How Navan is Simplifying Business Travel & Expense Management With AI

[00:00:00] - [Speaker 0]
A big thank you to Denodo for helping me make more than 60 monthly interviews possible across the Tech Talks network. And as businesses move from GenAI to Agentic AI, trusted data becomes everything. Everything from GenAI to Agentic AI, Denodo is helping organizations build intelligent, secure, and scalable AI solutions with data access, governance, and explainable results. So build AI that you can trust and do it with Denodo. And you can learn more by simply visiting denodo.com.

[00:00:38] - [Speaker 0]
What happens when the very people building the future of AI are also willing to question everything we thought we knew about software, user experience, and even human interaction with technology? Well, my guest today is someone who has spent the last decade doing just that. The Navan cofounder and CTO is gonna be joining me today to share how AI is reshaping business travel, customer experience, and the very way that we interact with software itself. And from that moment that ChatGPT appeared way back in, I think, late twenty twenty two, my guests believed that we were witnessing the beginning of a complete shift away from what he calls forms and tables and a move towards conversational context aware experiences that will be powered by AI agents. So today, he'll explain why he thinks the traditional software interface is already becoming outdated and how they are building systems designed to think, reason, and anticipate disruption and act on behalf and act on behalf of travelers in real time.

[00:01:46] - [Speaker 0]
I will also talk about the rise of agentic AI, the challenge of hallucinations, and why building trustworthy AI systems requires far more than plugging an LLM into a product. And he will share how their AI assistant, Ava, handled massive travel disruption during events like the Heathrow Airport Fire, which I personally still get flashbacks from to this day, and severe weather incidents and supporting thousands of stranded travelers without the long wait times that people have sadly come to expect in the travel industry. As someone that travels a lot, this is a fascinating discussion about smaller purpose built AI models and how they're outperforming much larger frontier models in specific business environments. But today is much more than just about travel tech. It's actually about curiosity, obsession, experimentation, and what happens when a CTO genuinely treats technology as both a profession and a playground.

[00:02:44] - [Speaker 0]
So if you've ever wondered where AI moves beyond hype and starts delivering measurable operational value in the real world, I think you're gonna love this one. But enough from me. Let me officially introduce you to my guest now. So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do?

[00:03:08] - [Speaker 1]
Yeah. My name is Ilan Twigg. I'm the CTO and cofounder of Navan. We started the company eleven years ago and six days. We just celebrated the our eleventh anniversary.

[00:03:21] - [Speaker 1]
I'm a software engineer in my education. I love technology. I'm really lucky that my hobby is my work, and my work is my hobby. I truly mean it. I truly love what I do.

[00:03:33] - [Speaker 1]
And my job is my day job is to make sure that we always use the best technologies to best serve our our users, All of them all the time. And yeah. So I'm I'm I I get to have my own playground, try new technologies out, and then build stuff, and then get it to production, and then move on to the next thing. And the last three years, I don't remember any anything remotely close to that in my entire career when it comes to technologies. I feel like I'm being, like, by the time I master the the technology that came out two days ago, there's a new one.

[00:04:10] - [Speaker 0]
Yeah.

[00:04:11] - [Speaker 1]
And very exciting times. Very exciting times. A lot of opportunities for me personally, for for Navan, for, I think, for any company today in the world.

[00:04:23] - [Speaker 0]
Well, congratulations on your anniversary. It makes such a difference when you genuinely love everything that you do. And listening to you there, it really comes across your passion for what you do. And there is so much excitement around all things AI and agentic AI and and the thirst for change right now. But I'm curious.

[00:04:43] - [Speaker 0]
For somebody right in the heart of this space, how do you see AI redefining how humans interact with software? How do you see this playing out?

[00:04:52] - [Speaker 1]
Yeah. So, actually, when ChechiPT just came out, as a reminder, it was November 2022. I think that still in November, the November 2022, I actually tweeted a thought that crossed my mind, and AI generated an image for me. Back then, it wasn't that great, but still. And it was it was terminal, and then next to it, the mouse, and then next to it, AI dot dot.

[00:05:20] - [Speaker 1]
It's like something new. I I knew that this technology would redefine how humans interact with computers. Two years ago, roughly two years ago in Helsinki, there was a big conference, Slush. And in Slush, my presentation was about all about that, and I shared my vision about how I think this technology would reshape how humans interact with machines. And I I re I even gave a demo that I was not allowed to do a live demo, so I did screenshots from a live demo.

