Curating trust for an AI future: an interview with Janette Roush on the evolving roles of NTOs and DMOs

Janette Roush: I’m Janette Roush, Senior Vice President of Innovation and the Chief AI Officer for Brand USA. I’ve been in this role for about a year and a half, and I see my work as falling into three areas. The first is internal upskilling: how we use AI at Brand USA to do our jobs more effectively. The second is taking what we’re learning and sharing it across the US tourism industry so that we remain competitive. The third is exploring how we can use AI tools to make the United States more bookable and discoverable. That includes initiatives such as generative engine optimization, but also the AI integration on our website. Through our Mindtrip itinerary builder, there are more than 600 entry points across the website that invite visitors to explore. If you’re interested in road trips through national parks, you can click through and immediately enter a chat conversation that’s already populated with that context.

I came into this without a technical background at all. I was a theatre major and spent 20 years on Broadway working in ticket pricing and advertising, including eight years as Head of Insights at an advertising agency, trying to connect the dots between the media we were running and the tickets a show needed to sell: at what price points, on which days of the week, in which parts of the theatre and at what times of the year. The goal was always to direct media spending more effectively so that shows could run longer.

When ChatGPT came out at the end of 2022, my brother sent me a link to it the day it was released. This wasn’t a world I had been following closely. At the time, I was working in marketing for New York City Tourism, so I wasn’t even in the more technical role I had previously occupied on Broadway. The first prompt I entered was: “Write me an ad for New York City from the perspective of a pigeon.” It did it. My immediate reaction was that I’d never seen anything behave this way before. Throughout December 2022, I kept searching online, thinking: this is clearly miraculous. Who’s using it? Who can teach me what to do with it? By January, I realized that very few people were seeing what might be possible with this technology. So, I made a commitment: every day that month, I would use it for something new at work and write about what I learned and whether it worked or not on LinkedIn. By the end of January, I had already received two keynote invitations. More importantly, I had developed a much clearer understanding of what this technology was, how it worked, what it could do and where it might be heading.

I think my Broadway experience helped me recognize its potential. The earliest versions couldn’t have helped me much in my day-to-day work, but I could see that this was a tool designed to find patterns in data. That’s exactly what I had been doing: looking at sales patterns to understand what a show’s future performance might be. That realization was really the beginning of my obsession with AI.

Janette Roush: I think it’s time to start thinking about AI not as a chatbot, but as your personal operating system. It’s difficult for me to call this a weak signal because, from my perspective, it’s already a very strong one. But not everyone is there yet. Most people still think of AI as something separate from their work: a chat invitation in the corner of a website, or ChatGPT, as a place you go to ask a question, rewrite something and then return to what you were doing.

What people who are following these developments more closely are beginning to realize is that AI is becoming embedded in the way work itself gets done. For example, when I sit down at my computer, I open Claude. Everything I do is saved in markdown files on my computer, organized in a way that Claude can understand. After a conversation like this one, I can use a command that prompts Claude to create files about the person and organization I’ve spoken with. The next time I interact with someone from your university, it can recall what we discussed previously.

It also captures new ideas that emerge in conversation and next time I am working on a keynote, it adds them to my speaking points, so they aren’t lost. If I commit to doing something, it identifies that as a task, records what’s involved and when it’s due and incorporates it into my workflow.

In other words, the system becomes smarter over time because it accumulates context about how I work. I genuinely believe that within three years, this is how everyone will be working. AI will make it increasingly easy to operate this way without requiring the kind of setup that I had to do to get there. Once you’ve experienced it, it’s very difficult to imagine going back to working any other way.

Janette Roush: I think one assumption that may become obsolete is the idea that destination professionals should spend so much of their time collecting and compiling information. Take measurement and reporting roles as an example. Traditionally, someone might look at Google Analytics, call OpenTable to obtain reservation data, pull information from multiple systems, put everything into a spreadsheet, build pivot tables and create graphs before anyone can begin making decisions.

Increasingly, AI can take over much of that process. Through data connectors and integrations, these systems can automatically gather information from different sources and generate reports. For example, you could simply ask for a daily update on how Restaurant Week is performing and receive a synthesized overview without manually assembling the data yourself. The real shift is that people no longer have to spend most of their time collecting information. Instead, they can focus on what the information means and how to improve the program. It creates an opportunity to move to the next stage of the work, which is often the more interesting part.

Another assumption we may need to rethink is that it’s enough to simply put information out into the world and assume that whatever happens next is good enough. A big part of what we’ll be doing moving forward is coordinating and organizing information so that it’s easily ingestible by AI agents, or, in the shorter term, easier to be understood by Google AI Overviews or ChatGPT.

