"The Taylor Swift of Travel AI" on Agentic AI, Organizational Upskilling, and Trust - Janette Roush, Brand USA

Josiah: I've been really looking forward to this conversation. I very much enjoyed our conversation about a year ago. We recorded at the very end of 2024. So much has changed in the world since then, Janett, and so I wonder if we can just sort of, we have a lot to get into in this conversation, but just very high level over the course of 2025. I'm very curious for you, were there any surprises when it came to AI? Anything that you felt was underwhelming? I know we were talking a lot about agents at the end of 2024, whether it's with agents or just generally with AI. Kind of what stood out to you or surprised you in 2025.

Janett: I will say definitely I went into 2025 reading and hearing that it was the year of agents, even though nobody really defined what do we mean by that? What does that look like as a person who uses a computer? And so maybe it should be unsurprising that it did not pan out that way. I found that very underwhelming, and it was all kind of marketing hype. I will say that I've done a 180 on that coming into 2026. I've spent a lot of time recently playing with Claude Code, and that has been blowing my mind in a really big way. So I've now started to say, okay, if somebody says what does this mean to me, I feel like I can answer that question now, which is good.

Josiah: Let's talk about that a little bit because for myself over the holiday break, I was experimenting a lot with Claude. And before we started recording, you said even over this past weekend you've been experimenting with it. What's standing out to you with Claude right now and Claude Code specifically? Maybe just high level for people who have no idea what we're talking about, what is Claude Code and what is standing out to you there?

Janett: So Claude Code is a version of Claude. Claude is a competitor to OpenAI. Claude is a chatbot made by Anthropic, so it is a competitor to OpenAI's ChatGPT or to Google's Gemini. And you can also access Claude through an API or inside the terminal on your computer. So the black coding box that as a non-programmer, to me, it's kind of intimidating, right? But it is, if you ask AI, you know, walk me through the steps on how to set this up, it's dead simple to use. And it allows you to, if you're familiar with Lovable, which is a vibe coding product, it allows you to describe in plain English a website or web interface that you want to build, and then it will just code it for you. And Claude Code works much the same way, which has been great. I've used it over the past six months to really build up a personal website for me where I host all of the video content from Brand USA's Agent of Change webinar series. I host it at janetteroush.com. So it's just like a little thing that I use as a playground for how does Claude Code work and what are the use cases for it. But now it has, you can audit your website to see how does AI look at and read the information that's on your website. I have a list of resources there, so if you want to learn more about using AI, you know, this is where I personally go to listen to podcasts and watch videos, so that's been fun. What I've pulled off just over the long weekend that just wrapped, I watched a podcast. I actually, I listened to the podcast at the gym and then I got on the treadmill. I'm like, I'm gonna watch the actual video of this. And then I came home. It's a podcast called How I AI, it's hosted by Claire Vo, and she had a guest on, her name is Teresa Torres, and she talked about how she created kind of an internal filing system for Claude Code that remembers information about you. You tell it every so often, you need to take everything you are learning from our conversations and add it to these context files. And so it wasn't hard to set up. What I ended up doing was taking the video, I fed it into Google Gemini, which can read and understand YouTube videos and said, I have a Mac, give me a step-by-step guide to building this from scratch. And then I just followed the steps and now I have kind of this chief of staff for my brain that also you can do work in. So I can give it notes throughout the day, and then at the beginning of the next day, I just type in a slash and then the word today and it will give me, okay, here's what's due today. Here's what's due this week. Next week you have a podcast at 10:00 AM. Today, it gave me a full dossier on you. So I have your psychographic profile. Your archetype is the intellectual collaborator and the empathetic storyteller. Could tell me some of this is writer, do webinar and conference intel. So you recently featured in No Vacancy News Roundtable with Katie Cline.

Josiah: Yeah.

Janett: Recurring themes, yeah, profitability concerns. The K-shaped economy, bifurcation in hospitality, opportunities for independence in smaller markets. We have warm icebreakers, we have the ego stroke, the deep cut, the webinar reference, the CMO bridge, the money talking points, landmines and things. So just, it's amazing. But in addition to that, because I have some information in here about, you could set up a project so.

Josiah: Yeah.