[00:05:54] - [Speaker 1]
And I showed how AI is fully capable of understanding well what the user wants and then generating the the UI needed for this operation. And I walked them through a mock up of how could a travel app look like. And I showed how it is full of context. You see what you need to see, when you need to see it, and then you don't see it anymore. And, really, like but I really put examples on on a, you know, Microsoft Word.

[00:06:24] - [Speaker 1]
I literally counted the how many different things could you do with Microsoft World 1.zero? And then in '95 and 02/2000 and in 2024, it's like I don't remember. It's like it was hundreds, if not more, times more than what you could have done, but the interaction is the same. It's the mouse, primarily the mouse. So it kind of breaks, and you kinda lose you kinda forget.

[00:06:55] - [Speaker 1]
You know, you I remember that I could do this thing. Where was it? Where was it? And then you go through all of the menus. And I showed how ChatGPT back then mastered Microsoft Word and could help me, guide me on how to change the footer notes in a multi page document, whatever, something super super esoteric that no one remembers how to do.

[00:07:18] - [Speaker 1]
Fast forward to today, in Navan, we fully fully embrace this concept of, I call it back then, conversational plus experience. We I think that the world kind of skipped the whole copilot experience. I I mentioned the Copilot approach, and I said that I don't think that I think I I said I I think it's like a midway, and I don't think that it would last for long. And the the real deal would be conversational plus plus UI. And if I may give you an example for something that I think demonstrates the difference between what I call forms plus tables, which is pretty much everything that we do today in software is forms plus tables.

[00:08:02] - [Speaker 1]
It's definitely in travel. When you when you search for some flight, you fill up a form, then you get a table with search results, and then you can resubmit the form. You can filter and sort. But that's the experience, forms and tables. It was revolutionary in 1995 when the Internet showed up.

[00:08:22] - [Speaker 1]
But later like, thirty years later, with AI, you can generate you can create a much, much more compelling experience. And so today in Avan, you can do pretty much everything through a conversation. So you can have the form and table, but you can click do it with AI. And when you say do it with AI, you talk to an agent. So I'll give you the example.

[00:08:48] - [Speaker 1]
A user says, hi. I need a a hotel in New Delhi on the tenth and provided the name of the exact hotel. The name was inaccurate. There was a spelling mistakes and also the name was inaccurate. But the u but the AI got what the user wanted.

[00:09:04] - [Speaker 1]
So the the AI said, okay. Sure. Let me help you with I can definitely help you with this hotel in New Delhi. However, I see that you land your flight arrives on the ninth. Should we look for availability starting the ninth?

[00:09:18] - [Speaker 1]
And the user was like, oh my god. Thank you so much. Yes. On the ninth. And that I think is a it's it's an anecdote, but it explains the difference between forms and tables, them, and the when the responsibility is fully on you versus an intelligent conversation plus UI.

[00:09:40] - [Speaker 1]
There is so many examples that that I can share, but that that's where I think we're we're not even headed. That's where Navan is at. I can talk about where we're headed because this is not this is not where we stop or where the world stops there. You mentioned agentic.

[00:09:58] - [Speaker 0]
Yeah.

[00:09:59] - [Speaker 1]
And, a, I like that you differentiated between LLMs that showed up in November 2022 that demonstrated unbelievable capabilities, but it was you cannot deploy a production application based on an LLM. The only way to make it work is you need a combination of LLMs. That's where it turns into an agentic system where each LLM has its own role or LLM plus a role is that's an agent. LLM plus base prompt, that's an agent. Very simple.

[00:10:41] - [Speaker 1]
It there's nothing more to an agent than that. It's an LLM plus a base prompt. You are a virtual you are a travel agent working for Navan, that's an agent. And you are the supervisor of agents working for Novan, and your goal is to review what they do and make sure that there's no hallucinations, that's a different agent. And they and and it's a real true story from when we built our first virtual travel agent, and we realized that LLMs hallucinate.

