Can Agentic Ecommerce Finally Fix Tourism and OTAs? (270)

This week’s podcast is about the likely impact of agents on ecommerce. And specifically on tourism.

You can listen to this podcast here, which has the slides and graphics mentioned. Also available at iTunes and Google Podcasts.

Here is the link to the TechMoat Consulting.

Here is the link to our Tech Tours.

Here is the mentioned McKinsey report.

----------

I am a consultant and keynote speaker on how to accelerate growth with improving customer experiences (CX) and digital moats.

I am a partner at TechMoat Consulting, a consulting firm specialized in how to increase growth with improved customer experiences (CX), personalization and other types of customer value. Get in touch here.

I am also author of the Moats and Marathons book series, a framework for building and measuring competitive advantages in digital businesses.

This content (articles, podcasts, website info) is not investment, legal or tax advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. This is not investment advice. Investing is risky. Do your own research.

Transcript

00:05
Welcome, welcome everybody. My name is Jeff Towson and this is the Tech Strategy Podcast from Tecmo Consulting. And the topic for today, can agentic e-commerce finally fix tourism and OTAs, online travel agencies? It's kind of a general discussion about what happens when you combine agents, which are just emerging with e-commerce. And it's kind of huge. I mean, it's a sweeping disruption, transformation, I think.

00:34
And I talked about it a little bit before in a couple other podcasts, but I wanted to kind of talk about tourism as an example, because it’s kind of an area of e-commerce that never really worked very well. It's always been kind of half broken. And it's an example people point to when they talk about agentic e-commerce and why it's different. So, I thought I'd go through that a little bit and also just sort of go through sort of three or four main concepts I think you want to keep in mind when thinking about this idea.  Agents meet commerce and what does that mean?

01:02
So that will be the topic for today. Let's see, no announcements, standard disclaimer, nothing in this podcast or my writing or website is investment advice. The numbers and information from me and any guests may be incorrect. The views and opinions expressed may no longer be relevant or accurate. Overall, investing is risky. This is not investment, legal or tax advice. Do your own research. And with that, let's get into the topic. Okay, I'm just going to run through this pretty fast, cause I'm a little bit.

01:31
sick and don't have my voice.  I'm going to go through this sort of point by point. Five points. Most of them have a concept which I think is worth uh keeping in mind. So, first concept, first point.  I break this into four levels. This is kind of a McKinsey framework. I've talked about this before. When you think about traditional software, digital, uh traditional machine learning, I generally

01:58
put those two things together, level one, level two, I call that digital strategy. And then level three would be generative AI, completely different, and then agentic AI, level four, I generally put those two levels together as AI agent strategy. So, most of my stuff is sort of, hey, here's a digital strategy, here's an AI agent strategy, but it's about those four levels. And really because they operate very differently, they have different tech stacks, the data architecture is completely different.

02:27
You need the models, all that stuff we've been talking about. basically, level one, this is McKinsey, rules-based systems.  So that's very traditional software, very deterministic, things like that. You put something in a database, you put something in a record, a Word document, an Excel, a contact. It's going to stay the same all the time. It's very exact, very structured data.  Level two, analytical AI, traditional machine learning.

02:55
Okay, you're still basically living off structured databases, but now you're doing sort of probability-based analysis, traditional search engines, things like that matching at TikTok. That's mostly analytical. We used to call that predictive AI, but it's also, it’s not as deterministic, but it's the data structure is very similar. You Google's been doing that since 1998. Okay, we moved to level three. This is the new world, generative AI. Yeah.

03:23
Suddenly we're generating content, we're generating code, we're generating really almost anything, this kind of content.  Very sort of non-deterministic, probabilistic. You ask a generative AI chat bot the same question, you're going to get a different answer.  Kind of like humans. uh the data structure's completely different, the mechanisms are completely different, and the central compute is a different type of math.

