Glenn Fogel

1. Key Themes

Booking Holdings Is Already Executing on Agentic Travel — Not Just Talking About It

Glenn Fogel is not theorizing about AI agents in travel; he is deploying them now. Priceline's AI agent "Penny" is handling complex, multi-leg, multi-passenger itineraries with real-time trade-off analysis (miles vs. cash, routing, hotels, ground transport), and adoption has doubled every month for several months.

"I can't wait until we, Booking Holdings in our companies, are offering up these personalized agents that are, that know everything about you, everything you want, and able to do so much more for you than any human travel agent could ever do. Because the machine never forgets anything." 00:16:22

Scale as a Structural Advantage That Outsiders Systematically Underestimate

Fogel repeatedly returns to scale as the thing newcomers miss. $186 billion in travel transacted, over 1 billion room nights, 8.6 million alternative accommodation listings — these create network and supplier density that cannot be replicated by an AI wrapper. He also adds regulatory complexity as a hidden moat reinforcer.

"If you think you're just going to come in and do this business and knock away these very big players, I'd say you should really understand what the business is before you decide to commit your capital." 00:00:24

Token Economics and ROI Discipline Are the Real AI Bottleneck Inside Large Enterprises

Fogel's candor about why Penny isn't being pushed aggressively at scale reveals a non-obvious constraint: the per-session token cost vs. lifetime-value math is not yet solved. Large-scale AI deployment in high-transaction-volume businesses requires rigorous cost-per-interaction accounting before full rollout.

"How many tokens are we consuming? Where? How many times are they coming back and forth? Tell me how much was the cost of us getting that trip for that person? And what is our ROI going to be? And the next thing is, what's the return long-term? Did it mean lifetime value?" 00:20:43

AI in Customer Service Is Already Generating Measurable, Dual-Sided Wins

Booking is already reporting a 10% reduction in customer service cost per reservation while simultaneously increasing customer satisfaction — a rare case of cost reduction and quality improvement moving together, not against each other.

"Our costs for customer service per contact are down. That's great. Customer satisfaction is up. That's even better." 00:22:57

The AI Boom Rhymes Structurally With the 1999–2000 Internet Boom

Fogel, who joined Priceline the week the NASDAQ peaked in March 2000, draws direct parallels between that era and today: euphoria, speculative capital, technology that is genuinely transformative but with a coming washout of most companies.

"You think like this is a plummet explosion of new things, everything just like in the late 90s. The optimism of technology is going to be wonderful. But then you get a little bit of the backlash... The numbers are much bigger. The issues at hand are much bigger." 00:07:37

Booking's Alternative Accommodations Business Is Globally Near-Parity With Airbnb — Almost No One Knows This

Elad flags this as underappreciated and Fogel confirms: Booking's alternative accommodation transaction volume is approximately 75% of Airbnb's, with 8.6 million listings globally, and has been growing faster than Airbnb for five consecutive years.

"When you look at our total amount of transactions on room night for alternative companies, you see that we are approximately three quarters the size of Airbnb. And that's just our alternative accommodation area. We're the whole much bigger hotel business on top of that. And over the last five years, we've been growing faster than Airbnb in the alternative accommodation area." 00:26:56

Supplier-Side Relationship Complexity Is as High a Barrier as Consumer-Side Scale

Fogel emphasizes that the supplier (partner) side of the marketplace — thousands of direct relationships with hotels and property managers, yield optimization consulting, systems integration — is as hard to replicate as consumer-facing scale.

"It's not just getting the inventory loaded into some database. Anybody can do that. That's nothing. We've got thousands of people who are dealing with hotels and other property managers. How can we do your business better? What do you need? Where do you need more demand? Where are you hurting?" 00:28:49

Capital Allocation Discipline: Return Capital Unless You Can Beat the Market With It

Fogel has bought back approximately 40% of outstanding shares over the last twelve years and frames the logic in stark investment banking terms: reinvest only if your ROI clears the bar, otherwise return it.

"If you can't do either of those, then get the money back to the shareholders because they can then invest it better than you can. And that's what I've always believed in." 00:24:37

The Speed Mismatch Between Job Destruction and Job Creation Is the Real AI Risk

Fogel is more concerned about the pace of displacement than displacement itself, and he draws on a first-hand example: translation jobs at Booking.com disappeared entirely as machine translation replaced 40-language human translation teams. He warns that if rejection of AI gains political momentum, the U.S. will be competitively disadvantaged versus China.