[00:11:09] - [Speaker 1]
And LLMs don't really follow instructions well, and and then we created the first agent. And this agent agent's role was to monitor the first agent and make sure that there is no no hallucinations, no violation of standard operating procedures, etcetera. Just like in real life, we have a call center, we have agents, and they're the supervisor, and they review the calls. So that's definitely and I AgenTik is the is a major revolution. And a year later, but AgenTik came out in January 2025 roughly.

[00:11:42] - [Speaker 1]
That's when I was in Davos and everyone was talking about it. Yeah. January 2026, open cloth. That was the that that's the new thing. And, yeah, definitely opens the door to a coop.

[00:11:57] - [Speaker 1]
To me, it is as big as the as the as the introduction of Chechi PT in 2022, OpenCLO.

[00:12:05] - [Speaker 0]
And you've been on an incredible journey here. Before we started recording, we're talking about I'm making well, I'm here in California covering an event. You're flying out in a few days to New York and California too. So I'm I'm curious from what you're seeing here. How do you see the the future of AI powered business travel just to further bring this to life for people listening?

[00:12:27] - [Speaker 1]
It just makes so much sense to use AI in in travel. So the first thing that we've done is the virtual travel agent that we released in a I truly believe that it was the first agentic application running in production. One of the first applications in front of the first in general. And it's amazing because it's it's always available. It has the context.

[00:12:51] - [Speaker 1]
It knows your travel. It knows who you are, and it starts to remember things that you care about and things that you don't care about. And the context the context is really important. The better and more accurate the context is, the more personalized the the conversation would feel and more and and quicker it would be, and the user satisfaction will go up. I can just share with you that the bot's name is Ava.

[00:13:13] - [Speaker 1]
We call her Ava, or in English, it's Ava. Yeah. So Ava Ava's user satisfaction is through the roof. It's like north of 8080% CSAT, which is very close to the humans. She manages, on average, 8,000 incoming support contacts per day.

[00:13:36] - [Speaker 1]
8,000 with a very high satisfaction, and she manages to solve north of 55%, meaning less than 45% would actually end up with an agent. So a tremendous value for for the users, no wait time, always available. When she cannot help you, she has no ego. It's in her prompt. She has no ego.

[00:14:01] - [Speaker 1]
If she cannot help you, she moves you to an agent. Users don't need to ask for it. She will just say, hey. For that, I'll just move you to a live agent. We see that happening less and less as we train Nava more and more and make her be able to cover more and more cases.

[00:14:18] - [Speaker 1]
And all all all obviously, she manages she allows Navan to scale our call center to levels that we couldn't even have imagined, and it is especially evident during the major flight disruptions. There was just a few major disruptions in The US. There was one not too long ago in in Heathrow. There was fire in the airport about than a year ago, and thousands of our our travelers got impacted. But luckily with Ava, there was no wait time.

[00:14:54] - [Speaker 1]
She just solves the problem for you. She knows what the airlines offer because she knows to read the website and digest it. So we really take advantage of what LLMs are really good at and turning it into real value for the users first and then obviously for Navan. It is a great example of how AI can generate undeniable business impact for us as a company. I I cannot I when there is major disruptions, Ava gets to thirteen, fourteen thousand chats a day.

[00:15:31] - [Speaker 1]
And we don't need to we don't need to there's no wait time, and we don't need to deal with staffing. Oh, no. No. We need now 50,000 more people in the call center. She's there.

[00:15:44] - [Speaker 1]
Huge value. Huge, huge value. That's about support. There is the, what I said, like, the agentic experience for managing your travel booking with the context. So, like, I I see conversations where users start with flights.

[00:15:59] - [Speaker 1]
They provide the dates. For the hotel, you don't need to provide the dates or deduced from the context and for the car. So the conversation flows so well. The LLM or the agent knows what you like. It know it remembers that you do not like carpets in the in the room.

[00:16:17] - [Speaker 1]
You don't like it, but it knows that it's very important for you that the gym opens early, as early as 6AM because that's your time. It things matter. All it knows that it that the room that you that you care about are the rooms that it's very convenient to to do a Zoom call like people are doing now. It's but it's good Wi Fi. Now how does she know that it's a good Wi Fi?

[00:16:42] - [Speaker 1]
She reads the reviews, and she just tells you. So the conversations are unbelievable. I see people the people want roll in shower. They just ask, oh, no. No.