Janett: It's like I barely, I am still wrapping my head around doing this, so I'm barely able to describe it. We have a repeatable process at Brand USA where we create state fact sheets, so you can see them at thebrandusa.com and it's just tourism information about the 50 states and six territories in the US. And then it's a PDF and some of the information comes from NTTO and some comes from Tourism Economics. So human beings sit down right now, collate all of that information. They pull reports from Salesforce to understand how many times is Utah featured in our social media or did we bring any fams to Utah? What members, what people do we work with are based in Utah? What press stories was Utah mentioned in that we had a role in? So that's right now manual to compile all of that. And we send it to designers to have it designed as PDF files. So I just described, so one of the people in that process wrote out all of the steps for me, and I just said, hey, working on this project, can you help me with it? And I pasted in his stuff and it wrote out a plan, said do you want me to execute this plan? I see you've given me access to the download files of press mentions, et cetera. In five minutes it had text files, 56 text files with all of the information that could be found on my computer. And then I gave it a copy of this PDF file and I'm like, can you recreate this as an HTML? Which, it still needs a lot of tweaking, but it's 75% of the way there.

Josiah: Wow.

Janett: And just like now, this is something that could live on our website, be updated once a month with a little more work. It could update in real time every time a press story runs. And I can't believe that this is possible. So when we talk about agentic AI, that's what this is. It's taking a process and moving steps of the process forward, and you could even go to it and I said, hey, can you make a presentation for me with buttons to click forward and back just to explain what I did and how it worked? Here you go. You've already uploaded your brand guidelines, so it has your logo and it's in your format with your font.

Josiah: So interesting.

Janett: So immediate. I mean, this is so recent, how good this has gotten.

Josiah: This past weekend. This is what I love about you and your work, is you're always experimenting. Yeah, I mean, for me, the last year agentic AI felt like a buzzword and then you start to get into the details and specific workflows, but you have to start building to sort of see this at play. But I love your notion of learning by doing. And I think when we recorded at the end of 2024, at that point you were talking about, I think it was you wanted to use Python or learn Python, and you didn't have that experience, but you were asking ChatGPT in that instance to take you step by step. And then you explained doing this with other tools. I found that really helpful myself. And so confession time, I think after our conversation, a goal that I had was to do something similar in 2025, right? And nothing happened until the last couple days of the year because what happened is I thought I had to go in and learn all these tools I was hearing about, vibe coding and stuff. It sounded easy. Then I started getting into it and I was kind of hitting roadblocks. And then I went back and kind of recalled what you and I were talking about. It's like actually just have a window up with an LLM, you just have it step by step. So I made a ton of progress in the last couple of days of last year doing that. I feel like I don't have any more coding knowledge than before, but I always just took this super stupid approach of just having it be step by step, where do I.

Janett: Tell me what to do next. What did I do wrong? Why didn't this work? I am still doing that.

Josiah: Amazing. It's really interesting to do this. I would encourage all of our listeners to think about experimenting. I'll also link in the show notes to our last conversation because you went into detail about your experimentation process. I want to change tracks a little bit and talk a little bit about progress that you've made towards three key areas of focus that you have in your work, and that is internal upskilling, enabling the industry, and making the US more discoverable. You have a very big important job with the impact of tourism. You know, we have people in the travel ecosystem that listen to the show, but also people in hospitality. And one key theme that has stood out is the importance of tourism, attracting people to our destinations. If you're running a hotel, if you have a restaurant, this matters a lot to you. And you are raising the bar not only in the US but around the world in doing this. I want to break down these three components in a bit of detail. Let's start with internal upskilling. I wonder if you could share with our listeners some progress that you've made over the past year in leveling up your teams.

