OpenAI has acquired Odessia, an AI agent startup for travel planning founded by Francis Davidson, the former founder and CEO of Sonder. The deal appears to be an acqui-hire: the Odessia service has already been shut down, deal terms are not disclosed, and Davidson is moving to OpenAI to work on 'the future of travel.' The acquisition strengthens the company's push toward agentic capabilities in consumer verticals, where it already has integrations with major booking services.

image
image

What Happened

OpenAI announced the acquisition of Odessia, an AI startup for travel planning created by Francis Davidson, previously the founder and head of Sonder. The deal amount is undisclosed, and the Odessia product has already been disabled, classifying the transaction as an acqui-hire: the team is transferring, not a working service. According to Davidson, Odessia raised $6 million in a pre-seed round led by Sequoia before shutting down and, by his own account, facilitated over $60 million in annual trip planning for tens of thousands of early users. OpenAI cited demand as the reason for the purchase, stating that ChatGPT is used by 1.2 billion people weekly, trip planning is among their daily tasks, and OpenAI's partners already include Booking.com, Expedia, and KAYAK.

Context

Francis Davidson is best known as the creator and long-time leader of the hospitality company Sonder, with Odessia being his next project at the intersection of AI and travel. The team built an agentic travel product not as a chatbot with recommendations, but through tools: workflows were assembled using MCP and CLI, and interaction partially involved voice prompts. This means OpenAI is gaining practical experience in tool usage and agent orchestration, tested on a live consumer vertical, rather than model weights or a recommendation system. The purchase fits into a broader shift: major AI labs are competing not just for models, but for vertical agentic products and teams, and OpenAI has openly named travel as one of the key consumer verticals for ChatGPT.

Why This Matters for the Industry

For the industry, the deal's value lies not in technology, but in the signal: it involves no API, pricing, or benchmarks, because OpenAI was buying people and their experience. This leads to three takeaways for builders. First, vertical agentic startups have received validation through two independent capital events: a round from Sequoia and an acquisition by the lab itself, meaning the direction no longer looks experimental. Second, self-reported metrics like pre-launch revenue are becoming currency paid for at exit, and the Odessia case sets a realistic acqui-hire scenario even at the earliest stages. Third, platform risk is increasing: if your agent operates in travel on top of ChatGPT, the platform has just hired a dedicated team for that exact space, so dependence on a single provider should be reduced in advance through a neutral tool layer and provider-agnostic integrations. Competitors are expected to respond with their own vertical agents or similar team acquisitions.

Why This Matters for Users

For readers, the key point is that trip planning via AI agents is moving from niche experiments to the mainstream. Currently, assistants help assemble routes and options, and integrations with booking services provide a bridge from recommendations to actual trips; the arrival of a team that built an agentic travel product with real revenue means that over time, selection, price comparison, and booking may be assembled directly within a ChatGPT dialogue, rather than in a separate app. The Odessia case is also indicative for those building their own agentic services: a small team with MCP and CLI-level tools and a voice interface brought the product to tens of thousands of early users. Those who used Odessia directly should note that the service has been shut down and no continuation of the product has been announced.

What Is Still Unknown / Limitations

There are many unknowns in the deal. The amount and structure are undisclosed, and available data does not suggest that OpenAI acquired or will support any Odessia product—the service has already been turned off. The figure of over $60 million annually is provided by the founder, refers to the period before the product launch, and lacks a public methodology or external verification; it is unclear what exactly 'facilitated' means—whether the user asked for advice, selected an option, or actually booked the trip. OpenAI's statement about 1.2 billion weekly users is a framework around demand, not a measurable assessment of how many people use ChatGPT specifically for trip planning. Finally, no timeline has been announced for the appearance of any travel features from Davidson's team in ChatGPT.

Sources

Author

Look at AI, editorial team