BackPerspectives

When AI books the trip, trust is the product

AI will book more travel. Winners won’t be clever chat UIs—they’ll be companies trusted to keep promises on price, booking reality, changes, and care.

5 min read

The next click disappears

We’re about to cross a line in travel where the click drops out of the flow. I won’t comparison‑shop for an hour and then type my card into three different forms. I’ll say what I need—dates, constraints, preferences—and an assistant will plan and book end‑to‑end. The interface will look like a conversation, but that won’t be the differentiator. Trust will be.

Trust means the price I’m quoted is the price I pay. The booking that shows up in my app actually exists. If plans change, the system handles it without turning my day into a support ticket. And when judgement is needed, a human can step in with the full context—no retelling, no loops.

I’ve built software long enough to know that the hard part isn’t the chat. It’s keeping promises through messy reality. In travel, that’s where advantage lives.

Before the trip: clarity and finality

When an assistant books on my behalf, the “pre‑trip” experience is where trust is made or lost.

  • One truth about money: I should see a clear total, including what can change and why. If a fare isn’t final until it’s ticketed, say so plainly and close the loop fast. Promise integrity beats a lower number that moves later.
  • One itinerary, many dependencies: Flights, hotels, ground, visas, insurance—each has its own rules. I shouldn’t have to learn them. A trustworthy system explains the constraints briefly and coordinates them behind the scenes.
  • Clean confirmations: A confirmation number isn’t a trophy; it’s a commitment. The itineraries in my calendar, email, and wallet should match the supplier’s reality. No “phantom holds,” no ghost segments.
  • Straight lines to finance: If this is a work trip, the ledger should accept the booking without translation. Policy and approvals should be visible up front, not rediscovered after the swipe.

None of this requires magic. It requires ownership of the model behind the interface. In a previous essay I argued that the future of travel is architecture, not inventory—the winners treat a unified system as the product, not the list of things they can sell. See: The future of travel is architecture, not inventory.

During the trip: quiet competence

The best travel systems are quiet when they can be and decisive when they must be.

Quiet looks like this: my seat and meal preferences stick; my loyalty numbers are applied; my hotel knows I’ll arrive late; my ground transfer waits because the inbound is delayed. I don’t notice coordination because it happened before I needed it.

Decisive looks like this: when a connection risks misconnecting, the assistant offers a re‑route that respects my constraints (no red‑eyes, aisle seat, arrive before noon)—and executes after a quick confirmation. When a hotel oversells, I’m already re‑accommodated nearby on a comparable rate with my preferences noted. I still want the option to say “talk to a person,” and that lane should exist without friction. But the system should default to taking care of me, not asking me to become the orchestrator.

When something breaks: judgement with a human lane

Travel breaks for boring reasons: weather, crew timeouts, maintenance, oversells, strikes. In those moments, trust is the product.

  • Authority to act: An assistant should have clear authority (within policy) to rebook, cancel, separate travelers’ bookings when plans diverge, and arrange alternatives without trapping me in approvals.
  • Context that travels: If a human steps in, they should see the full state—what was booked, what changed, what policies apply, what options were considered. I shouldn’t re‑explain my own trip.
  • Make the aftercare automatic: Refunds, credits, and receipts shouldn’t require me to chase. The promise is “we’ll make you whole, and you’ll see when it’s done.”

Judgement is the dividing line. Machines can propose, but people decide under uncertainty. A trustworthy platform makes it easy to escalate the decision while keeping the agentic system doing the work around it.

What businesses must get right for agentic AI to be dependable

If assistants are going to plan and book on behalf of customers (and other businesses), travel companies need to become dependable components. That has little to do with a witty chat tone and everything to do with the plumbing beneath it.

  • Clear source of truth: One consistent model of a trip across products and surfaces so the assistant isn’t reconciling five versions of reality. I’ve staked my own work on this idea—building from Kochi since 2017, we’ve treated a unified core as the foundation, not an afterthought.
  • Price integrity by design: Quotes and charges should match by default. Wherever they can drift, the system must declare it up front and close the gap quickly. “Got it, booked” has to mean “ticketed and settled,” not “we’ll see.”
  • Booking finality: Confirmations should be auditable artifacts, not screenshots. Assistants (and customers) need the same durable identifiers the supplier trusts.
  • Observability and self‑healing: Interfaces drift. Reliable systems watch their own behavior, propose safe fixes, prove them, promote carefully, and route with intent. When this loop runs against a coherent model, assistants don’t paper over issues—they improve the platform.
  • Human‑in‑the‑loop surfaces: Escalation paths should be designed, not improvised. Give people the context and authority to make the few decisions only people should make—and let the system do the rest.

Do these well and you become the partner other agents prefer, not because your chatbot is charming but because your promises hold.

The next decade: compete on kept promises

Over the next decade of the travel economy, AI will make planning and booking feel effortless. That will make the product behind the scenes more visible, not less. Founders and operators will compete on the rate at which promises are kept:

  • The quoted price is the paid price.
  • The itinerary you hold is the itinerary that exists.
  • Changes get handled without making customers do admin.
  • A person can step in—fast—when judgement is needed.

Metrics will shift from “time on page” and “conversions” to “promise‑kept rates,” “time‑to‑finality,” and “time‑to‑care.” Your reputation will become an input other agents use to decide whether to route volume your way. In that world, the reliable win the right to grow.

My bias is founder‑led and systems‑heavy. I started ElkanIO Research Labs in 2017 with a deep focus on AI and analytics, and I build from Kochi because it keeps me honest about design. If you want a sense of my background and what I’m building, the About page has more.

Closing

I don’t think the travel winners will be the ones who make the cleverest chat. They’ll be the ones whose systems keep promises when the world is messy. That’s what customers—and agents—will bet on.

Get in touch →

Continue the conversation
Have a travel systems problem like this? I’d love to help.