Janett: So one of the things I implemented when I first started at Brand USA was create guidelines for the organization, provide access to paid tools. So if you haven't done that yet, that is step zero. And it's what helps people feel comfortable and confident that they're allowed to do this, because they also, they will understand what they are not allowed to do with this. And people want to hear it. They don't feel safe if they're not doing that. And then conversely, you're going to have one or two people who are going to town with their own tools and possibly free tools that are open for training future models. If they leave your organization, they're taking all of the information about your organization with them inside their personal LLM. So you don't want any of these things. You want people to feel comfortable talking about this stuff and sharing it. So take care of that. And then you want to provide an environment for ongoing learning. So we have a company Slack channel where you can exchange thoughts and ideas, what you're doing in AI in that channel. And I also do, I'm not as religious about this as I need to be, but I try to do weekly trainings with the staff in our international offices just so that we are again continuing to see and share what we are doing. And now my goal for the next year is I want to take us from like there are some people who are now really great and skilled and comfortable working on their own with AI. And it's taking that and it's turning it into departmental processes. So we now have a process for working with RFPs where we create a project in ChatGPT. When you are, you can use that project when you're initially writing the RFP. You can then upload the completed RFP as a knowledge file inside the knowledge base for this project. If you have a team account in ChatGPT, you're allowed to take a project, which is just a folder that everybody can look at, and you can add members of your department or members of the RFP committee to that project. And that way everybody can read everyone else's chats. And so you want to use it to create a rubric for measuring how does somebody's response to the RFP measure up. You can use the original RFP to write that, and then you can build inside of this chat. Before we had some RFPs where maybe half of the RFP committee was using a language model to help them evaluate the responses, but each of them were doing it on their own, in their own accounts. And that gets time consuming because if there's 40 responses, you have five people each uploading and asking the same questions or different questions. So it's like, why don't we just do it once, standardize it, and then have it out in the open so that everybody can read it, everybody can see if they agree or disagree with how the RFP was evaluated. Then also, somebody in our team built an artifact in Claude, which is just like a private, it's like a webpage inside of Claude that they use to collect people's own grading rubrics for an RFP response. So before we were doing that in Excel, nobody likes to open up and save 40 different Excel files to judge 40 different RFPs, and you have to email all of those files to the person running the project. And then they have to go through and calculate everything. Not only is it dumb, but it's not enjoyable. It's not an enjoyable process. So this, you fill out the form in the Claude artifact, you can read how other people on the committee, when you're done, you can read how they responded to each of the RFPs. It keeps a running tally for the person who is ultimately pulling together all of the scores, and it guides you through how can you use the original language model using that to make its own score. It helps you be the human in the loop because it encourages you to look at, well, what did the language model think? And then how is your response different from what the language model thought? And in some cases it might encourage you to look at your own biases, because of course the language models are biased. We wouldn't want to just hand this job over to it and say, I'm out, I'll work with whatever vendor you tell me. No, nobody's advocating for that. But humans have biases too. We have a friend who works at this vendor or another vendor that you may evaluate differently, right? This encourages you to say, well, if you look at ostensibly a neutral viewpoint, does this change the way that you think about the response?

Josiah: Interesting. I remember in our first conversation you talked about identifying big organizational problems to solve. But even on a personal level, thinking about what do you enjoy in your work and what is a little tedious. And I think for many, or all people, processing RFPs is not high up on the list of fun things to do. And you talked about you enjoy writing, it's a side project, but it has to be done. And so I think kind of thinking about what's an important or expensive organizational problem to solve and then what is a little tedious to do? And that's maybe a good candidate to solve with AI. So I love you invest a lot in upskilling your teams internally, but you're also focusing on leveling up the industry as a whole. And I find that work really interesting. With regards to that, I know you have launched a webinar series. I wonder if you could explain a little bit about why you're investing in leveling up the industry's use of AI. You are called the Taylor Swift of destination AI, which is phenomenal. It might need to be the title of this show. But why are you investing not only in your own organization but the industry at large?

Janett: Because when it comes to attracting international business into the United States, we're not going to always be able to outspend every other national tourism organization or every other global destination. So I think it's incumbent upon us to be able to all come together and say yes, when it comes to the traveler getting here, or if it comes to domestic business, which we're not responsible for as Brand USA, yes individual destinations, attractions, businesses may be competing with one another, but when it comes to bringing that international traveler into the United States, that's when we all need to be able to pull together. Make the best possible use of our limited resources because I don't think you're going to find an organization who listens to the podcast who says, nope, I got plenty of money. I have all the time and people and resources I need. We're all constrained and we're only going to get more constrained. So how can we be a step ahead when it comes to taking advantage of these tools that make us, doing our jobs more and make it a little easier, a little faster, maybe a little better.