"We know also the other side, that new jobs are going to be created. We all know that also by history... but the speed of job disappearance and new job creation. Those rates are not happening probably at the same rate." 00:35:58

2. Contrarian Perspectives

There Is No Such Thing as a Moat — Not Even at a $130 Billion Company

Most people assume that at Booking Holdings' scale the moat is self-reinforcing. Fogel explicitly rejects this framing and uses it as a management philosophy, not false modesty.

"There is no such thing as a moat. There is no such thing as somewhere you're going to be protected against innovation. Today, we have a competitive advantage on areas. Absolutely. But those can go away tomorrow." 00:00:00

OpenAI Exiting Checkout/Travel Commerce Was Not Necessarily Bearish for AI in Travel

The market read OpenAI canceling its checkout feature as good news for Booking (+8% on the day). Fogel argues both reactions were overreactions — the underlying agentic trend is intact and is proceeding through incumbents like Booking, not being abandoned.

"Then people say, oh, well, I was wrong. I'm not going to worry about it as much. It flips the other side. The truth is, the way we look at AI is an incredible, beneficial tool and a way for us to be able to do our mission easier, cheaper, and better for our customers." 00:13:05

Government Retraining Programs for Displaced Workers Have a 50-Year Track Record of Failure

While the conventional policy response to technological unemployment is government retraining, Fogel dismisses it directly based on observed history, placing the obligation instead on employers.

"Retraining by governments over the last 50 years really hasn't worked out so well. So I'm not sure that's the right way to go either." 00:38:19

Consumer Sentiment Surveys on AI Are Deliberately Misleading — People Actually Love It

Elad Gil surfaces research showing that framing entirely determines the result: ask people if they enjoy using AI tools and they say yes; ask if AI will destroy their jobs and they say no. The "anti-AI consumer" narrative is a survey artifact, not a real finding.

"It turns out it depends on what the question you ask is. If you ask people, do you love using ChatGPT and Gemini and Claude and all that, they're like, we love it and we'll pay more for it and it's wonderful." 00:39:15

Wealthy People Already Have AI-Level Travel Planning — It's Called a Human Concierge

Fogel reframes the AI travel agent not as a futuristic concept but as the democratization of something the ultra-wealthy already buy. This reframe has significant investment implications for who the addressable market actually is.

"Many people would like something else to do for them. In fact, that's why you'll find very wealthy people have travel concierge people who are actually human beings who really understand the needs of that customer. People aren't quite in that wealth zone." 00:15:22

3. Companies Identified

Booking Holdings

Global online travel company (Booking.com, Priceline, Kayak, Agoda, Rentalcars.com). Mentioned as the primary subject of discussion — $186 billion in travel transacted last year, over 1 billion room nights, 8.6 million alternative accommodation listings, approximately 40% share buyback over twelve years. Market cap peaked around $180 billion.

"Last year we did $186 billion with the travel. That's a lot of travel. We did over a billion of room nights." 00:20:15

Priceline (Penny)

Priceline is the Booking Holdings brand that houses the AI agentic travel assistant "Penny." Penny adoption has doubled every month, delivers lift in conversion, faster search, shorter path to booking, lower cancellations, and higher customer satisfaction.

"We are doing it right now. In fact, if you go to Penny, which is Priceline's authentic AI system. And I just did it the other night. I put in a very complex need for a travel with the family." 00:16:22

Airbnb

Online alternative accommodations marketplace. Mentioned as the reference point for Booking's alternative accommodation business — Booking is approximately 75% of Airbnb's transaction volume and has grown faster than Airbnb for five consecutive years.

"When you look at our total amount of transactions on room night for alternative companies, you see that we are approximately three quarters the size of Airbnb. And that's just our alternative accommodation area." 00:26:56

OpenAI

AI lab and ChatGPT developer. Mentioned because its decision to cancel its checkout/travel commerce feature caused an 8% single-day move in Booking Holdings stock, and because its agentic systems are cited as both a potential threat and potential channel.

"OpenAI had checkout in ChatGPT. And one of the main use cases was travel. And then they canceled that feature. And I think at the time, booking went up 8% on the news." 00:11:33

Anthropic

AI lab and Claude developer. Mentioned by Elad Gil as having a rumored $30–$50 billion revenue run rate, alongside OpenAI, as examples of AI companies with genuine revenue bases versus those with high valuations but no revenue.

"OpenAI and Anthropic are rumored to have $30 to $50 billion in revenue run rate each." 00:08:06

Decagon

AI-native customer support agent company. Cited by Elad Gil as an example of the emerging category of agentic work being deployed in specific enterprise functions.