03:48
So very different, and then agentic AI, which is what we're talking about, which is kind of generative AI, but it can act on its own. And it can access tools and it can go around. And agentic AI kind of looks like humans. And when you let it sort of run and do its thing, give it control of your browser, let it go through your emails, write replies, summarize your emails, things like that. It's pretty good at that stuff already. And more and more, agentic is going to be out in the real world, so that's robots.

04:16
which are kind of agentic agents with their own bodies. Although some people call that a different thing, maybe that's true. Anyways, I put it in four levels, well, McKinsey does and I buy it. So that's kind of idea number one. I'll put a slide in the show notes. I think that's a good way to think about it.  And the way you can do this as a business is you just start mapping out your workflows and then you, you know, you sort of, okay, step one, customer acquisition, we post some content.

04:45
We measure the clicks, we get them to sign up to our email, we send them a thank you, welcome message, and you can map any workflow out into steps. And each step you can sort of assign to level one, two, three, or four, depending on what's most important, or more than one. But when you start to reply to someone's email to you, you're probably going to use generative AI that's going to customize it and things like that. So anyways, you can map out your workflows against the four levels. That's pretty useful. um

05:14
I'll put in another slide. This is also from, I think this is McKinsey.  They talk about sort of the difference of traditional AI, generative AI, agentic AI in terms of adaptability, memory and learning, multi-system integration, human involvement required, decision-making, things like that. So, you can break it down by capability the same way. I'll put both slides in the notes. I think it's pretty useful to have as a reference.

05:44
Okay, so that's kind of point number one, break it into four levels. We're talking about agentic AI, but really you kind of got to start talking about generative AI to get at this. All right, point number two, which is about generative AI and agentic AI. It's this idea that, you know, these are all about coordination. That's really their strongest power.  This is like Sanjit Chaudhary, the strategist, he wrote a good book about this.  And his main point was like,

06:13
Generative AI, the real superpower is coordination. It's enabling things to connect and interact with other things. know, traditionally in strategy, we would say you can create value in one of two ways.  There's a linear process like a manufacturing line where you take the metal, then you bend the metal, then you put it on the car, and then lots of goods go in, and at the end, you add value, linear, step by step by step. At the end, a car comes out. That would be a type of value creation by a business.

06:41
The other way is by connecting entities to each other. So that would be your typical marketplace platform. know, Alibaba, Lazada, the third-place marketplace, they're not selling anything themselves. They're just connecting buyers and sellers and enabling them to do transactions or other types of interactions. So, you can sort of create value by connecting things together and enabling sort of, we would also call that coordination. Okay. Well, the argument.

07:11
which I think is true, is it turns out most things are not connected. I wouldn't have said that a couple of years ago. I'd said, oh, everything's digitized and when things get digitized, you can connect them.  Identity, merchants, consumers, people chatting, dating, whatever. Well, it turns out that's not really true. Like the way we connect things for the most part is by platform business models. And it turns out they're the exception.

07:41
that it turns out if you're going to connect a bunch of buyers with a bunch of sellers, merchants, you have to really standardize the inputs on both sides. And you have to sort of all be working on the same database. So, if you want to sell something on Lazada as a merchant, you have to input all your SKUs into their data fields and then those data fields are standardized.  everybody who's a merchant is putting their things in one format and then.

08:08
you know, the interaction, the algorithm can then go between that standardized format on one side and a consumer query, usually a keyword on the other side. You really need a lot of standardization of the data and it doesn't do well with messy types of data. Images are messy, video are messy, audio are most, most data in the world is unstructured. It's messy and crazy.  It turns out those things are

08:38
really hard to coordinate.  The information's all sloppy, we don't know how to do it. Well, it turns out generative AI is really good at taking different silos and different pockets of information, expertise, and enabling it to coordinate with other types of information that may be in completely different formats. And that's kind of what humans do. This is why people start thinking generative AI, and especially agentic AI, starts to look like a human.

09:08
If I'm on, let's say a website, don't know, Lazada or something like that or Taobao. Okay. I'm dealing with a sort of platform business model where I'm a consumer on one side and the platform itself has everything sort of siloed and appropriately standardized such that I can hunt for what I need.  But let’s say I also want to watch a video about the project or the product I'm buying. Well, then I got to go over to YouTube and watch it there.