Josiah: I love it. I guess like whether it's internal training or external training for the industry at large, I'm curious how you think about balancing some of the basic with more specific pointed applications of AI. I remember our first conversation you were talking about, sometimes conferences will do here's the 70-year history of machine learning and computing. It's like, well, great, but it's kind of hard to action that. How do you think about what topics to cover, what specifically to cover in training or a webinar? And the reason I'm asking is I want people listening to also think about training their teams. I'm curious how you think about that.

Janett: So I'll say I follow largely, if I'm keynoting, I follow today largely the same format from the very first time I ever spoke about AI in 2023. And that is start with kind of what are we talking about? What do I mean when I am talking about AI? Here are some quick ways for you to get your head around how would you use it, and then let's go into some case studies. And so that can look like the example I just gave you of using Claude Code. But generally it's going to be, here's a small one. Now let me, if everybody's with me on that one, let's move up to a medium one, and then something that is a little more spectacular. Then I will typically talk about what does it mean for the group of people that I am talking to. So which typically is how travelers are using AI. So even if you, because there's multiple ways of attacking AI, right? It's the individual or organizational, how am I using it? But it's also how is it changing my job because travelers are now using it, or there are now AI overviews in Google search, and so DMOs are seeing a decline in organic traffic to their websites. Well, what does that mean for DMOs? What is our job going to start looking like? So I try to really, I don't want to dive into the most extensive, difficult to understand use case. Hopefully that didn't drive away the entire audience, but at least even if you didn't follow along with what I was describing, you know what it means to make 56 state fact worksheets. People understand, okay, this is work that there's now a way to take steps of it and automate it without personally knowing how to code in Python. And so I just, I try to really ground it in the, we're going to come with me on this journey and I'm not going to leave you behind. And then kind of end with a vision of, and this is what it could mean for, if we are all using MCP servers to connect information about our destinations to consumer LLMs or to B2B, what does that mean? What does that look like? And take you on the journey so you can see where the finish line is.

Josiah: I love that. I mean, speaking of case studies, let's talk about one of your other focus areas, and that is making the US more discoverable and bookable with the use of AI. This has been a big initiative I know for you and your teams over the past year. I wonder if you could explain a little bit about what you did and why you did it.

Janett: So we launched, Leah Chandler came on board as our CMO right around St. Patrick's Day last year and said, working with the agencies, I want to launch a new campaign promoting the United States, and I'd like to launch it at IPW, which was in June. And I don't think I want to launch it on our website that dates back to 2016 or something like that. So she tasked Miles Partnership with, we're creating this campaign. The campaign is called America the Beautiful, and we should probably take the next six weeks to make a website for it to live on. Which anybody who.

Josiah: Six weeks is a very tight timeline.

Janett: Yes. Even today, in the age of vibe coded websites, this was not that. And if you're talking about a big destination website, you wouldn't want to do that. So it was already a quick turnaround time. And to be able to have all of the content you need to fully explore the United States in six weeks, we wouldn't have been able to migrate information from our Visit the USA website in time to make that happen. And so Miles actually said, and we'd been doing an RFP at the same time, but they're like, we think the solution here is to embed AI throughout the website and not just have it be a little button in the corner of the website that says click here to chat with us, because we're all trained to, oh God, I don't want to talk to a website chatbot, no thank you. But if there's an invitation throughout the website where you say, oh, tell me more about this, or I'd like to explore sun and sand in the United States, or things that invite you in. And we created a partnership with Mindtrip to bring that to life. And so that website was able to launch in six weeks at IPW when we announced the new campaign, America the Beautiful, and the website. It didn't have thousands and thousands of pages of content about the United States, but it could still answer, serve that role because of the information that Mindtrip was able to bring to the table inside of their proprietary system.

Josiah: Janett, I want to unpack this a little bit because I was playing around with America the Beautiful over the last few days and it was interesting, both high level, it was interesting, but also, you know, an area I know really, really well, my hometown San Francisco, going deep into neighborhoods. So I kind of tested it on a bunch of different levels and found it really interesting and engaging. But I want to build on what you were just saying because in our first conversation, we were talking a little bit about how some marketers in this case might think that their job is content creation, right? And it's actually a more important job. And you've also talked about opportunities that you see as returning to first principles thinking. I feel like this is very connected to that. I wonder if you could unpack a little bit more, maybe through the lens of what you were describing, building around what is first principles thinking, how did it apply in your situation, and how might it be an opportunity for our listeners?

Janett: Well, part of the thinking was that if there are fewer people coming to your website, you need a way to milk more out of the people who are coming in terms of first party data. And so what having this chat interface on your website does is allow you to learn more about the people who are coming to your website. So we are able to collect all of those chats that happen, analyze them, and eventually my goal is to make it something that the industry themselves can log into a platform in order to access those insights that come out of the chats. So that's part of, I really see the DMO moving to a more data and information focused role in the future. If our role from a first principle standpoint is that we are the one organization that sits kind of at the hub of the tourist, the resident, the local government and tourism businesses, businesses that benefit from the tourism economy. So we're well poised to be a nonprofit that could be a source of non-biased data about those destinations and to share that with those four entities and to make it easier to access that kind of information.

Josiah: But what you're describing feels really important in a time of rapid change, right? Because I think this feeling of somebody make a new website or somebody create something that describes what we do is so pervasive across travel and hospitality. And coming back to why does our organization exist, right? And so it up levels, it allows you, I think, to navigate these new opportunities faster and more effectively than somebody make a new website.

Janett: Right, or tell me how many visits we had last week, or how many referrals did you send to my business? If we keep chasing that, it's not going to go well for DMOs and that was never our job to begin with. Those were just indicators that we were doing the work. So now we have to define what our new indicators that we are doing the work. And I think part of that is we have to accept that content, the content is still important. Content is king, but where people access that content isn't always going to look like your website. And I think eventually it will be inside of the language models where they already do their work or do their travel planning. So how can we make sure our content and trusted, verifiable information about our destinations can be accessed by a private, on-premise language model that a business event planner uses in Europe, as well as an app inside of ChatGPT, the way that Expedia and Booking and TripAdvisor all have ChatGPT apps. That's just using an MCP server to plug into their website experiences and bring it inside of ChatGPT. So that's how, yeah.

Josiah: If we could, I want to make sure our listeners understand how important this is because it is a distribution and a visibility challenge you're solving for, but it's actually even more important than that. And you referenced this a few moments ago, but this sense of being the source of truth and this trusted place of information feels exponentially more important in a world where anyone can create anything and there's this explosion of content. And lies. I wonder, there's been a few instances recently where this is underscored. I wonder if you could share some of those just to help people understand being this a trusted source of truth is so important right now.

Janett: So over New Year's Eve, somebody on TikTok, and I don't think this was even an AI hallucination, sometimes the problem we need to solve for is not just hallucinations, it's human beings who are liars. So somebody posted a TikTok video saying, oh, make sure that you stand in Brooklyn Bridge Park on New Year's Eve at midnight because there's going to be fireworks over the Brooklyn Bridge. That was a lie. And so there were a number of New Yorkers, I guess, gathered in that spot and then they themselves were on TikTok saying, yeah, I don't see any fireworks. A similar story was over the holidays in London, somebody on TikTok posted that there was going to be a Christmas market, a holiday market outside of Buckingham Palace. There was not. And so the BBC sent a reporter out there where they're interviewing people in their coats and they're like, yes, I thought that was the perfect place for a market, but now I'm just going to find a cheeky Nandos. Or you could even think back a couple of years ago in Scotland, there was somebody created a Willy Wonka themed interactive experience and advertised it online and it looked fantastic, but the images were all AI generated. And then when you got there, it was just like a warehouse and one Oompa Loompa and bad face paint. It was not remotely like what had been advertised. And in a world where we both can't trust that the information just provided by the language model is going to be correct, and we also can't trust people to be honest, we need somebody who is able to put the stake in the ground and say this thing is real. And I think the DMOs are the correct organizations to do that because we sit in this place of public trust. And so I think if we can start to figure out how to take our information and put it in databases that can be accessible through APIs and through MCP connections, which is just an API that a language model can read without coding, taking care of that, it can still feed your website, but then it can feed all of the other places where people get information today. And I think consumers are going to start wanting this. How many times are you going to get burned before you're like, well, how do I find out if this is a real event or not? We should start solving for that now.