[00:16:52] - [Speaker 1]
This room doesn't have it. The other room is not in policy, but this room has everything that you need and really fantastic conversations. Again, I compare it to forms and tables. There's no way back. It's just like there's no way back to to, I don't know, a black and white TV.

[00:17:08] - [Speaker 1]
Once you once you experience colorful TV, there's no way back from smartphone to an old Nokia phone. There's no way back from an intelligent, agentic conversation plus UI to forms and tables. And then there's OpenClub.

[00:17:23] - [Speaker 0]
And Absolutely. Love it. And and and also, I would say, I mean, you mentioned the Heathrow incident there, and I was stranded at I I remember that incident very well because I was stranded at JFK for fifteen hours. It was Oh, man. It was awful.

[00:17:36] - [Speaker 0]
And there has been unprecedented disruption in travel, I would say, over the last five years in everything from pandemics to fires to wars now. But how are travel agencies better handling crisis like this and and better supporting customers? What do you see here?

[00:17:53] - [Speaker 1]
I I think in the end, it is all about getting you to where you need to get

[00:17:59] - [Speaker 0]
Yeah.

[00:17:59] - [Speaker 1]
In the most comfortable way despite the disruption. So for example, prior to the capabilities that we developed internally with Eva that can I can tell you what she can do? But prior to that, I I have an admin. And the admin does not wait for me to call her and tell her, oh my god. I just got a notification from Lufthansa.

[00:18:25] - [Speaker 1]
The the flight is canceled. She knows that the flight was canceled, and she already looks for alternatives. And then she tells me, you need to fly out instead of Heathrow from and I forget the name of the other the other airport that is smaller, but there's still international flights. Or you need to fly first to a different destination in Europe and then and fix the whole thing for me. And in the end, what I want when I wake up, the whole drama happens while I'm asleep.

[00:18:51] - [Speaker 1]
When I wake up, I want to get a text saying your taxi will show up at nine and not 09:30, and it will take you to the other airport. You have a different flight. That's what I want, and that's what we're we're trying to get to with at scale. And, thankfully, with the technologies that we developed in the last eleven years, which are first the connectivity layer. So AI is AI, but without the ability to access the richness of the inventory of the world when it comes to flights and lodging and cars and rail, without that, the LLM would be somewhat limited.

[00:19:31] - [Speaker 1]
So we keep investing a lot. That's the key. Once we have access to the richest inventory in the world for everything related to travel, now we need to put a layer that, a, knows you well, and we noticed that it is if the AI will come up with a solution, a generic solution, that does not take into account your preferences, you're not gonna like it. You will not like it. If you find yourself compromising on so sometimes when there's disruption, you need to compromise on some things.

[00:20:06] - [Speaker 1]
But if you compromise on the wrong things, for example, you do not want to have a connection in a certain country because you know that always there for whatever reason, they stop you, and and there is like it it becomes like a two hour saga. Or and I can come up with a long list of travel preferences. We all have them. If you travel a lot, you have them. If you don't travel a lot, it's not a big deal.

[00:20:29] - [Speaker 1]
Then there's a disruption. Okay. Not a big deal. But the challenge is is to satisfy the frequent travelers. These are our audience.

[00:20:38] - [Speaker 1]
These people that travel three, four times a month, that that's the bar. We need to make sure that they are happy with with our service. And with them, personal preference is a huge deal. And I remember when we started the journey, I knew nothing about travel, literally nothing. And by now, I I see that all of the things that I've seen in the past, I became one of them.

[00:21:02] - [Speaker 1]
Now I travel a lot, and I I care about if it's five d or four d because five d goes like this and four d goes like that, and I don't want this. I only want that. And it's call call me spoiled, but it's not. When you travel a lot, these little things matter. And that's the bar.

[00:21:20] - [Speaker 1]
If you produce if you almost satisfy them, you fail. There's no such thing as almost satisfying them. So the the AI must really know you well, must have all of the options available, must reason well enough, and only then can you can you start visioning, wow. I'm gonna provide the best travel experience, including handling disruptions at scale. The user would gain from it.

[00:21:49] - [Speaker 1]
Navan, obviously, would gain from it. Very exciting. And one last thing on that, by the way, it's an endless world with endless opportunities. That's how I see it as a CTO. We started with generic models.