09:37
and I watch the video there, maybe I watch some reviews. Ooh, is this a good phone or not? I watch a bunch of reviews, then I go back to Taobao and maybe I buy it there, or maybe I go on Reddit and I talk about, but I'm basically jumping between different ecosystems and platforms as a sort of human middleware to put it all together so I can decide what I'm going to buy. Those systems are uncoordinated except for humans jumping back and forth on tabs on their browser.

10:05
Well, it turns out generative AI is really good at that. You can give it control of your browser and it will jump from one tab to the next. It'll, well, I can't really watch videos very well yet, but it can read Reddit reviews. It can go on forums. It can go into Taobao or Lazada. can hunt there. It can do all that, put it all together and then say, okay, here's the one you should buy. And then if it's an agent, it can buy it. So, it turns out like platform business models,

10:33
as a mechanism of connection, coordination, were really kind of the exception.  Most of society, most of business is messy and uncoordinated. Well, it turns out generative AI is pretty good at putting those things together, kind of how humans do it. So anyways, that's kind of point number two. um and this is where tourism comes up. Tourism, online travel agencies, doesn’t work very well as a digital service.

11:01
You can go on a website or an app and look at booking.com and you can hunt through various hotels one by one by one. It takes a long time because they're very different. Especially if it's like Airbnb where you've got lots of different types of accommodations. But then you might buy your plane ticket there or you might buy it somewhere else. Usually, people bounce between multiple sites. You might watch videos on where you want to go on a travel.  When you get to a town you might look at, what is it, Viator.

11:28
You might look at activities you can do. You might talk to friends. should I do? when you go on a trip, it's super uncoordinated. Even though these websites, the OTAs have tried to sort of pull all that together. It just turns out the information's all messy. Everybody has a different sort of interest level. People want different things. They want to go to Paris differently than other people want to go to Paris. It's not really very standardized. So.

11:54
Sort of online travel is a good exception. It's like a good example of like when platform business models don’t really work very well. So that's kind of point number two and I'll get into tourism and how agents can sort of play out in that space.  Point number three, agents are a huge transformational disruptive issue within e-commerce. I mean it is just massive. Like it's my number one single question in life right now is,

12:23
how does generative AI and specifically agentic AI change commerce? Because it looks to me like it is going to be everything, like dramatic transformation. And there's a good slide, this is by BCG, where they basically had, I'll put it in the show notes, they basically say, look, retailers are going to get disrupted in normal ways. Now my standard sort of strategy is kind of same as everybody's, which is like, look,

12:51
If you're selling anything online, you're a retailer, you're a product, you're DTC, whatever you're doing, you're a YouTuber, you're putting out videos, you're driving them to your website to buy stuff, you've got a Shopify, whatever. It's kind of the same plain book for everyone. It's all sort of omnipresent, omnichannel, which is like, you got to get direct traffic, right? Either to your store within Shopify or Amazon or to your own website, but you got to drive traffic. And then you sort of have to...

13:21
convert that traffic into sales and in the process you're trying to get information about the customer, you're trying to figure out what they want and the more data you have, the better off you're going to be because you can start to tailor the messaging, the content, the bundles you might offer them, the products you might offer them. The whole CRM playbook is about hitting the right customer with the right offer or the right bundle at the right moment in time in the right manner. Well.

13:51
That's part of it. And then from there you would sort of iterate and you would improve your products and services over time. But you know, the game is getting the attention, get a direct relationship, gather data, increase your engagement, add touch points, try to personalize, try to bundle, try to cost sell, try to get them into your loyalty, your member program, all of that. Well, agentic AI kind of hits a, a sledgehammer on that whole strategy. If someone has an agent shopping for them.

14:21
That's who's showing up at your website. It's not the person anymore. You're sort of dealing with their assistant and you're no longer may have a direct relationship with the customer. All this game becomes very difficult if you don't have a direct relationship with your customers. You're not getting data about what they like.  You can't personalize. Your kind of getting cut off. So, here's the list from BCG, which is pretty good. What is the problem? They say there's several levels of impact.

