OAG argues aviation's AI agents will answer to developers, not travellers, as MCP spreads

OAG has published an analysis arguing that the rise of the Model Context Protocol (MCP) — an open standard Anthropic released in November 2024 for connecting AI systems to external data — is reshaping who aviation data providers actually need to serve. Author Ben Maynard frames MCP's value in efficiency terms: rather than each AI application needing a custom integration with each data source, a shared protocol turns an N-times-M integration problem into an N-plus-M one. He points to Expedia, Booking.com and TripAdvisor's decision in October 2025 to build AI integrations inside ChatGPT on top of MCP as evidence the approach is already being adopted at scale, and cites OAG's own operational load — some 400,000 flight-schedule changes processed daily — as the kind of scale such integrations need to handle. Maynard notes that today's AI travel tools mostly function as discovery layers rather than transactional ones, since PCI compliance, legal agreements and liability concerns still push users back to a human-facing platform to actually complete a booking. He also cites Anthropic research showing that a well-optimised MCP implementation cut token consumption by 98% compared with a naive one, from roughly 150,000 tokens down to 2,000, as a sign of how much of the current overhead in AI-agent integrations is avoidable engineering waste rather than an inherent cost. His central claim is that "the developers building the agents are the channel" now, not the end traveller directly — a shift he argues aviation data providers should design their APIs and access models around.