[00:22:02] - [Speaker 1]
They're one of the frontier models to serve the, for example, the flights, so to manage a conversation with you about flights. And today, we have in production, not not for all of the conversation, but some portion of it because we still are testing it. We trained our own model, and we expect this model to be the best flight agent in the world, better than the frontier models that are generic, are not trained on data that is really unique to to Navan or to any travel agency if they have the data on these conversations. And what we see is amazing. These models are much smaller.

[00:22:44] - [Speaker 1]
I'm talking about 27,000,000,000 parameters compared to all of the frontier models, the Geminis of the world, the the Anthropic, Cloud, and and the ChatGPT. We don't know the number, but it's be between one and ten trillion parameters. 27,000,000,000 parameters runs faster, cheaper. And with the data that we have, which is unfair, it is also performing better. Very exciting times for for us at Navant.

[00:23:15] - [Speaker 1]
I wish that we could talk more, and I'll share it. We we haven't talked about OpenCLO and what we do with it, but I can just give you a hint. We call it TravelClaw, and it takes the whole identity experience to a completely new new level. It is less about what it can do, all of the features that has another feature and another feature and another feature. We just put responsibility on it, and it needs to figure it out.

[00:23:41] - [Speaker 1]
And it can also talk to Ava if it needs to, or to talk to a live agent, or to call the restaurant, or to do many, many things on your behalf to make sure that it has one goal. It's your responsibility that Neil travels comfortably, and it's your responsibility to anticipate problems and come up with solutions. The world is really exciting. The world ahead of us is really exciting.

[00:24:05] - [Speaker 0]
I absolutely love what you're doing, Kian, how you're using technology to solve very real problems and the latest technologies as well, like OpenClaw that you mentioned there. There are so many different travel apps and different solutions and AI promises out there. What is it that makes what you're doing with AI so so difficult to replicate? Tell me what makes you stand out.

[00:24:27] - [Speaker 1]
I think I I think that I'll I'll tell you something about myself when it comes to everything, not just technology. I don't see things as difficult. I if I came I studied I read a lot about quantum physics, and it's really complex. But in the end, you need to find a way to explain it to yourself. If you can, then you understand.

[00:24:55] - [Speaker 1]
Or to someone else, if you can, it means that you really understand it. And if you cannot, you don't you do not understand it. And I that's how I am since I was a since I remember myself. I need to explain things to myself. Oftentimes at school, for example, I did not see it the way that the teacher saw it, and there were conflicts.

[00:25:16] - [Speaker 1]
But in the test, I always had high really, really good grades. And at some point, they let you go. There was a teacher that, I don't know, did not like my being, and she was she would enter the class. I was in seventh grade or ninth or eighth grade, and she would say, Ilan, get out. First thing, get out.

[00:25:34] - [Speaker 1]
Smile and get out. She did not like the the arguments that we had. Okay. I got out, but I then had a in the class in the end. And I carry this way of being and thinking until today.

[00:25:47] - [Speaker 1]
When it comes to AI, there are many complex things. To me, there's not nothing is complex. It's okay. Break it down. Explain to yourself.

[00:25:54] - [Speaker 1]
Break it down. That's how I play the piano, by the way. I started playing the piano just like that, and I play the piano. And and I'm not saying that I think that if you really care about if you have I I'll tell you how we win. Passion, a lot of passion, genuine passion for technology, genuine passion for travel and expense, and genuine passion for having happy users.

[00:26:21] - [Speaker 1]
And we can double click on that maybe in the next session, but that's I think what's that's the secret sauce. I don't I've been working for a lot of other organizations, and I have not seen that. So it may be easy to say, more difficult to implement, because you can't fake it, by the way. You cannot fake it. It's either either it's either there or or it's not.

[00:26:43] - [Speaker 1]
So that that's my answer.

[00:26:46] - [Speaker 0]
Oh, I certainly love it. I wonder if that teacher that you spoke of there could listen to this and see how far you've come. And and and when we're talking about AI here, I think there'll be so many people listening, especially in business scenarios where there is that pressure to deliver ROI on every tech project now. So how can businesses better turn AI potential that we're talking about here into to real return on investment? What what do you see here?

[00:27:13] - [Speaker 1]
In general, not just in travel.

[00:27:15] - [Speaker 0]
Yeah.