14:49
Number one, you get a loss of direct traffic to retail websites. True.  You become dependent on generative AI search platforms to drive traffic and to do your sales. People are shopping within ChatGPT now. Then maybe they're clicking over at the end, but even then you're going to be able to buy within ChatGPT and not even go to the website.  You get reduced insight into customer behavior and intent.

15:17
due to the fact that you no longer have a direct customer relationship.

15:23
You lose the loyalty. You know, a lot of retail, e-commerce strategies about loyalty and memberships and retention.  you don't really have that. Agents have no loyalty whatsoever. And a lot uh of sales are based on emotion, on loyalty. If you're dealing with an agent and not the person, there's none of that.  You lose the opportunity to cross-sell.  Usually, bots are going to be very specific in what they're looking for.

15:53
And then you'll probably lose the ability, if you're doing media retail, you know, if you're doing advertising and stuff, maybe you got some ads on your webpage. Well, they aren't worth very much if most of your traffic is agents and not humans. So, this thing is just massive. Like it is a major potential disruption for a lot of businesses.

16:15
What else could that be? m Your revenue will probably go down. We already see that. If you look at Google search traffic going to an independent website, I have a website, your website traffic is going down because people are just getting their answers within ChatGPT and not clicking over from a search link. uh You have much less capability for personalization, for data gathering.  You're going to get...

16:42
kind of an intensified level of competition. It was bad enough when you were trying to become one of eight to 10 links that show up in a search.  That was already pretty competitive, but now, mean, a typical answer in a chat GBT, they're going to give you one to two options. So, the competition just went way up. Probably going to get a decreasing average order value, uh margin pressure. Yeah, so it’s a major deal. So that's something to think about.

17:12
And now let me get to point four here, which is kind of related to that is this idea of SEO and AEO. If you have a website, you're doing search engine optimization, right? And people are clicking over depending on what you're selling. If you're doing videos and stuff, your traffic is all human. But if you're selling socks, a lot of your traffic is going to end up being bots’ agents, not really bots’ agents. So.

17:41
how do you optimize your website? Or let's say you're doing content. I do a lot of content. Okay, it used to be I would show up in search engines and then people would click over and read my articles. But now it's an agent. And I need to not just show up in the search engine results, I need to show up within the answer that ChatGPT provides to whoever sent in the question. So that's the idea of AEO, answer, search engine optimization, SEO, agent.

18:11
engine optimization, AEO. Well, how do I do that? I'm already actually doing this. Like, whenever I put up an article on my website, I'll do a little search engine optimization. It's just all handled pretty much automatically. It'll put in the meta description. It'll put in the keyword phrase. It'll put in a three-sentence expert. That all gets logged into the thing. You put the backlinks and blah, blah, blah.  And everyone knows how to play the SEO game. Okay.

18:40
The way I've heard this described was by one of the founders of HubSpot. And he basically said, look, the webpage is no longer sort of the atomic unit of this world. It used to be search engines would search webpages. So therefore, you had to make your webpage, that's the atomic unit, appear a certain way and standardize the information such that, you know, it would search all the webpages and all their metadata, and then it would provide a list of ranked pages. So...

19:09
you know, the right atomic unit would probably be an 800-word blog with good SEO put into it.  Now, sort of the atomic unit is no longer the webpage, because it’s more like a clear answer to a specific question. It's more like a blurb. It's more like three sentences, because what does chat GBT do? Well, the first thing it does is it probably looks, it does a search.

19:35
So, it starts with a search engine. It does actually look for the pages that are highly ranked and therefore indicate a higher level of quality. But within those pages, then it synthesizes the information and then it basically generates a small, short answer to whatever question was put in and that's what goes within its answer. It doesn't give the link to your page for the most part.  So, the biddable unit, the atomic unit is not the web page.

