From Travel Decisions to Travel Operations: FortripAI at Arabian Travel Market 2026
FortripAI is demoing at Arabian Travel Market 2026 in Dubai — not another AI that writes itineraries, but one that optimizes the trip you already planned and turns finished itineraries into sourcing-ready requirements for the operators who have to actually run them.
On September 14, I'll be in Dubai demonstrating FortripAI at the AI Innovation session at Arabian Travel Market 2026 — running September 14–17 at Dubai World Trade Centre, with more than 55,000 industry leaders from 166 countries expected across technology, sustainability, and the future of travel.
I'm Boyuan Dong, founder and CEO of FortripAI. This demo matters to us less because it's a new product showcase and more because it's a chance to make an argument we think the industry needs to hear:
Plan a trip with Fortrip
Use Fortrip to draft and validate itineraries with travel-specific reasoning — not generic chat answers.
A scenic routing request added 228 miles and pushed four driving days to roughly ten hours each. An itinerary with a textbook route landed on the central coast at the peak of flood and typhoon season. A park fee quietly doubled between one half of the year and the other. Rome2Rio answers how to get there — here's the layer that decides whether you should go that way.
Rome2Rio can confirm in about a second that four cities connect. It will not notice that one of your flights departs before the flight bringing you there has landed. Route search engines and AI optimizers are built for different questions, and most multi-city trips need both answered. Here's what each one solves, where each genuinely falls short, and how to use them in sequence.
the next valuable layer of AI in travel isn't itinerary generation. It's decision-making and execution.
That belief shapes both halves of what we build — the optimization tools travelers use, and the operations layer we're now building for travel businesses.
Why Another AI Itinerary Generator Isn't the Answer
Ask an AI for seven days in Japan or a multi-country Balkans trip and it'll produce something convincing in seconds. That part of the problem is basically solved. What isn't solved is what happens next — a traveler rarely trusts the first itinerary they see, and starts asking a longer list of harder questions: should I spend three nights here or four, would a different city order lower the total cost, is this train connection actually still running, is this itinerary technically possible but needlessly expensive or exhausting?
And once an itinerary looks right, a different question shows up: can I actually execute it? That question — not "can AI write a plan" but "can AI make the plan better and then help make it real" — is why we think of FortripAI as travel decision infrastructure rather than a planning tool.
For Travelers: Better Decisions, Not More Suggestions
The consumer side of FortripAI starts with a conversation, but the conversation isn't the product — the value is in what happens after a first itinerary already exists.
Multi-city trip optimization. Take Tokyo → Kyoto → Osaka → Hiroshima. There are several valid sequences, and they aren't equivalent: one might land a city on a weekend or holiday date where hotel prices spike, another might create longer transfers or backtracking that a traveler wouldn't choose on purpose. FortripAI's Transport Optimization Agent evaluates alternatives — a different order, a shifted date, a lower total hotel cost for the same destinations — instead of accepting the first reasonable-looking sequence. We've written up several real examples of this, including a case where reordering the same three hotels changed the total by $650 and one where shifting a departure date by two days saved $400 — the pattern shows up often enough that it's worth checking on nearly any multi-city trip.
The goal isn't always the cheapest possible trip. For a luxury traveler it might mean keeping the same hotels but finding better dates; for someone budget-conscious it might mean cutting total accommodation cost; for someone else, saving six hours of unnecessary transit matters more than saving $200. Optimization only helps when it's aimed at what the traveler actually cares about.
Hotel structuring, not just hotel search. Hotel search has traditionally answered one question — where should I stay. Multi-city and longer trips raise a different one: how should the stay be structured? Five nights split across two hotels can be booked as 2+3, 3+2, or any other combination, at slightly different dates, in either order — and because hotel prices move by night, these combinations can land on very different totals for the exact same hotels and the exact same five nights. We found a $327 spread doing exactly this for a real Costa Rica trip. FortripAI treats that as a decision problem to compare, not a single recommendation to accept.
Validation, regardless of who wrote the plan. A well-written itinerary can still contain an attraction that's closed that day, a transfer that doesn't actually work, or travel time that reads fine on paper and falls apart in practice. The FortripAI Validator checks an itinerary for exactly this — and it doesn't have to come from FortripAI. Bring one from ChatGPT, a travel advisor, or your own spreadsheet, and the question becomes "is this itinerary actually good," not "which AI wrote it."
One thing we learned early is that most travelers don't describe planning itself as their biggest pain point — plenty of people genuinely enjoy discovering restaurants, comparing hotels, and imagining the trip, and automating that away removes part of what makes travel planning fun in the first place. What people don't enjoy is uncertainty: finding out after booking that the city order cost them money, that a connection doesn't actually work, or that an AI confidently invented a detail that wasn't true.
That distinction became a working principle: don't automate the joy of planning — automate the uncertainty, comparison, and verification underneath it. That same principle is now shaping the B2B side of the product.
For Travel Companies: The Problem Starts After the Itinerary Is Finished
Travel businesses face close to the opposite problem consumers do. For a tour operator, agency, or custom-trip provider, an itinerary isn't the end of the workflow — it's usually the beginning. A travel designer writes something like "Day 4: depart Shanghai, visit Wuzhen, continue to Hangzhou," and an operations team has to translate that single line into vehicle requirements, driver hours, hotel check-in coordination, guide bookings, supplier availability, and cancellation terms. Today, that translation is mostly manual.