Josiah: It's so interesting. It feels like there is, the rise of social media with the democratization of anyone's voice can be heard. And I think that is powerful and can be good. It almost feels like now there's sort of this counterbalance with this next wave of AI-powered technologies where maybe it's more of a second order effect, but the power of brands and a trusted voices and maybe institutions, it feels like it's kind of maybe going or there's an opportunity for trust to move in that direction. So if you're working at an organization, you have this opportunity. I think the practical way that comes to life is exactly what you said though. It is the organization of the data. And so I think when people talk about organizing data and databases, it doesn't sound that cool or sexy necessarily. But if you think about modern web development, it's all driven by databases. And this is the mechanics of how you bring what you just described to life. You have to get good at this. This is a mission critical skillset for your organization, right?

Janett: Yeah. And so when we talk about jobs are changing, it's like, yeah, AI will make jobs change, but there was still going to be a need for people to figure this out and to build and to manage these databases. It's that, that's something I want to bring to life this year. I want to create a recommended structure that people can use, and then ultimately, if you are a tour operator based out of Italy, I want you to be able to access destination, attraction, et cetera, information from the United States from one source, and I want Brand USA to be able to be that source.

Josiah: I love that. I mean, to this point it seems like there are so many new things to learn, to dig into. I appreciate you breaking some of this down. I'm curious, what is inspiring you or where are you going to learn these days? In our first recording, you shared some resources. I have followed religiously every day since then and learned so much. At the beginning of our conversation, you talked about How I AI, which I'll link to in the show notes. Is there anything else that you are enjoying or learning from these days that our listeners might enjoy as well?

Janett: I don't remember if I mentioned this last time, but there is a media and product company called Every. So the website is every.to. And it's like a new AI-first business model where they are at their heart, they're a media company, so they write great articles, both very practical about how to use AI, but also kind of more philosophical about what is this going to mean for the future of work? Where is this going? And then they also create products. So they have a product called Monologue that is, it's an alternative to Whisper or WhisperFlow, which is a voice transcription tool, along with some products to organize your email inbox and other things. So it's an interesting ecosystem where you can learn things and then actually work with tools to do the things you're learning. I've also really been enjoying a podcast called Marketing Against the Grain, and that is, it's two CMOs, it's the CMO of Zapier.

Josiah: And HubSpot.

Janett: HubSpot. Do you listen to that?

Josiah: Yeah. And they also have, it's the audio podcast, which I tend to find my default mode. But also, as you mentioned with How I AI, there's also a lot of, I find a lot of interesting stuff that's happening on YouTube too, because people are screen sharing and so it's not just talking, it's like, let's go build something. And then I feel like that's really interesting to dig into.

Janett: Yeah, that's definitely treadmill time for me. I'm so craving multitasking that I have to be in an environment where I can't do a single other thing and I'm forced to look at the screen and make it work.

Josiah: It's funny, the voice to text is interesting though. On my run this morning, a couple things came up that I wanted to talk with you about. So I'm transcribing as I'm running, which is, I don't think anything would understand. But AI's gotten really good at transcribing. It's like, wow, these notes are formatted and it's great. So it's interesting. Before we go, Janett, I'm curious, we're recording this at the beginning of 2026, we have a year ahead of us. Is there anything you're looking forward to or excited about coming up in the year ahead?

Janett: Gosh, I'm excited about finding more time to figure out how to talk about what I am learning and then share it more widely across the company and across the industry, because I do think this Claude Code piece, I know they have started to release products to make it more digestible for people who aren't coders, but I want my entire company to be using this tomorrow. And I know that is, that's not practical because it's still technical to set up. And then I look externally at the industry and I know some people are still using free AI tools as fancy Google. So to try to get every, the analogy that I'll make in speeches is that it's okay if you start at just the tiniest, smallest use cases because you just gotta be in there and using it to get to the point where you are installing this program in your terminal. You can't skip any steps. You just have to do it. Now I'm like, okay, please everyone hurry up and do it so that I can talk with you about it. So I'm looking forward to finding my group of people to talk about this with in the next year.

Josiah: Amazing. Well, time is the most precious resource, and I appreciate you taking the time to speak with me, to teach all of us how we can learn about AI, how we can experiment with it, and leading the way for Brand USA and for teaching so many around the world. So thank you, Janett. This was a real treat.

Janett: Thanks so much.