I’ve been exploring this idea as what I call “destination as a service.” I think the provenance of information and content is going to become more important to users in the future. The internet was always full of slop. It’s becoming even more saturated with additional AI-generated slop. If people, or their agents, are getting help planning a trip, I don’t think it’ll just be a case of, “Whatever is in the training data of the language model is good enough,” or “Whatever sources it pulls in are good enough.” I think people might want to say, “Actually, I want to pick some sources.”

My hope is that one of those sources would be the National Tourism Organization. I’m not a technical person. I have that Broadway background. So, I’ve been trying to work through what this means in practice. How do we organize the content we create so that, if a user says, “I want DMO content to be the primary information source for this search,” it’s easy for an AI agent to find it? An AI agent could come to visittheusa.com, recognize that someone is looking for family excursions in Brooklyn and route them to New York City Tourism and Conventions as the official source. When the response is given back to the user, it includes the source. That creates trust for the user.

Janette Roush: NTOs and DMOs occupy a unique position because they sit at the intersection of government, residents, businesses that benefit from tourism and visitors. As AI evolves, our work will evolve as well, and the question we need to keep asking is whether the activities we undertake are truly worthy of our time and attention.

I also think part of our role may shift towards validating information. A few years ago, someone in Scotland promoted a Willy Wonka-themed immersive experience using AI-generated images. The event itself existed, but it looked nothing like what had been advertised. In the future, perhaps part of what DMOs do is help verify that the experiences people discover through TikTok or language models are genuine and accurately represented.

At the same time, I’m excited about the potential of smart cities and the ways information might be connected at the governmental level. Imagine a visitor standing in Times Square asking ChatGPT, “What should I do next?” Without the visitor realizing it, the system could draw on multiple sources of information: official destination content from New York City Tourism and Conventions, weather data, subway information, foot-traffic camera data from Times Square in Lower Manhattan and even real-time information from attractions such as the Empire State Building. It might determine that it’s cloudy now but expected to clear later in the day. Based on current visitor flows, it could recommend starting with the Staten Island Ferry and views of the Statue of Liberty because that area is less crowded than usual. It could then suggest the best subway route to the Empire State Building later in the afternoon, when conditions are ideal and wait times are likely to be shorter. The integration of these different information streams could give DMOs a much greater role in encouraging visitor dispersion.

Janette Roush: I think branding is going to be really important in an AI future because, again, I’m betting that the provenance of information is going to be important. That means either the AI models themselves need to think, “I should select a DMO for this search because they’re the official source,” or because, reliably, they’re providing content that people are satisfied with. A city DMO does not want every hotel to determine not just its own brand, but the brand for the entire destination. DMOs create that umbrella for everybody to sit underneath. Nobody else really sits in that spot. It is a unique role that DMOs get to play.

In the age of AI, there’s a fear that everything will get homogenized. People say, “Well, if I put the same prompt into ChatGPT as you, we’re going to get the identical response, and we’re all going to be told to do the same three things.” I don’t think that’s true. AI is increasingly going to have more context and know how to give different responses to different people. But the thing that keeps everything from sounding exactly the same is our brands. DMOs are in the perfect position to create, curate and capitalize on that.

Janette Roush: I think having those data coordinators is going to become really important. At the same time, one of our strengths is that we have a lot of people who genuinely love working with people, and I think that’s only going to become more important.

If you think about a membership team at a DMO, part of their job is keeping the CRM updated. If a venue has accessibility information, for example, someone needs to log into the CRM and update it so that it appears on the website. I don’t think that part of the job is going to be helpful in the future because that information is becoming commoditized. Google already owns things like places, addresses and opening hours. Why are we all duplicating that same work? Instead, you could tie a connector to Google and have that information flow automatically. Then people could spend that time doing the work that really matters: being out with restaurants and tourism businesses and supporting them directly.

I think AI training is going to play an important role for DMOs in the future. New York City Tourism has a program called Tourism Ready, which helps small businesses understand how tourism works and what they need to do in order to be visible to tourists. How do FIT sales work? What does it mean if you want to work with groups? What do you need to have in place? I can imagine extending that to AI. How do you optimize content on your website for generative engine optimization, not just search engine optimization? As a small business, how can you use AI in your operations to keep your overhead down and stay more competitive so that more of these places can stay open? Because that’s what makes our destinations interesting and unique.

So, I think it’s this simultaneous pivot. You need people who know how to work with data and keep it organized. At the same time, you need people who love hospitality, love working with people and can lean into that, because the machines can’t do that.

Janette Roush: I think we can all invent the future at this moment, which is what is so exciting. So, I don’t have any particular bets on whether OTAs are all going to go out of business or exactly how all of this is going to work.

I think it’s going to come down to brand. If I’m planning a trip, I’m not asking Claude to simply find a hotel in a destination. What I want is for Claude to connect to my Marriott Bonvoy account and see how many points I have, which hotel offers the best value for those points, who’s coming on the trip with me, what kind of room I need and where that room needs to be located.

But there are other types of searches. Take rental cars: I have no allegiance to any rental agency whatsoever. If the cheapest and most convenient option comes back through Expedia, great. But if there were a brand that was strong enough in the car rental universe that I cared enough to be loyal to, that would change the equation.

I think branding is what’s going to make the difference for many of these organizations in the future. If you don’t have a strong brand for a particular consumer at a particular decision point in their journey, then everything is going to be reduced to price and convenience.

Janette Roush: My bet is towards decentralization, just because of the way I’m seeing myself work. Claude is what I’m using right now, but none of my information is actually in Claude. All of my information lives in markdown files on my computer. If a competitor came along with a harness that worked better, I would simply point that harness at my files instead of Claude. That’s a danger for all of these companies because, if language models become commoditized, they’re all competing on price, efficiency and how well they work, rather than relying on brand lock-in. I think that definitely creates a threat for Google.

But I also know that there are more Google searches happening right now than ever before. Even as people can search in other places and use language models for that, the need to search is increasing because people are now having these rich conversations with Google. It’s not just a matter of putting in a few keywords and seeing what gets spit back out. So, there are huge opportunities for everybody right now, which is what makes this moment so interesting.

Janette Roush: I would resist using AI for content generation. Certainly, at the moment, it’s too smooth. It’s not bumpy enough. You need a little more friction when you’re reading something. That friction is what makes content thumb-stopping. It’s when somebody says something in a weird way that makes you stop and reread the sentence. That’s how you know it came from a human being. That’s the kind of content the world needs. We don’t need smooth content that your eyes just roll over and that means nothing.

Janette Roush: DMOs and NTOs are in an interesting spot because we don’t necessarily capture a lot of private information. We don’t sell anything, so we’re not sitting on some kind of treasure trove of private information. Where this comes into play, I think, is in the example I gave earlier of that super-connected trip in Times Square. You have to ask: what cameras are there in a city? What information do those cameras capture about me? And what can AI do with that information? That’s not really a DMO question right now. Even in the example I described, where different systems connect to one another, that doesn’t exist today. A DMO doesn’t have access to that information. And even if it did, I would assume it would be anonymized. You’re not looking at people’s faces and saying, “Oh, that one’s a tourist.” But you can also imagine that, in ten years, perhaps something like that could become possible. It’s a really interesting thing to think about.

For this moment, our AI policy hinges on transparency. Using AI isn’t inherently good or bad, moral or immoral. What we don’t want is for people to feel tricked. You wouldn’t want to find out that I had secretly recorded this call, generated a transcript and then used it in ways you didn’t expect. You wouldn’t want to see a photo and think, “That’s a great photo of Times Square,” only to find out later that it was AI-generated and feel tricked. So, if we’re going to use AI, we say that we’re using AI. I think that is essential. That’s how you create, keep and earn trust. I think that responsibility only grows as DMOs become the curated, trusted sources I was describing earlier–if we’re asking people to rely on us for the provenance of their information, we have to hold ourselves to a higher standard than the slop that surrounds us.

Janette Roush: I’m really excited about this idea that everything I need can live here on my phone, giving me perfect guidance for having a perfect experience in a destination, where you’re not necessarily waiting in line. Discoverability isn’t just, “Here are the top ten restaurants near you,” but “This is exactly the kind of restaurant you like. If you go here at this time, you won’t have to wait in line. It’s similar to places you’ve enjoyed before, but different in ways that I also know you’ll appreciate.” I like the idea that AI could do all of that in a way that doesn’t contribute to overtourism.

When I’m travelling in Europe, I’m very cognizant of the fact that I don’t want to be just another dumb American checking off the same three things on a list. As a New Yorker, I want to be cool. That’s really important to me. I don’t want to wait in lines to have the same experience everyone else is having. AI contributing to that, I think is really cool.

Janette Roush: I would love to know whether people care where the information comes from when they’re asking a language model a destination-related question. Is it important to them if that information comes from an NTO or a DMO, specifically because those organizations represent the locals, too. We hear that people care about overtourism, respecting locals and traveling in a responsible way. But do they care whether the response they get from a language model incorporates that point of view?

I know it’s still early. I don’t know whether people think that deeply about where language models get the information they use to generate these answers. But the answer to that question would change my work quite a lot. If people said, “No, I want the DMO to be part of that response,” if they know what a DMO is in the first place, and they want that perspective included, that would matter.

I’d also want to know whether people want more visible sourcing. Do they want an answer that says, “According to TripAdvisor, you should do this” and “According to Brand USA, this is the best time of year to go,” alongside other sources? Is visible sourcing something that will be important to people?

Janette Roush: It’s provenance of information. How important is that to people?