This is the gap we're calling itinerary-to-operations: FortripAI reads an itinerary — built in FortripAI or brought in from elsewhere — and asks what actually needs to happen for the trip to be fulfilled, turning it into a structured set of Fulfillment Requirements.
A fulfillment requirement is not the same thing as a supplier product. A product is something a supplier already sells; a fulfillment requirement describes something the traveler needs to happen, whether or not a matching product exists yet. If a traveler wants a car for the afternoon, dinner at a specific restaurant, then a concert with the driver waiting afterward to take them back to the hotel, there's no supplier catalog item called "restaurant plus concert plus late-night transfer" — and there doesn't need to be. FortripAI can describe that as a single evening private-vehicle requirement (say, 17:00–00:30, hotel → restaurant → venue → hotel, driver waiting required) and hand it to an existing transportation supplier as an RFQ. Standard inventory covers some requirements; others need a quote, a manual reservation, or turn out to already be bundled into another service.
Continuous services should aggregate, not multiply. Three consecutive nights at the same hotel is one accommodation requirement, not three. Four days needing the same vehicle is one transportation requirement, not four separate day-by-day bookings. Recognizing that boundary — instead of generating one line item per day regardless of what's actually being purchased — is most of the difference between a usable procurement list and a 30-item mess nobody wants to work through.
AI should flag what's missing, not just extract what's written. An itinerary that says "morning mangrove kayaking, then public bus to San José" doesn't say what happens to the travelers' luggage while they're kayaking, or how they get from the water to the bus terminal. Those questions rarely make it into a traveler-facing document, but an operations team has to answer them eventually — which means the more useful question isn't "what does this document say" but "what needs to be resolved before this trip can actually run."
Missing information should become a short decision queue, not a long checklist. Instead of showing an operations team 30 incomplete fields, we want the system to identify which few decisions unlock the most requirements — confirming travel dates might resolve 17 open items at once, confirming room configuration might resolve 5. The employee doesn't see "26 missing fields." They see three questions that, once answered, make most of the trip actionable.
The sales-to-operations handoff. A salesperson creates the itinerary, the customer approves it, and someone else has to operate it — today usually by re-reading the whole plan and manually translating it into hotel, vehicle, and guide requests. We want that handoff to look like: itinerary approved → FortripAI generates fulfillment requirements → operations sees a short list of sourcing requirements and open decisions → RFQs go out to suppliers. Fewer re-reads, fewer missed details.
We're not trying to replace the supplier network. A lot of the best travel inventory — a trusted driver, a local guide, a restaurant contact, a small community operator — exists through relationships that were never going to fit a standardized catalog, and we don't think forcing every request into predefined products is the right answer. FortripAI's role is translating traveler intent into supplier-ready requirements; the agency or DMC keeps its own supplier relationships and owns the decision of who to book.
The Same Problem, Viewed From Two Sides
The consumer optimization tools and the B2B fulfillment layer might look like two different products. We think of them as the same underlying problem from opposite directions. For travelers, there are too many possible versions of a trip and the question is which decision to make. For travel businesses, there's an unstructured itinerary and the question is what needs to happen next. Both are really about turning messy travel information into something structured enough to act on — for travelers that looks like compare → optimize → validate → decide, and for businesses it looks like understand → structure → resolve → source → operate. The common layer underneath both is reasoning, not generation.
Frequently Asked Questions
What is an itinerary-to-operations layer?
It's software that reads a finished travel itinerary and converts it into structured, sourcing-ready fulfillment requirements — the vehicles, accommodations, guides, and tickets an operations team needs to actually book — instead of leaving that translation to be done manually.
What's the difference between a fulfillment requirement and a travel product?
A product is something a supplier already sells as a package. A fulfillment requirement describes what the traveler actually needs to happen, whether or not a matching product exists yet — some are met by standard inventory, others need a custom quote or manual booking.
Can AI turn a travel itinerary into supplier RFQs?
Yes — by identifying the accommodation, transportation, guide, and activity needs implied by an itinerary, grouping continuous services (like several consecutive nights or days needing the same vehicle) into single requirements, and flagging gaps the itinerary doesn't explicitly answer before sending requests to suppliers.
Does an AI itinerary validator only work on itineraries it created itself?
No — a validator can check an itinerary regardless of where it came from, including ones built with ChatGPT, a travel advisor, or a spreadsheet, checking for issues like closed attractions, unrealistic connections, or underestimated travel time.
Can changing the order of cities in a multi-city trip actually lower the total cost?
Often, yes — because hotel and flight prices vary by date, reordering the same destinations can change which dates each stay gets assigned, sometimes by hundreds of dollars for the exact same hotels and nights.
See FortripAI at Arabian Travel Market 2026
I'll be demonstrating FortripAI on September 14 in Dubai during the AI Innovation session at Arabian Travel Market 2026, which brings together travel companies, destinations, DMCs, hospitality companies, and technology buyers from across the industry — a fitting place to show where we think AI travel technology is heading, on both the traveler-facing and operations side.
If you're attending ATM 2026 and work in travel technology, tour operations, custom travel, DMC services, or travel procurement, I'd like to meet and compare notes.
The question we're interested in isn't whether AI can plan a trip anymore — it clearly can. It's whether AI can help make that trip better, and then help make it actually happen.
