The Travel Platform That Ditched Old Processes

Omio, a multimodal platform coordinating over 3,000 transport providers across 47 countries, decided to scrap outdated processes. CTO Tomas Vocetka mandated all internal functions to redesign their operational frameworks from the ground up to operate as a native AI enterprise — not just layering technology on top. The first step: giving every employee access to ChatGPT to build familiarity. Then OpenAI Codex was embedded directly into engineering operations, mandatory across the entire software development lifecycle: from preliminary research and architectural planning to coding, automated testing, code reviews, and maintenance. Developers build custom connectors linking proprietary data with the tools, allowing them to bypass information retrieval and jump straight to execution. The result: the effort to build specific products dropped to roughly 20% of previous levels. Projects that once required several developers for a full quarter now take a single engineer about a month.

Conversational Commerce: Booking Travel by Speaking

In 2023, Omio launched one of the earliest conversational travel booking interfaces by connecting OpenAI models to its transportation inventory. The system parses natural language requests for complex multimodal routes — like “fastest way from Rome to Florence” or “compare flights and trains between Paris and Barcelona.” Omio aggregates trains, buses, ferries, and flights, replacing the fractured web of traditional sites with a unified interface that interprets intent. Generative models analyze text and query booking systems to construct directly bookable itineraries, grounding responses in live pricing and availability to avoid outdated suggestions. The integration was later extended into a dedicated ChatGPT experience accessing the company’s global network. Omio dubs this “conversational commerce”: the AI becomes the primary layer between consumer and provider network, signaling a move beyond search-based interfaces toward generative customer experiences.

Governance: Humans Stay in Charge

Despite deep automation, corporate policy insists that human personnel retain full accountability for all deployed code and business outcomes. AI tools serve strictly as acceleration engines. “The responsibility and accountability stay with people. AI helps us develop faster, analyze faster, and make decisions faster, but people stay in charge,” Vocetka explains. This governance prevents automated systems from executing irreversible changes to the booking infrastructure or core routing algorithms. The mix of broad AI access and strict oversight creates an environment that balances speed with systemic stability.

What This Means for On-Premise Evaluations

Omio’s case demonstrates how cloud API adoption can slash development timelines, but it raises questions about data sovereignty and vendor lock-in. For a company handling real-time travel data at global scale, using external models means considering where data resides and how it’s processed. Omio’s decision to keep critical decisions under human control serves as a counterbalance, yet technical dependency remains. Replicating such integration in a self-hosted environment would involve significant trade-offs: dedicated hardware and serving frameworks like vLLM or TGI bring infrastructure costs and management complexity that can erode speed advantages. AI-RADAR explores these compromises in its guides, offering frameworks to assess when on-premise deployment truly makes sense. Omio’s cloud-plus-governance path points to a pragmatic direction: maximizing AI’s competitive advantage without relinquishing control.