"That could be specific companies like Decagon having agents that do customer support." 00:14:37

Google (Gemini)

Mentioned alongside OpenAI and Anthropic as a major platform developing increasingly self-driven agentic AI systems that could interact with or compete in the travel booking layer.

"The larger platforms like OpenAI, Anthropic, Google, etc. providing increasingly self-driven systems, Codex or Claude Code or some of the things Gemini is doing." 00:14:37

4. People Identified

CEO of Booking Holdings. Twenty-seven-year veteran of the company, joined Priceline in 2000 when it was worth a few hundred million dollars, helped grow it to a peak market cap of approximately $180 billion. Background spans IT operations, investment banking (Wall Street until 1995), trading at Morgan Stanley, and corporate development.

"Our stock went from where we were — when we went public, in a week or so, we were worth $30 billion. By the time I joined, we're probably down to about $15 billion. In nine months, our market cap is now down to just a couple of hundred million. And in last summer, we came very close to $6,000 [per share]." 00:05:49

Elad Gil

Co-host of No Priors, investor and entrepreneur. Surfaces key data points about Booking's business (Penny adoption doubling monthly, $550M–$700M investment in tech, $4 billion Q1 return to investors) and provides the AI industry market context including the internet-era comparison of ~1,500 IPOs of which roughly two dozen survived.

"Something like 450 companies went public in '99, 450 went public in the first few months of 2000... You had 1,500 companies of which what, two dozen are left at most?" 00:08:06

5. Operating Insights

Proactive Disruption Management Before the Domino Falls

Fogel's articulation of AI in customer service goes beyond reactive support — his stated goal is a predictive system that identifies travel disruptions before they cascade and re-routes customers proactively. This is an operational north star that any high-transaction-volume service business should be building toward.

"My goal is to have a system that actually we are able to predict well enough what the problem will be before it happens. And so just changing things... travel is like dominoes. One thing falls over and it all starts falling over." 00:19:11

Upskilling Employees in AI Is Both a Retention Tool and a Business Obligation

Fogel frames mandatory AI literacy training not as a defensive HR cost but as a dual-benefit investment: it raises internal productivity and creates portability for the employee if their role is displaced, which in turn creates goodwill and reduces attrition during AI transitions.

"Even if we end up that we can't replace or retrain or put someone else, at least they are better skilled for a job somewhere else. And I feel a real obligation for that... It's good for our company. It's a positive for our life for somebody to be able to use new tools in a better way." 00:37:21

Model Selection and Token Economics Must Be Managed at the Feature Level, Not Company Level

Fogel's operational concern is not which LLM to standardize on but rather which model is right for which specific task and at what cost — a more granular level of AI procurement discipline than most companies are currently applying.

"The whole thing of token economics now — which model should we be using? For which purpose? And when? And obviously, you can get tokens a lot cheaper. It's a lot cheaper. Different models can be a lot cheaper. And that's something that we have to look at very closely." 00:21:13

6. Overlooked Insights

Machine Translation Eliminated an Entire Job Category at Booking.com — And Nobody Noticed

Fogel casually drops a real-world example of AI-driven total job category elimination that has already happened inside Booking.com: all human translation roles across 40 languages were eliminated as machine translation matured around 2005 onward. This is not a hypothetical. It is a completed case study of a white-collar, language-specialist job category going to zero at scale inside a single company — and it passed with essentially no public controversy or policy response.

"In 2005, all the translation are being done by human beings. There's no machine translation at all. And all those customer service human beings had people in all these 40 different languages. That was a big deal. Now, we have machine translations. All those jobs are gone. There's nobody doing that." 00:35:01

The investment implication: language services, localization firms, and any business whose core value proposition is human multilingual content production is already in secular decline. The AI job displacement debate is treated politically as a future risk; it has already run its course in at least this category.

Booking Holdings Is Quietly Running a Yield and Demand Consulting Practice for Hotel Partners at Scale

Embedded in Fogel's description of the supplier relationship is something that is not marketed or widely understood: Booking has thousands of employees engaged not in listing properties but in active demand optimization consulting with hotel and property partners — effectively acting as a revenue management advisory firm at scale. This is a B2B services layer sitting inside what most investors model as a pure marketplace.

"We've got thousands of people who are dealing with hotels and other property managers. How can we do your business better? What do you need? Where do you need more demand? Where are you hurting? What can we do in terms of your systems better?" 00:28:49

This embedded advisory relationship creates switching costs and data network effects on the supply side that are entirely separate from — and arguably more durable than — the consumer-facing brand. It is also a function that AI can enhance rather than displace, making it a compounding rather than depreciating asset.