When AI in travel first hit the mainstream, it mostly looked like a chatbot. A search box with a friendlier face. A polite assistant that could answer questions but couldn’t really do anything.
That’s changing fast and what’s coming next is what the industry has started calling agentic AI.
This week, our Director of AI Products and Personalization, Brianna MacNeil, joined Dan Christian on the Travel Trends Podcast for Part 2 of his Agentic AI series. She was joined by Ben Manzi, Co-founder and COO of Maya, and Michelle Denogean, CMO of Mindtrip, three companies looking at this shift from different angles, but agreeing on the core point: AI in travel is moving from conversation to action.
What “agentic AI” actually means
The shorthand “agentic AI” can sound like marketing-speak, but the working definition the panel landed on is concrete. An AI agent is software that can take real steps on a traveller’s behalf: pull live pricing, generate quotes, draft emails, build full itineraries, and increasingly help make the decisions that travellers used to grind through alone.
The old internet of filters, dropdowns, and twenty open tabs is starting to feel dated for a reason. People want help with the hard part of travel, the paradox of choice, not just help with the search.
Where the value shows up first
One of the more honest threads in the conversation is about where agentic AI delivers value first. It’s not the traveller-facing demo that gets the keynote slot.
It’s the unglamorous middle of the operation: repetitive, low-judgment work clogging operator inboxes, internal workflows that lose days to manual handoffs, and the gap between what large platforms can afford to build and what smaller brands can realistically ship. Agents close that gap. They let leaner teams operate with capacity that used to belong only to the giants.
Trust is the unlock
The other recurring theme: none of this works without trust, and trust is harder than it looks.
Three things came up repeatedly across the conversation:
- Grounding agents in real data. A trip planner that hallucinates hotels that don’t exist is worse than no trip planner at all.
- Cleaning up outdated content. Agents are only as good as the inventory and information they can pull from. Stale data poisons the experience.
- Choosing the right models. Cost and latency aren’t side concerns — they decide whether an agent is usable at scale or a research project.
None of this is glamorous. All of it is the difference between a demo and a product.
A new “listening layer” for marketers
For destinations, hotels, and operators thinking about discovery, the panel surfaced a shift worth paying attention to. As planning moves into conversational interfaces, a new listening layer is emerging — one that surfaces what travellers are actually asking, not what marketers think they’re searching for.
This is where AI SEO and generative engine optimization come in. Ranking in the new layer depends on whether your content answers real traveller questions, in real language. Evergreen content written for the old search index isn’t enough.
What stays human
The conversation closed somewhere worth ending: agentic AI changes a lot, but it doesn’t change everything.
Taste still matters. So does community validation. The friend who’s actually been to the place, the operator with twenty years of on-the-ground knowledge, the moment in a destination that no model can simulate. The long-term winners won’t be the brands that bolt a chatbot onto a legacy stack. They’ll be the ones that understand a traveller across contexts, and know which parts of the experience to automate and which to leave alone.
That’s the conversation worth having, and one we’re glad to be part of.
