At the recently concluded fourth-quarter earnings call, Airbnb CEO Brian Chesky revealed his ambition to completely "AI-enable" the world's largest short-term rental platform. He not only announced that AI-driven search functionality has entered the testing phase, but also revealed that in the US and Canada, one-third of customer service cases are currently handled entirely by AI.
To accelerate this transformation, Airbnb even poached Ahmad Al-Dahle, a former Meta executive who was in charge of developing the Llama language model, to become its new Chief Technology Officer. All these signs indicate that the future Airbnb will no longer be just a "find a house" app, but an "AI travel concierge" that understands you and can help you plan your trip.
Say goodbye to keywords; AI search enables apps to understand "human language."
In the past, when we searched for accommodations on Airbnb, we usually had to enter the location and dates, and then check off a bunch of filter options (such as: has a pool, has a kitchen). But the new AI-powered search will completely change this experience.
According to the official statement, this feature, which is currently being tested with a small number of users, allows users to describe their needs using "natural language." For example, one could directly type: "I'd like to find a cabin suitable for family trips, with a large backyard for barbecues, and less than a 10-minute walk from the beach."
The system uses a large language model to analyze the meaning of the sentence, automatically matching it to the characteristics of the property, reviews, and location, and then accurately recommends it to the user. This means that users no longer need to set search criteria like operating a database, but rather feel like they are having a conversation with a professional travel agent.
Brian Chesky described this as a key step in the evolution from a "search engine" to an "AI-native experience."
Recruiting top Meta experts to create an app that "understands you".
To support this AI vision, Airbnb announced a major personnel appointment: Ahmad Al-Dahle will serve as Chief Technology Officer.
Ahmad Al-Dahle is no ordinary person; he worked at Apple for 16 years before joining Meta, where he led the team developing the Llama series of large language models—one of the most powerful large language models in the open-source world. Brian Chesky stated that under Ahmad Al-Dahle's leadership, Airbnb's goal is to create an app that doesn't just help you search, but truly "understands" you.
This suggests that Airbnb's future AI will possess personalized memory and inference capabilities. It might remember that you prefer a firm mattress, are allergic to cat fur, or prefer rooms with morning sunlight, and automatically take these factors into account during your next search.
Customer service automation has yielded remarkable results and significantly optimized the cost structure.
In addition to front-end search, AI has also delivered amazing results in back-end customer service.
Airbnb revealed that in the United States and Canada, one-third of customer service requests are now handled independently by AI-powered customer service assistants. These systems typically combine Natural Language Understanding (NLU) with automated processes to quickly handle standardized issues such as cancellations, changes, and listing inquiries.
Brian Chesky confidently stated that the quality of AI customer service will even "surpass human" in the future, and plans to extend this feature to all language markets with human customer service this year. For a two-sided platform like Airbnb that heavily relies on human resources, if AI customer service can handle more than 30% of cases, it will significantly reduce operating costs. This is one of the reasons why Airbnb has been able to continuously optimize its profit structure while growing its revenue.
Furthermore, Airbnb's engineers have fully embraced AI. Currently, 80% of engineers use AI-assisted programming tools (such as GitHub Copilot), with a goal of achieving 100% coverage, which coincides with the trend recently mentioned by Spotify's CEO.
Unafraid of Google and OpenAI, data is the biggest moat.
While investors worry that general-purpose AI chatbots like ChatGPT or Google Gemini might steal Airbnb's business, Brian Chesky sees it as a positive sign.
He pointed out that data shows that when consumers are directed to Airbnb through AI chatbots (rather than traditional search engines), the conversion rate is actually higher. More importantly, Airbnb has 2 million verified user identities and 5 million exclusive reviews, with up to 90% of communication occurring within the platform.
This proprietary data is a "moat" that external AI cannot easily crawl or replicate. General AI can tell you "which ward to stay in when you go to Paris", but only Airbnb's AI knows "which host responds the fastest and which bed is the most comfortable to sleep in".
Analysis of viewpoints
Airbnb's financial report and strategy announcement have given all online platforms a shot in the arm: AI applications in vertical fields may be more monetizable than general-purpose AI.
While general-purpose AI (like ChatGPT) may seem to understand everything, consumers still tend to return to dedicated platforms when it comes to "transactions" and "trust" (such as booking, payment, and dispute resolution). Airbnb's clever move is that it doesn't reject general-purpose AI; in fact, it has poached developers of general-purpose AI (such as the head of Llama) to optimize its vertical models.
From "search" to "conversation," and from "intermediary" to "concierge," Airbnb's AI strategy is very clear: use AI to reduce costs and increase efficiency (customer service, programming), while using AI to improve user experience (semantic search).
If Ahmad Al-Dahle can successfully combine Llama's technological expertise with Airbnb's massive database, the future Airbnb may become a super real estate agent with a "God's-eye view," not only knowing where you want to go, but also eliminating all potential pitfalls for you before you even set off.
