OpenAI has acquired AI-powered travel startup Odessia, with its founder Francis Davidson confirming that he will join the company to work on the future of travel.
The acquisition comes less than a year after Odessia was founded, during which the startup claims to have helped tens of thousands of early adopters plan more than $60 million worth of travel over the summer.
“Over $60M/year of travel was planned by Odessia over the past summer by tens of thousands of early adopters. Using AI to plan and book extraordinary trips is now a reality,” the founder said.
The founder also indicated that more details about the company’s future direction would be shared in the coming days.
He thanked the Odessia team, Sequoia, the angel investors who supported the venture, and its partners across the travel industry for backing the startup even before its official launch.
How Odessia Built an AI-First Business
Alongside announcing the OpenAI acquisition, the founder shared insights into how Odessia operated as an AI-first company, explaining the strategies and working methods that helped it build and scale its business in less than a year.
According to the founder, Odessia was established at a time when AI agents were becoming increasingly capable of handling complex tasks. Rather than adapting conventional startup operating methods, the company chose to build its operations around AI agents.
The founder previously established Sonder, which grew from zero to $600 million in revenue over a decade and required a workforce of more than 1,000 people.
With Odessia, however, he adopted a significantly different approach.
“This time, my approach with Odessia was radically different. We burned the old playbooks and radicalized ourselves to truly embrace the new agentic era,” he said.
He outlined six key principles that shaped Odessia’s operations.
1. Deploying AI Agents Across Business Functions
Odessia adopted AI agents across nearly every business function, moving beyond the conventional use of AI for software development.
The founder explained that agents equipped with relevant context, specialised skills, detailed voice prompts and advanced AI models could significantly increase productivity and, in some cases, complete entire projects in a single attempt.
The company deployed agents across strategy, partnerships, product design, marketing, customer support, legal and finance, among other functions.
2. Giving AI Agents Access to Organisational Context
Another important element of Odessia’s operating model was providing AI agents with access to the information available to employees across the organisation.
The founder described this approach as “context maxxing”, which involved connecting agents to company information through Model Context Protocol (MCP) integrations and command-line interfaces (CLIs).
These connections covered meeting recordings, Slack conversations, emails, Google Drive and other software tools used by the company.
Odessia used dozens of plugins, MCP integrations and CLIs to enable agents to retrieve information from multiple sources while working on a task.
3. Using Voice Prompts and Reusable AI Skills
Odessia also relied heavily on voice prompts to communicate instructions to AI agents.
According to the founder, agents still required corrections and feedback to deliver high-quality work. However, manually writing detailed prompts could slow down the process.
He said voice-based instructions were generally three to four times faster than typing.
Odessia also converted repeated corrections into reusable AI skills. These skills could then be automatically loaded by agents whenever relevant, reducing the need to repeat instructions.
4. Combining Iterative Workflows With Frontier AI Models
The fourth principle behind Odessia’s operating model was the use of advanced AI models and multiple rounds of iteration to improve productivity.
The founder explained that tasks requiring approximately 40 hours of manual work could sometimes be completed in two hours using agents powered by frontier models, even after several rounds of feedback and corrections.
He described this as a potential 20-fold productivity improvement.
According to him, the cost of running AI agents was relatively small compared with the productivity gains, making it worthwhile to increase usage rather than impose restrictive limits.
5. Managing Multiple AI Agents Instead of Individual Tasks
Odessia’s fifth operating principle involved managing several AI agents simultaneously rather than assigning and completing individual tasks sequentially.
The founder described a typical working day that began with reviewing the work completed by agents overnight, providing corrections through voice prompts and launching new agents to handle additional tasks.
Instead of waiting for one agent to finish, employees could initiate multiple tasks and work across different areas simultaneously.
Odessia also developed proactive agents capable of initiating tasks without requiring employees to manually launch them.
These included:
- Daily analytics reports, including new analyses beyond standard dashboards.
- Investigation of every production error.
- Reviews of customer feedback and user interaction traces.
- Accounts payable matching.
- Updating Linear following daily stand-up meetings.
- Overnight smoke tests.
6. Maintaining a Faster Execution Pace
The final principle was maintaining a rapid pace of execution.
The founder said that although AI agents significantly increased productivity, the result was not necessarily a reduction in the effort required from employees.
Instead, the increased capacity to complete work made every working hour more valuable.
He noted that Odessia was among the startups operating six or seven days a week, explaining that the team wanted to avoid losing ground to competitors.
Odessia Acquisition Marks a New Chapter for AI-Powered Travel
Reflecting on the experience of building Odessia, the founder described the transition to an AI-driven operating model as an opportunity to bring ideas to reality much faster than traditional business processes allowed.
He said projects that might previously have taken months or years could now move from an initial idea to real-world implementation within hours or days.
The founder also questioned the limitations of conventional business roadmaps, highlighting the trade-offs involved in prioritising projects and the difficulties of executing even relatively simple ideas through traditional processes.
“The bells and whistles that could never make their way onto the roadmap will become a competitive imperative,” he said.