[00:27:17] - [Speaker 1]
First, you need to really understand AI, what it can and cannot do. The way for me to understand is, as I said, is to touch it and decompose it. My engineers know. I'm I'm old, but but last year, I was the second on the leaderboard of I will not say it's either Kaltzor or Cloud Code. I don't want to say.

[00:27:39] - [Speaker 1]
We use both, but I was the second in the leaderboard of the entire company as a CTO. And imagine, we have engineers in the company, and I out use them on on these tools. So I that's I really believe in it. I touch the technology to understand it. So you really need to get it.

[00:27:57] - [Speaker 1]
You need to really get it. And the second thing is you need to be honest with yourself. Identify areas that you can see value. We immediately felt that support. It was like a match made in heaven.

[00:28:11] - [Speaker 1]
Like, you we have this technology that can chat. It can chat and talk, and, it's really good. It's really talkative. You need to tone it down and calm it down, but it can the whole Siri the whole part with Siri and Alexa and the Google AI assistant failed to do for years, it just does it and does it really well. Siri doesn't get me until today.

[00:28:33] - [Speaker 1]
It's like one out of three. She will do something that I didn't ask her to do, and I'm like so, yeah, I think I identified the use case. And I can tell you one thing that I really believe in. Sometimes it's not obvious, but it's there, and you need to think well. And the other thing is one, when you implement something, don't lie to yourself.

[00:28:54] - [Speaker 1]
Don't don't don't do something to please someone or to please yourself and say, did it. You need to then look have a KPI before you even start the project. Have a KPI to say, this is success, and that's anything else but success. And then then be honest with yourself. That's, again, a recipe to get to real results with with anything and also with AI.

[00:29:17] - [Speaker 0]
And one of the things that I absolutely love about what you're doing here is you're challenging the status quo there and what everyone just accepts, especially in travel and actually creating a real solution, solving a real problem. And what would you say is that one hot take that you have on AI? Or what's one thing that the rest of the industry might believe right now that you disagree one of the reasons that why you do what you do. Is there anything you could share there?

[00:29:45] - [Speaker 1]
It's a good question. I can tell you again, it's it's very personal. For for me, it's really personal. When ChatGPT came out, I was unprepared for this technology. And I I don't think I know.

[00:29:59] - [Speaker 1]
The world was unprepared, including its own creators that I met. They were unprepared for the for that. Dropped their jaws when they saw the performance of a 175,000,000,000 parameters model first time. And it's like discovering radioactivity for the first time. And just like radioactivity, by the way, you know what you know about it, but there were so many things that you did not know.

[00:30:23] - [Speaker 1]
You did not know, for example, that that it destroys the tissue. It destroys the the atoms. You did not know that. And then I think that madame Curie paid with her life because she used to carry a radioactive source in her pocket all the time. She thought that it was cool.

[00:30:41] - [Speaker 1]
She knew what she knew, and she did not know what she did not know. And I'm saying it because we are at the same spot today with AI. There is don't know really what happens in the brain. There was a research that was published a month ago by Anthropic, and it is it is a really, really exciting topic. We already obviously got into it.

[00:31:00] - [Speaker 1]
It's called interpretability. Don't try to understand the word. It's a new word in English as far as I know. They coined the term interpretability, and it is like doing a functional MRI to the to the model. Don't assume that people know better than you.

[00:31:18] - [Speaker 1]
Just go and do your research. Get to the point that you really, really understand the technology, which is what we've done. I bought a $30,000 GPU back then, and I deployed, and I and I I isolated myself from humanity for four months. And I was obsessed about this technology, and I and and I did not think that I know more than anyone else. I assumed that I know the least.

[00:31:41] - [Speaker 1]
And five months later, I, completely by accident, got to spend a full day with the leaders of of AI in the world. Jensen Huang and was there not the whole day. Sam Sam Altman was there. Dario Amoudi was there. Really?

[00:32:01] - [Speaker 1]
And I was there. And we were and I was talking about what we do with AI, and I explained, and it drew a lot of attention. And I then realized that it is really you know, go and figure it out. You you want to be good at it, go be good at it. There is it's it still is such a new technology.

