A NEW DYNAMIC AT GBTA, DRIVEN BY AI

In a TTH talk at the GBTA Business Travel Forum São Paulo in June, which gave rise to the "From Hype to Hands-On" playbook, the pace of change was framed in three boxes: what was once a cycle of years became a cycle of months and now runs in weeks, sometimes days. Anthropic, the company behind Claude, shipped 74 releases in 52 days, roughly 1.4 per business day. A feature can become obsolete while it is still profitable, because the profit reflects ground already covered. The sooner you see the cycle, the sooner you start moving. Chicago confirmed the thesis two months later: a market that had been discussing AI as a future and a doubt can now see the results and press real buttons.

The list of launches was long: FCM Travel presented conversational booking inside FCM Booking, filtered by policy, profile and traveler history, running first inside Teams and Slack with planned expansion, via MCP, to ChatGPT, Claude and Gemini. Accelya launched SPRK Assist, covering 37 NDC servicing use cases in which the agent describes the desired outcome in natural language and the system locates the booking, validates contractual conditions and calculates fares. TripGain announced its MCP server. Direct Travel rebuilt Avenir as a multi-agent architecture on top of Spotnana. Cerebri presented expense agents with Moderna as its first named client. Conversational search, NDC, agents: exactly the package any good observer had predicted, now with an availability date attached.

In practically every launch, the human, largely the travel manager, remains the final authority by product design, not by technical limitation. In SPRK Assist, the agent prepares and the human confirms. In FCM Booking, policy filters first and the consultant stays in the loop. "We didn't build FCM's technology ecosystem to replace our consultants' judgement. We built it to take the routine work away", said John Morhous, Chief Experience Officer at FCM. That is neither accident nor modesty: fully replacing human teams with AI has already proven unsuccessful wherever it was attempted, both in quality of delivery and in cost, promising efficiency and delivering rework. The market learned, and the product design of 2026 carries the scar.

It is the same conclusion the most provocatively titled debate of the convention reached by another route. In the session "NDC Is Dead: Long Live King AI", the screen opened with a challenge: "NDC has been 'the future of distribution' for over a decade. AI got there in 18 months." Along the way, the session weighed the NDC promise against the delivery: rich content and branded fares stalled on GDS economics, personalization at the point of search remains uneven and, as the screen itself put it, every airline implemented NDC its own way, and the standard never truly standardized. Forty-five minutes later, the closing slide dismantled the provocation: coexistence, not a cage match. The verdict, written on screen: "AI on top of NDC, not instead of it". AI as the interface and decision layer; NDC and GDS remain the plumbing.

In more than one education session, travel managers described the same scene: travelers using AI to build itineraries, compare prices and, in growing numbers, to buy. The data released that week showed the scale of it.

Amadeus called it shadow AI: 62% of American corporate travelers have already used their own AI to plan or book outside approved channels, 20% do so regularly and only 9% say they never would.

SAP Concur, surveying 3,300 travelers across 21 markets, found 72% willing to use unsanctioned tools, with 85% of CFOs concerned and 73% considering it likely that AI is already being used to falsify expenses.

Travelport mapped the funnel: 32% research with AI, 8% complete the booking through it. Trust to assemble options, 76%; to spend on their behalf, 44%. And 80% require seeing the tool's reasoning before letting it act.

The detail linking the data to the product showed up on media day: in the FCM conversational booking release, every recommendation comes with the reasoning attached, showing why it fits policy, why it matches the traveler's usual route and loyalty program. "Conversational booking only works if travelers trust what they're being shown, and that trust comes from context, not just convenience", said Jessica Dunderdale, Global Head of Product at FCM. It is the direct answer to the 80% who require seeing the reasoning before letting the tool act.

AI is the new Google, remember (and for the older among us, it is also the Office suite on a résumé). The comparison deserves to be taken seriously in everything it implies, because the change is not one of channel, it is one of behavior. It changes how every system delivers value. It changes how travelers look for the first information about a trip and a destination. It changes how they compare fares, hotels, flights and neighborhoods. It even changes the calculation between renting a car and calling an Uber. It therefore changes the dynamics of the entire market. And it should change how we generate value as an industry and as participants in it.

