AI Can Book the Trip: What Clients Still Need From a Travel Advisor

In September 2026, the Indian travel platform Travelxp launched an agentic AI trip concierge called Marco. According to PhocusWire's reporting on the launch, Marco does not stop at recommendations. It searches live fares and room rates, builds an itinerary, completes the booking and executes payment within spending limits the traveller sets in advance, and then stays with the trip afterwards, adjusting the itinerary when a flight is delayed or a price changes, without the traveller stepping in. Travelers reach it through WhatsApp, Instagram or the Travelxp website.

Travelxp itself described the transacting piece as "the piece that has been missing from most AI travel tools to date, the ability for an agent to not just recommend but to transact." Trade coverage of the launch also noted that Marco can coordinate group trips, opening a group message and booking against preauthorised amounts from each traveller.

Two qualifications matter before anyone draws conclusions. Marco is currently available in India, built around Indian payment authorisation rules, and it is not a product landing on North American advisors' desks this quarter. And the company's own framing of the market gap is telling: Travelxp cited 2026 research finding that while 75% of travellers using AI rely on it for recommendations, only 13% use it to complete a booking. Read Marco as an industry signal about where consumer tools are heading, not as a competitor that has already arrived.

The signal is still worth taking seriously. The direction of travel is that search, assembly and transaction are becoming cheap and automated. If your business is built on doing those three things competently, your position is getting weaker every year, and that has nothing to do with whether you like the technology.

An itinerary is an artefact. It is a document listing flights, hotels, transfers and activities in a defensible order, and software is now genuinely good at producing one. Judgment is a different product. Judgment is deciding that a client should not take the itinerary she has asked for, and being able to explain why in a way she accepts.

The distinction shows up in the choices that have no correct answer on paper. Two properties price identically and both look right; one suits a couple who want to be left alone and one suits a family that wants staff who engage with their children. A routing saves four hours and costs the only sensible buffer before a once-in-a-lifetime sailing. A destination is beautiful in the month the client is free and unbearable in the way brochures never mention. Software can retrieve all of this. It does not carry the consequence of choosing wrongly.

That is why the useful question for an advisor is not whether AI can build an itinerary. It obviously can. The question is how much of your fee is currently justified by assembly work that is being automated, and how much is justified by decisions you are accountable for.

Supplier relationships remain valuable because they are relationships with people, and people respond to history. An automated system can request a room type. It cannot call in a favour from a general manager who remembers the client you sent last year, or get a firm answer at ten at night from a destination management company partner who knows your name. Access built on reciprocity is not something a booking engine can query.

Real destination knowledge holds value because it is specific and current rather than descriptive. Which wing of the hotel is under renovation this season, which guide is worth the premium, which market closes on the day your client planned to visit. That knowledge decays and has to be maintained, which is exactly why it is scarce.

Complex trip design holds value because complexity multiplies failure points. A multi-country trip with private aviation, a chartered vessel, elderly travellers and a fixed event date is not a harder version of a simple booking. It is a different task, where sequencing, buffers and fallback plans matter more than price comparison.

And the client's unstated priorities hold value because clients rarely lead with the real brief. She says she wants a villa in Tuscany. What she wants is one week where her adult children actually sit down together, which changes almost every decision you make. Drawing that out is interview work, and it happens in conversation with someone who has heard a thousand versions of it before.

Your clients are arriving with AI-assisted plans, and the plans are usually plausible. Plausible is the problem. A confident draft can quietly contain a property that closed for the season, a connection that works on paper and not in the terminal, a visa or entry assumption that is out of date, an activity unsuitable for a traveller's mobility, or a hotel whose online reputation reflects a standard it no longer holds.

Treat a client's AI draft the way you would treat a quote from an unfamiliar supplier: useful input, unverified. Confirm availability and current condition at the source. Check that the sequencing survives contact with real transfer times. Flag the choices that conflict with something the client told you she cares about. This is not adversarial work, and it should never be framed as correcting the client. It is the part she cannot do herself, because verification requires knowing which claims are worth doubting.

It is also worth naming the value of a second judgment on a large purchase. An advisor who can say "this is sound, and here are the two things I would change" has given a clear, paid service, even on a trip the client largely assembled alone.

Supplier service is automating quickly, and mostly for good reasons. Salesforce announced at Dreamforce 2026 that Air India is expanding its use of the Agentforce platform across customer service, beyond its first refund use case, into automated email resolution, passenger name changes and an internal knowledge assistant for service teams. Salesforce reported that refund turnaround fell from roughly 14 days to about four hours, and that eligible name-change requests now take around 30 minutes instead of about three days.