[00:32:21] - [Speaker 1]
I think that the there is this there is the creation of the models. I'm I'm I don't have intentions to be better than OpenAI or all of these guys. Don't care about it. There is a new tool in the world, and there is these companies, and that's most of the world is how to best use these tools, how to understand these tools and and put them to your benefit in the most in the maximum way. And I think that this is a completely new ground, and go figure it out.

[00:32:50] - [Speaker 1]
Be the best in the world and and believe that you can do it. Yeah.

[00:32:56] - [Speaker 0]
It is such a great story, and thank you so much for sharing it. And everybody listening here, Navan is the global AI powered business travel and expense platform that is really using technology to make travel easier for frequent travelers. And for anybody listening that wants to find out more information, where would you like to point them? I see on your social channels. I mean, LinkedIn alone, you've got over 200,000 followers, so I'll include links there.

[00:33:20] - [Speaker 0]
But anywhere else you'd like me to point everyonelistening?.co.

[00:33:25] - [Speaker 1]
We do have the blogs there, and, yeah, stay actually, tuned. We're gonna publish a a few really cool things about TravelClaw as well as about hoping to have good results from the research. It's a pure research on interpretability in the real world. Very, very exciting stuff. How you can get even more out of the same LLMs, the same technology?

[00:33:48] - [Speaker 1]
How you can can you get more? How we can make other, for example, perform better by using this functional MRI of the brain. It's it's fascinating stories, fascinating research. Again, Anthropic came up with it. Now I think that OpenAI and probably Gemini are doing also go getting into that.

[00:34:07] - [Speaker 1]
But the paper that they published is fascinating. To me, it was really fascinating.

[00:34:12] - [Speaker 0]
Well, I'll include links for everybody listening to everything that you mentioned, and urge people to keep an eye on that, especially with some of those big announcements coming out. But more than anything, just a big thank you for you taking the time to sit down with me, sharing your story. Really appreciate you.

[00:34:27] - [Speaker 1]
Thank you so much.

[00:34:28] - [Speaker 0]
Wow. I love the honesty behind the innovation and the chat about technology here. Because there are plenty of companies talking around AI right now, but there are far fewer that are willing to openly discuss the messy realities of hallucinations, experimentation, failed assumptions, and the work that is really required to build systems people can really trust. And I think his perspective felt refreshingly practical, especially that belief that businesses need to stop treating AR like a buzzword and instead focus on where does it genuinely deliver value. And I think his comments about the future of software interfaces will strike a chord with many of you listening.

[00:35:10] - [Speaker 0]
Because the idea that we may look back on endless menus, forms, and tables, maybe we'll look back the same way we now look back on black and white TVs. It does feel surprisingly believable after hearing so many of those examples shared there. But perhaps the biggest takeaway from this episode is that companies succeeding with AI right now are often the ones willing to get closest to the tech itself, experimenting with it, stress testing it, breaking it apart, and rebuilding it until it solves those real world problems for real people. And if today's conversation sparked any thoughts about the future of AI agents, business travel, or conversational interfaces, or where enterprise software is heading next, I'd love to hear your perspective. Are we entering a completely new era of human computer interaction, or are we still just scratching the surface of what these systems will eventually become?

[00:36:06] - [Speaker 0]
Remember, techtalksnetwork.com. You'll find everything that you need over there, and I'll await your thoughts. Send me a recorded message. Connect with me on socials. Just at Neil c Hughes on all channels.

[00:36:17] - [Speaker 0]
You can see, pics of my recent travels to tech events during silly season over on Instagram too. A quick thank you to NordLayer for supporting the podcast and helping me make these daily conversations possible. And if you are listening and you're responsible for security or IT, you will know the reality. The reality that most of your risk now sits inside SaaS apps and browser activity. That gap is exactly what NordLayer is addressing with its new business browser.

[00:36:48] - [Speaker 0]
So instead of bolting security on from the outside, it builds it directly into the browser itself. This means you can control access, monitor activity, enforce policies, and reduce shadow IT all from one single place. And most importantly, it does it without adding deployment headaches or complex onboarding. You get things like browser based data loss prevention, SaaS access control, and zero trust browsing, but delivered in a way that your team can actually use. So if you've been trying to simplify your stack while improving visibility, please check it out at nordlayer.com/browser.

[00:37:31] - [Speaker 0]
But I've taken up far too much of your time today, so I will speak with you all again bright and early tomorrow. Bye for now.