20:03
What you want on your webpage, which is what I've been doing, my articles down at the bottom now have a series of questions, usually five to 10.  Specific questions with specific answers, one to two sentences each. Those are biddable atomic units that can be plugged right into the answer that ChatGPT might provide to somebody. And the clearer they are, the better. If they're a little punchy and interesting, that's actually better.  But they got to be

20:32
crystal clear answers to highly, very common specific questions that basically, chat GPT can just plug in to its answer. So, I've been sort of redoing some of my articles and at the bottom I'm starting to put five, 10, 15 uh questions with really clear little answers and the idea is those can just be grabbed and pasted in. So anyways, you got to start doing, and you still want to do the SEO, because you got humans, but you got to do the AEO as well.

21:01
So that's kind of point number four.  That's a good start in dealing with, you know, this sort of agentic world and generative AI world really. So that's point number four.  All right, point number five.  I'm going quick here because my voice is giving out as I'm speaking.  My framework for agents, which I've talked about before, is just ABCs. When it comes to commerce, there's assistants, brokers, and concierges, ABC.

21:31
So, you know, when I go online, I've been doing this. If you want to play with this, what you can do is you can go over to perplexity and download comment, which is a browser with an agent built into it. And you can basically just open tabs and do things in the browser and you can tell the agent what to do. You know, search for an email by this person. It finds it. uh Draft a reply telling them I can't make a meeting next week. I'm out of town, but how about the following week? And then send.

22:01
Usually, you want to check it first. And you can start playing with agents within a browser very, very easily. Well, those are all sort of basically assistants. So, I can do the same thing if I open up a webpage for let's say Shopee. I can have the agent just go in and hunt for what I'm looking for. Find me some Christmas ornaments that look like this and it'll hunt all through Shopee for me. And find and bring up the, now it doesn't work too seamlessly yet because I think Shopee doesn't want external agents.

22:30
crawling around, they want you to use their agent.  You can sort of play around with assistants.  and the analogy I use is that’s how billionaires’ shop at Walmart. They don't go themselves; they have their assistant go and buy everything. Well, we can all get assistants now. You can have one on your browser right now. If you don't want to use perplexity comment, can try uh Microsoft Edge has Copilot.  That works pretty well. Those are kind the most popular ones. It's good to play around with.  The C, Concierge.

23:00
It's the same thing, but from the merchant side. You, you click into Lazada or Shopee or whatever.  They're going to have their own concierge pop up to help you. What are you looking for? You know, how can I help you? And you start to have a conversation with it and it hunts around for you. Sort of like the concierge in a hotel is there to help you, you know, stuff within the hotel. But also, if you want to go down the street and go to a restaurant, it can help you with that too.  So, you sort of have agent, buyer side agents and seller side agents.

23:29
which I call assistants and concierges. And in the middle you have brokers, the B, A, A broker is sort of an independent agent that you can have go buy something for you, like health insurance. Most people use brokers to buy health insurance or their employer. So, you know, we've got sort of three types of agents that we're seeing in e-commerce and everybody is sort of fighting to be your best friend right now. They want you to become the agent you start to rely on.

23:57
I'll probably use about five to 10 agents. I've already got two or three I'm using every day. I'll start to do more. The capabilities are still sort of emerging, but they're getting better and better.  The difference between like generative AI and let's say an agent is an agent can access tools. It can access data. It can access payment protocols. You can start to talk with other agents. It can deal with websites. So, the capabilities of the agents, it's like they have a toolbox.

24:25
and depending on what type of agent you're using; they will have a different toolbox that they can draw on. So, some of them can do payments, some of them are good for planning travel, some are good for just going through your emails and your calendar, things like that. I'll probably task up about five.  Now that whole story, the ABCs, it's not a bad way to think about agents in commerce, not other parts of life. What's interesting is when you start talking of agents of agents, when

24:55
the tool that the agent can do is to deploy other agents, right? It's exponential. So, you could have three to four agents that you're using and in theory they could command a hundred others. That's where it starts to get weird.  So anyways, that's not here yet, but it's something people will talk about. All right; those were my main points. I'll get into the tourism case now, but I'll repeat them really quick. Point number one, think about four levels of digital machine learning.

