Something Is Changing Underneath the Surface
Corporate travel has spent the last few years absorbing ever-evolving AI tools. Chatbots, automation workflows, agentic assistants.
But there’s a more foundational shift happening underneath all of it. One that isn’t about which AI tools you choose. It’s about how those tools connect to the data they need to carry out workflows on your behalf.
That shift has a name: Model Context Protocol, or MCP. It’s the connectivity standard that lets AI agents plug into external systems, booking tools, policy engines, profile databases, without needing a custom-built integration for every single one.
Anthropic introduced MCP in 2024. Adoption across the tech industry has moved quickly since, and corporate travel is now firmly in its path. Oversee’s co-founder and CTO Ami Goldenberg joined BTN’s April News Desk alongside industry experts Steve Clagg and Gee Mann to unpack what it means in practice.
The Promise: Data That Was Siloed Can Now Be Reached
Start with the problem MCP is actually solving. Right now, getting an AI agent to interact meaningfully with a travel program requires stitching together connections to multiple different systems, each with its own integration requirements. Profile data lives in one place. Policy in another. Preferred rates somewhere else. Every connection needs work.
MCP changes that by establishing a common interface. An agent platform that implements MCP can connect to any system that has an MCP server and, critically, discover what that system can do. As Ami put it during the session, that discovery piece matters. An LLM that can explore an API and learn what it can and can’t do is a fundamentally different kind of tool than one that only executes predefined instructions.
For corporate travel programs, the practical upside is real. Data that previously required significant technical resource or vendor roadmap dependency to access becomes reachable through a consistent, structured interface.
The Catch: Connectivity Doesn’t Fix Bad Foundations
MCP makes it faster and easier to reach data. But it doesn’t improve the quality of that data. If your traveler profiles are incomplete, your policy is inconsistently documented, and your preferred rate structure isn’t clean, an AI agent working through MCP will surface those problems faster, not work around them. Steve Clagg put it plainly: “MCP can accelerate both good outcomes and bad ones.”
The companies that will get the most out of MCP are the ones that treat it as a reason to get their data house in order now, not later.
NDC Isn’t Going Away. Neither is the GDS. But Both Have to Adapt.
Two questions that came up repeatedly in the BTN session: does MCP make NDC obsolete, and what does it mean for the GDS model?
On NDC: no, it doesn’t make it obsolete. NDC solves a content problem, letting airlines distribute richer offers through intermediaries. MCP solves a connectivity problem, letting AI agents reach those systems through a standard interface. They operate at different layers and will coexist in most architectures. Ami’s view was that MCP should push airlines and GDSs to simplify how they expose their APIs, making NDC content more accessible, not less relevant. Steve’s practical take: don’t use MCP as a reason to defer NDC decisions. The value of NDC content your TMC is delivering today is real, regardless of where agentic AI lands in three years.
On the GDS: the panel was more pointed. If airlines and hotels expose their own MCP servers directly, AI agents can reach them without routing through a GDS at all. The transaction-fee model that GDS businesses rely on faces genuine structural pressure. Gee Mann pushed further: if an IT department can whitelist MCP connections and those agents simultaneously query a TMC, a GDS, an OBT, and an airline direct connect, pricing transparency across every channel becomes unavoidable. “It’s not just a technology evolution. There is also a commercial model evolution.”
The GDSs that adapt by opening up and repositioning as trusted aggregation layers have a strong hand to play. The ones that respond by building locked-in, proprietary MCP setups risk fragmenting the standard and losing ground faster.
TMCs Are Better Positioned Than They Might Think
The TMC’s position in an agentic world isn’t under threat. If anything, it gets stronger. When a traveler interacts with a TMC through an AI agent, they don’t just get a flight search. They get policy enforcement, profile management, duty of care integration, preferred rates, and payment handling, all in one place. A GDS or airline MCP server can’t deliver that. The TMC can, because the TMC knows who the traveler is and what their program requires.
Ami made the case directly: the TMC’s MCP layer can provide a richer, more complete experience than any single supplier endpoint. Steve framed the strategic opportunity: TMCs sit at the intersection of traveler profiles, policy, preferred rates, and servicing relationships. Those that build MCP infrastructure around that intelligence, exposing a genuine policy engine alongside negotiated content, become the intelligence layer that AI agents rely on.
What to Do With This Information Right Now
Audit your data. Not just understand where it lives, but actually assess whether it’s clean enough for a machine to act on. An AI agent working from inconsistent profile data produces an inconsistent traveler experience, at pace.
Get your policy into machine-readable formats. Intent documented in a Word file works for a human agent who can interpret it. An LLM needs structured, deterministic rules to enforce policy reliably. Starting that conversation early means your program is ready when the tools arrive.
The Stack You Build On Matters
Oversee operates as an AI optimization layer running quietly inside the workflows TMCs already use. Across air and hotel, the platform continuously monitors booked travel, identifies savings opportunities, and acts on them automatically, without rebuilding operations or adding manual work. AgentSee extends that same logic to agent productivity, automating ticket handling and service requests so teams respond faster and scale without scaling headcount. Travel Sourcing Optimization adds a further layer of intelligence, auditing supplier contracts, benchmarking spend, and surfacing the data TMCs need to negotiate with confidence. Together, the products give TMCs a single embedded intelligence layer that captures value across the full travel lifecycle.
As MCP becomes the standard interface between AI tools and travel data, the value of having clean, well-integrated optimization layers already in place only compounds. The TMCs building toward that future now are the ones the panel described: clean data, clear policy, and the right technology already embedded in the right parts of the stack.
Curious how Oversee fits into your existing setup? Book a walkthrough.