Corporate Travel's Rulebook Is Its AI Booking Advantage
TravelEverything travelers hate about booking a business trip—the mandated tool, the approval chain, the policy caps—is exactly what let an AI agent complete that booking without a human clicking “confirm.” Amex GBT and Kayak for Business are proving it first, ahead of any consumer travel app.
What Happened
Amex GBT rolled out the Egencia AI connector inside Anthropic’s Claude on July 27, 2026, its first deployment of a proprietary agent-to-agent (A2A) architecture. Evan Konwiser, Amex GBT’s Chief Product and Strategy Officer, described the mechanics plainly: “an employee or an enterprise AI agent can now search, book and manage air and hotel for a traveler” without leaving policy constraints. The connector is landing in Google Chat for early adopters this month, full availability by year-end, with a Microsoft Teams pilot already running for Neo customers. SVP of Product and Engineering John Sturino framed the goal as channel-agnostic: “No matter what solution you’re on, Amex GBT intelligence is available to you.”
Kayak for Business took a different structural path. Rather than embed itself inside a third-party assistant, it shipped its own agent-to-agent connector: client companies plug their existing enterprise AI assistant into Kayak’s platform, and Kayak supplies the policy engine, live inventory, and transaction tools on the back end. The traveler’s assistant carries the context—who’s meeting whom, where, on what budget—while Kayak’s systems supply the rules and the ability to actually execute.
Why It Matters
The distinction most coverage of “agentic travel” misses is that completing a booking autonomously requires an agent to resolve ambiguity the way a human travel coordinator does: what’s allowed, what needs approval, what counts as compliant. Consumer travel AI has none of that scaffolding—it has to guess at budget, preference, and consent with no deterministic rule set behind it, which is why most consumer “AI travel agents” still stop short of the purchase and hand off to a human.
Corporate travel already had that rule set, built for compliance and expense-audit reasons that predate AI by decades: policy caps, preferred-vendor mandates, approval workflows, spend audit trails. Those constraints are precisely the input an autonomous agent needs to transact without oversight. What looked like corporate travel’s biggest usability liability turned out to be its readiness advantage for agentic AI.
It echoes a broader pattern in managed travel: infrastructure built for control is being repurposed as the foundation for automation. Hilton’s direct data feed into Navan, bypassing the GDS entirely, is the same dynamic one layer down—structured, rules-based data pipes are what let software (human-facing or agentic) transact reliably, while looser consumer-facing systems still require a human to sanity-check the output.
The So What
The more useful question for tracking this shift isn’t which company has the flashier agent demo, but who keeps ownership of the transaction. Amex GBT is meeting travelers inside AI environments they already use for work—Claude, then Google Chat, then Teams—while still completing the booking itself, per Skift’s reporting. Kayak is doing the opposite: opening its policy engine and inventory to a client’s own agent while keeping the booking experience inside its portal. Both bets assume the same thing—that the compliance rules corporate travel already enforces are the unlock for autonomous booking, not an obstacle to it. Consumer travel doesn’t have an equivalent rulebook yet, and until it does, expect the most capable AI booking agents to keep showing up in expense reports before they show up in vacation itineraries.
Source: Skift
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →