Connect with us


The week GenAI took off for travel



This week was easily the biggest week so far for generative artificial intelligence – especially for the travel industry. Let’s look at three of the big announcements.


On Monday, OpenAI introduced a new AI model called GPT-4o. The event was small and fairly low-key, and the updated products looked like mostly iterative improvements on many earlier versions. The demo was good, but we’ve been taught recently not to trust these demos. GPT-4o as large language model (LLM) looks like it’s the best model yet, though not a huge step up from the latest GPT-4. It’s free to chat for users, and 50% cheaper via API. It’s also much faster, which is important.

The other feature they focused on was the new voice-to-voice agent. 

The new voice agent has memory (which has been building for a while now), sounds lifelike and has almost zero lag time. It’s also completely hands-free. None of these sound like huge upgrades on their own, but the overall experience is revolutionary. 

You have to experience it to understand it. It’s just like speaking to a human. The intonation is perfect. You can have a normal conversation, at a natural cadence. But it’s an LLM. An LLM that knows everything. This is your 24/7 assistant, to help with everything in your life. Your business strategy, your sales prospects, your issues, your secrets. Or just ask useless football trivia – whatever you want. It remembers everything. Until you tell it to forget.


The Expedia Partner Conference has become a bit of a tech launch conference. This week, they did the on-stage CEO handoff to Ariane Gorin. 

The most significant product launch was Romie – your travel assistant/agent. The angle now is that Romie can join your chats with friends. Funnily enough, Expedia had a human-powered version of this in the “local expert” days. 

In this new product, you can invite Romie to your chat on Messenger, Whatsapp or Text. Romie (who needs pronouns, or no?) sits quietly and “monitors” the conversation, jumping in to help at the right time, with a full LLM of information, as well as some handy flight, hotel and activity options. 

It’s actually really clever when you think about it. The humans chat very naturally with each other, revealing all of their preferences for an upcoming trip. Romie steps in when it’s time to help with inspiration, discovery or transactions. 

What a perfect way to gather relevant personalized data for a specific trip. This is everything that personalization should be.

They did go and spoil it and say they’ll save those preferences for next time, so you can seamlessly bring your bachelor weekend plan and apply it to your honeymoon. Never mind. The core concept is great. It’s on Beta right now (EG Labs) – or you can just communicate from a message without even needing the app.


You’d need a month to properly look at everything Google launched Tuesday. And another month to work out what was just conceptual and what is live. Then two more months to work out which Google account has access to which tool on which browser login. 

All that aside, Google is up at the front these days, right alongside OpenAI/Microsoft. Everything looked impressive. 

Google Gemini already leads the pack for trip-planning, and they continue to talk about that. It occurred to me that they might not actually care about trip planning despite using it for demos at every event. 

The trip-planning solution is a proxy for fixing space and time. They have to fix space because humans move and because Google Maps is such a huge advantage, and possibly a trillion dollar moat. They have to fix time because humans value time. 

So producing the ideal trip-planning tool is a perfect way to demonstrate mastery of both space and time in one very relatable tool, which is fantastic in a 30-second demo. 

The problem for everybody in travel, is that even if it’s only a byproduct, it’s likely going to attract a huge portion of the funnel. This is fine though – people love paying Google.

All three companies spent a lot of time on agents. Right now, it’s “Agent in-the-loop” – a nice way to ease things in. The assistant joins the conversation amongst the humans. Everyone can bring their own agent if desirable. The agent just happily speaks when spoken to – no ego to feed. The humans are happy and unthreatened.

In the early 2000s, most websites were colorful and noisy. Google went with a white, clean, simple design. Ebay had a bright yellow background but wanted to modernize and switch to white. 

On the day they switched it, all of the humans (customers) screamed and shouted, so they immediately switched back to yellow. The engineers then programmed it to get 1% lighter every few days, and in a few months it was white. No fuss. Everybody was happy.

About the author …

Christian Watts is founder and CEO of Magpie.

Phocuswright Europe 2024

Join us in Barcelona to hear Christian Watts, CEO of Magpie, and Matt Barker, founder of Horizon Guides, discuss generative AI in travel during the Executive Panel: The Great GenAI Debate.

Continue Reading