Those are real improvements, and advisors should welcome them. But read the detail in Salesforce's own account: automated responses are generated only when the system's confidence exceeds 95%, with human oversight retained where required. Automation is built to handle the cases it recognises. The cases it does not recognise are the ones your clients call you about.

That is where advocacy lives. A disrupted itinerary with a non-standard fare, a medical situation abroad, a supplier error that falls between two policies, a goodwill request no rule entitles anyone to. Resolving those requires knowing which person to reach, what to ask for, what leverage you actually hold and when to escalate rather than resubmit. Salesforce's broader Dreamforce message about grounding AI agents in business context makes the same point from the vendor's side: an agent performs within the context and permissions it has been given. Your clients' worst travel days are, almost by definition, outside that context.

People do not buy professional help in proportion to task difficulty. They buy it in proportion to consequence. Nobody needs an advisor to book a weekend hotel. They need one when the trip carries substantial money, multiple travellers with competing needs, accessibility or medical requirements, or a milestone that cannot be rescheduled: a fiftieth anniversary, a graduation trip, a first family journey after a loss.

In those situations the client is not buying an itinerary. She is buying a named person who will answer, who has assumed responsibility for the outcome, and who can be held to it. An automated agent operating within preset limits offers convenience. It does not offer accountability, and it cannot be embarrassed by a bad result.

This is the honest commercial position for advisors. The low-consequence end of the market has been leaving for years and will keep leaving. The high-consequence end is where professional planning was always strongest, and it is the part technology is least able to absorb.

Automate what the client never sees or never values: intake forms, document collection, passport and expiry tracking, calendar reminders, deposit and final-payment notices, first-draft research summaries, trip document assembly, follow-up sequences, review requests, anniversary and re-book reminders, and internal note-taking after calls. Every hour reclaimed here is an hour available for work that justifies your fee.

Keep human review on anything where being wrong costs the client something: the final recommendation and the reason behind it, property and supplier selection, pricing and fee conversations, anything involving medical, accessibility or safety considerations, complex sequencing and buffers, disruption handling, and every message sent in your voice to a client during a live problem.

A workable rule: AI may draft, retrieve and remind. It may not decide, and it may not be the last set of eyes on anything a client will act on. Applied consistently, that rule lets you use the tools aggressively without letting them quietly replace the part of the work you are actually selling.

When a client arrives with an AI-built plan, she has not insulted you. She has told you what she wants and saved you the discovery shortcut. Say so plainly, then add what she could not add. "This is a good starting point. Here is what I would keep, here are the two things that would not work in practice, and here is what I can get you that is not on a booking site." That conversation positions you as the professional without a word of defensiveness.

Avoid the two failure modes. The first is dismissiveness, which tells a capable client she is naive and makes the rest of the conversation a contest. The second is anxious over-explaining about how AI cannot replace human advisors, which signals that you are worried and invites her to wonder whether she should be too.

Talk instead in terms of what you take on. You verify. You hold supplier relationships that produce outcomes she cannot request. You design for what happens when something fails. You are accountable when a trip with real money and real meaning attached does not go to plan. That is a description of value, not a defence of a profession.

First, document your planning process. Write down every step from first inquiry to post-trip follow-up, including the checks nobody sees. Most advisors do far more than they can describe, and a process they cannot describe is a process they cannot charge for. The written version becomes your consultation script, your fee justification, the basis for your automation decisions, and the standard you hold when you are busy.

Second, tighten your client data and CRM practice. The advantage you hold over any automated tool is accumulated knowledge of the person: anniversaries, dietary and medical needs, what they disliked last time, who in the family actually decides, which properties they have already seen. That advantage exists only if it is recorded in one place rather than scattered across your memory and your inbox. One client record, updated after every conversation.

Third, revise your consultation and website language around judgment and advocacy. Read your current copy and count how much of it describes things a capable AI tool can now do: finding options, comparing hotels, building itineraries, saving the client time. Replace that with what you are accountable for, the depth behind your recommendations, and what happens when something goes wrong at eleven at night in a country where the supplier's automated system has no answer. That is the language that still holds in 2027.

None of this depends on predicting how far consumer AI agents go. It holds whether Marco expands beyond India or stalls, because the work being described is the work that was always the most valuable part of the job. The change is simply that it is now the part that is clearly worth paying for.

If your client can get the information herself, search live rates herself, and increasingly complete the transaction herself, what can she only get by working with you? If the answer is convenience, that answer is depreciating. If the answer is judgment, relationships, verification, advocacy and accountability, you have a business that gets more valuable as the tools get better, because capable tools raise what clients expect from the human in the room.

That is the case for building the business deliberately rather than defensively. If you want practical training on the positioning, client service, supplier relationships and business systems that support this kind of practice, explore Unrivaled Travel Academy. The work is unglamorous and it compounds, which is the only kind of advantage worth having.