In corporate travel, the agent picks the hotel now

TMCs are wiring their booking tools for AI agents — and for hotels, eligibility now gets decided before a person ever sees the list

For a corporate traveler, building the hotel shortlist has always been a human act: open the booking tool, scan what fits the trip, pick one. That act is moving to the machine. As travel management companies connect their platforms to AI agents, the agent assembles the set of eligible hotels — filtered to policy, traveler profile, and rate — and the person chooses from what the agent has already decided to show. For a hotel, the consequential moment now happens before anyone looks.

Driving the news. BCD Travel, one of the three largest TMCs, has deployed Model Context Protocol (MCP) across its TripSource platform. By the company's account, the framework lets AI systems reach shopping, booking, policy, and spend data through a single interface in plain language rather than through bespoke integrations. The practical effect BCD describes: instead of handing a traveler hundreds of options, an agent presents only the hotels that already clear corporate policy and the traveler's preferences.

The pattern. BCD is not alone, and that is the part hoteliers should register. American Express Global Business Travel rebuilt Egencia around agentic search this year, with a conversational assistant living inside Microsoft Teams. Spotnana, the infrastructure layer beneath a growing roster of TMCs, has shipped its own agentic features and keeps adding partners onto the same plumbing. The corporate booking stack is converging on a model where an agent does the shopping. Managed travel — structured, policy-bound, data-rich — is the first place that becomes the default rather than the experiment.

What it means for hotels. Hotel spend is roughly 30% of managed corporate travel budgets (GBTA), and enterprise programs route most of it through booking tools (around 82% adoption). When the shopping inside those tools shifts to an agent, the qualifying step shifts with it. A property gets considered only if its rates, content, and policy terms are clean enough for a machine to read and match against a corporate program. Gaps a human booker would have squinted past — a missing rate load, an ambiguous cancellation term, thin structured content — now read as non-compliance, and the agent moves on. The hotel is not rejected on price. It is left off the list the traveler sees.

This rewires where corporate visibility is won. For years the work was ranking and rate parity inside a booking tool built for people. The work now is being selectable by the agent that shops for the traveler — accurate, machine-readable, policy-matched, every time. In managed travel, that selection is the booking, and it is settled before the traveler looks.

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →