told about their destinations. And they end up paying for access to their own customers.

And so the choice that DMOs need to make right now is: Which future do we want? Do we want to lead the way in publishing our data and maintaining control? Or do we want to cede that control to third-party platforms?

And my strong recommendation is that DMOs need to move quickly to establish themselves as trusted MCP servers. Because the window of opportunity is closing. The companies that move first are going to be the ones that define the standards, that set the expectations, and that maintain control over their data.

So what are the next steps? If you're a DMO or you're working in the tourism industry, here are some things that you can start doing right now:

First, audit your data. Take a look at the data that you have about your destination. Is it accurate? Is it up-to-date? Is it well-structured? If not, you need to clean it up because AI systems are very literal. If your data says that a museum is open on Mondays and it's actually closed on Mondays, the AI is going to give travelers wrong information and they're going to have a bad experience.

Second, think about what data you want to make available. What are the most valuable pieces of information that travelers are looking for? What are the things that only you know as the official destination marketing organization? Those are the things that you should be publishing as MCP servers.

Third, start experimenting with MCP. If you have access to Claude, you can already start building MCP servers and testing them out. There are tutorials available online. Morten Rand-Hendriksen has some great LinkedIn Learning courses on this. And Anthropic has published documentation about how to build MCP servers.

And then finally, start thinking strategically about your role in this new ecosystem. How do you want to position your organization? What data do you want to control? What partnerships do you want to form? These are big strategic questions that leadership needs to be thinking about now.

So just to wrap up: The Model Context Protocol is a fundamental technological breakthrough that solves the trust problem with AI. It allows AI systems to connect to verified sources of truth so that they can give travelers accurate, reliable information. And it represents a major shift in the role of destination marketing organizations from content creators to data stewards.

The organizations that move quickly to adopt MCP are going to be the ones that maintain control over their data, over their narrative, and over their relationships with travelers. And the organizations that wait are going to find themselves paying for access to their own customers.

So my encouragement to all of you is: Start learning about MCP now. Start experimenting with it. Start thinking strategically about how your organization is going to adapt to this new world. Because this is not a "someday" technology. This is happening right now. And the decisions that you make in the next 6 to 12 months are going to have a huge impact on your organization's future.

Thank you so much for your time today. I hope this was helpful. And I'm excited to see how all of you are going to use MCP to transform your destinations. Thank you.