25:25
Generative AI agents, that's pretty helpful. uh Point two, the big superpower of generative AI is coordination.  And yeah, it turns out platform business models were a bit of an exception, that most things have not been coordinated because they required standardization. Well, generative AI doesn't require that really. So yeah, that's pretty interesting. And then tourism is an example of something that people have tried to sort of get to work, never really did.

25:53
Generative AI might be the thing that makes it work.  Point number three, agents are a huge transformative issue within e-commerce.  Freaks me out to tell you the truth.  Point number four, think about doing SEO and AEO. Probably got to start doing both. need to, whatever business you're in, commerce, content, whatever, you got to start to think to the fact that you're going to deal with non-human entities as a significant potential part of your business.

26:23
And that's pretty weird to think about. And then the last one for agents for e-commerce, think ABC's, advisors, brokers, concierges. Okay, let me get into the tourism question.  Okay, so I'm going to read from a McKinsey article that was sort of digging into this subject, pretty interesting. They basically say, consider the example of a partner at a global law firm who needs to fly to New York to meet with colleagues. Okay, a London based partner, let's say.

26:54
Lots of partners at the firm make this trip all the time. Okay, so it's the same route against sort of clear company policies. So, you can see there's a sort of a degree of repetitiveness within this that is something AI is good at. However, uh the partner has a few specialized requirements, like she doesn't like to walk long distances. She maybe needs assistance to get to her gate to get up the stairs, prefers to sit in the aisle.

27:24
more elbow room to work. She likes to use the same flights to get her loyalty points. There's the visa issues and how she may or may not have to do anything for a visa prior or you know arrival documents and things like that. When she gets to New York she has hotel rooms, hotels that she likes and doesn't like. Within the hotels she likes queen beds for more space. She doesn't like to be on high floors. Maybe she doesn't like flights, things like that.

27:54
So, you can see even for a of a basic business trip, there's an interesting mix of sort of multiple steps, right? It's not like I'm looking for something to buy online, shoes, I find it, I buy it, it's not, no, this is a multiple step process throughout the transaction. There's a certain degree of repetitiveness of this.  The partners do this all the time, it's within the company.

28:21
You know, does this many, many times. But also, there's a high degree of personalization.  There could be also an evolving nature to the trip. When you get there, you might want to do something else. You might change your mind. You might want to go to the restaurant. You might want to go to another.  So, it’s kind of way too complicated for a simple website to handle traditionally. So, what they argue is basically, agentic systems which are, you know, kind of like humans.

28:48
are well equipped to deliver a trip such on this because it’s sort of high frequency happens over and over and over across basically fragmented disconnected systems. The hotel system is going to be different than the company policies, than the airline, than the airport management, then the taxis, you know, let alone like restaurants and other things you might do in there. So, you've got sort of fragmented systems.

29:17
The systems are a mix of structured data, credit cards, flight times, and a lot of unstructured data. I don't like that street. That street's not where I like to stay. I like to stay near Times Square. It's more fun. There's more, know.  So, you have this mix of sort of structured and unstructured. You've got a high degree of personalization. So, it's kind of an interesting.

29:43
scenario to think about in terms of can you have an agent do all this?  So, here's their takeaway.  The ideal agentic use case has a clear objective that can only be accomplished by using open-ended multi-step reasoning and decision-making. Right, that's the key thing with agents as opposed to generative APIs is they can make decisions. And in this case you're going to have to make decisions all the way along and the decisions may change based on what's happening. They go on.

30:13
Agentic AI thrives when presented with a high volume of tasks that are generally repetitive and basic structure, but still require personalization based on the context of a user individual person.  Workflows within travel, within hospitality, uh booking trips, things like that, hotels, they do have a pattern, but.

30:36
And so, you don't have to completely reinvent them every single time, but at the same time, you're not going to use a one size fits all standardized approach.  That's kind of, you the way I always think about it is just a human operating his middleware, you know, sitting on their computer with five tabs open and going between the tabs, watching videos about restaurants, looking at hotels.

30:58
Booking your thing, looking at the travel and visa information, looking at your company policies for reimbursement. You got five or six tabs open and you're bouncing between them. Well, that's what you want the agent to do. A couple other things to think about.

31:11
This is from the article.  Unlike traditional automation tools that can only access structured APIs, agentic AI can navigate user interfaces directly. They can basically use your browser.  Clicking through websites and legacy systems much as a human would.  This enables the integration of information in fragmented situations that previous technology simply couldn't handle.  By acting as connective tissue in this way, agentic AI can retrieve and update data across silos.

31:41
perform multiple workflows and save humans from doing sort of tedious interpretations between disconnected systems, the middleware thing. Last point, agentic AI tends to shine in situations that depend on retrieving and reasoning across different types of data sources. Structured data, flight schedules, pricing, booking that you need to execute, but also unstructured, transcripts, customer reviews, travel vlogs that are sort of

32:11
Crucial for understanding context, nuance, sentiment, that sort of thing. So anyways, that's the basic idea. But that's really all from the consumer point of view, the advisor, right? But what about brokers? What about concierges? What about people running hotels who are thinking about how to manage their internal workflows and handle customers? mean, if you show up at a nice resort,

32:36
your external agent is probably not going to be able to manage the internal operations of the resort.  They're not going to be able to move you from room to room, they're not going to be able to change your meal, might not be able to change your pillows because you're allergic to whatever, no.  In theory, you'll have internal agents of the hotel that do all of that. And suddenly you're talking about agent-to-agent interactions. And there's protocols for this now.

33:02
So, at a certain point, the agent within a hotel or the agent within an airline is going to take over parts of this and is going to coordinate with the external agent. You can kind of see how this is all going to play out.  And again, you can make the same argument for improving the workflows, the quality of service within a hotel because again, it's the same mix of structured and unstructured data. If you want to make things more efficient, you want to have your customer have a better experience.

33:28
Okay, you're going to look at things like check-in times and having the data right and making sure they can check in seamlessly. But you're also going to look at a lot of the contextual data. Oh, this person really likes this type of event. This person was here before and they talked to someone and they really liked that we gave them fruit for breakfast. You're going to capture a lot of that data. And agents in generative AI are good at synthesizing that with the more structured data to improve the experience and probably to make it much more efficient.

33:56
You could move people around in rooms much more efficiently and take care of the workload problems. So, you can kind of see it plays out in the A, it plays out in the C. Are we going to see independent brokers that are just basically agents for travel? Well, that might be the OTAs. Maybe, maybe it might be that these online travel agencies are going to become amazing.

34:21
and we'll be able to hand everything and they'll basically act as brokers between the hotels and the flights and all of that and the consumer. Maybe.  I think that's interesting. I would be sort of watching for the A, Bs and the Cs to all emerge. I think we're going to see all three. Okay, that is it for the content. My voice is shot.  I keep pausing to cough here. Yeah, I think I'm done. I think I've made the point. There's several concepts I think that are important to get familiar with regards to agents in general.

34:51
but also with this agentic e-commerce idea, which is kind of huge. And tourism is a nice example of some of the nuances and, you know, some of this probably will work out well. A lot of this will end up being, well, that didn't really ever happen in practice. You know, we'll see. This is all people theorizing about how it could be in tourism. We'll see. But I think there's enough there that I'm probably a 50 % believer. And then we'll see what happens in practice.

35:17
Okay, that is it for me. I got to go lay down here. My voice is gone and I'm getting a little dizzy. It's this flu that everyone's got.  Literally everyone I know in the world got the flu in the last week or so.  Every country, it was kind of surprising. My family got it in the US, then I talked to people in Europe and they had, and everyone got it like the same week. It's very strange. It was just suddenly everywhere. Anyways, that's it for me. I hope everyone's doing well and I'll talk to you in a couple days.  Bye Bye.