The market has already lived through a dress rehearsal of this, and the name is worth recalling: open booking. Last decade, the term described a booking made outside the preferred channel, and programs responded with an arsenal of capture: itinerary integrations, email parsing, reimbursement policy conditions. Shadow AI is open booking 2.0, with one aggravating factor that changes the category of the problem: AI is not merely another channel outside the preferred one, it is a channel that reasons outside policy.

Hence the question every travel manager asked in the hallways: how do you catch up with that adoption? Chasing the traveler is running a race already lost, with the added cost of turning the travel program into a police force. There may be another way: to embrace AI not as a single miraculous tool but as part of a set, an ecosystem of tools that supports and advances the corporate travel ecosystem as a whole. The traveler comparing neighborhoods in ChatGPT is not an offender, and may well be an unpaid beta tester for the travel program of the future, or something close to it.

This avalanche of new capability forces a rethink of the value a travel manager can generate for the organization. The CFO wants to know why the program costs what it costs. The C-level wants to know what the function delivers beyond a cheaper ticket. And the answer that worked ten years ago, "savings on travel", has shrunk: savings became a commodity that every tool promises. The proof of value moved from the trip itself to the orchestration of the whole ecosystem: travelers, tools, suppliers, stakeholders (CFOs and C-levels at the head of the table), budget and an ROI tied to the business, not to the fare or the savings.

The GBTA survey quantified the chaos every global manager already knows: on average, 5.3 OBT instances and 5.3 expense instances per multinational program. Five environments, five sources of data that do not talk to each other. No AI orchestrates something fragmented that way, and no conductor leads five orchestras playing different scores. Before the agent, the inventory; before the automation, the tidying up. It is the corporate version of the garbage in, garbage out of those 30 systems that Vinicius Gonçalves, of TourHouse, brought back from a session.

How should that new role be pursued through technology and AI? The answer I argued for in São Paulo, and which Chicago reinforced, has three layers. First, a real increase in analytical capacity: the travel manager who does not read data outsources the decisions to whoever does. Second, a change in posture: turning the travel program into a compass for value creation rather than a set of hard rules to be obeyed; the yardstick stops being "how many exceptions did I block" and becomes "how much value did the program create for the business". Third, AI attached to the routine, not as a special project with its own committee and budget.

A provocation, then, and it is individual rather than institutional: look at your skills, at your set of tools and at your role in the company, and ask honestly how to absorb this avalanche. Yes, avalanche. It is not an exaggeration for the volume of change this paper documents on every page, and an avalanche is not managed the way a breeze is managed: it demands revisiting the role, not merely defending it. The professional who takes that inventory now gets to choose what changes; whoever delays will receive the change fully formed, chosen by somebody else.

The session "Brains + Bots" gave the same movement a framework name: AI Readiness, Intelligence & Governance and Human Boundaries, the three pillars of the reimagined travel buyer. The third pillar is what holds up the other two: identifying where human judgement is indispensable, and the boundary no panelist at the convention contested is that crisis management is not delegated to an agent. The model standing in every product and every framework is the same, and I summarized it in this month's Travel Leaders Hub column: the agent suggests, the person decides.

Historically there was a wide distance between the American market and the Brazilian one. Technology, processes, tools and best practices were born and developed there; we consumed them as novelties. It was understood that this was the best on offer, and we visited GBTA as the mecca of the new. The 2026 edition buried that script. The trip became less about learning and more about connecting, debating and understanding a way of working. Carolina Gaete, of Voetur, who has attended the convention for 18 years, sums up the arc: there was a time when talking about OBTs felt remote; today "we are very close, and Brazil has something to teach". Aline Bueno, CEO of Argo Solutions, with more than 15 GBTAs behind her, says she spent 13 of them wondering when all of it would reach Brazil, and that this year she watched the distance "shrink considerably". Manuela Bernardes, of Flytour, went further: "we are not behind in anything".

There are even areas where we are further along, out of necessity. Payments is the clearest example, and it is not only about PIX: virtual card numbers embedded in the transaction, automatic reconciliation, anti-fraud layers that the American market is only beginning to discuss, while Conferma launched at the convention itself the first global benchmark for virtual card performance.

Even Brex, exhibiting at GBTA, tells that story: you cannot say it was created in Brazil, but Brazilian creativity is in the product's foundation and nobody can take that away. Pagar.me was born to solve the complex problems of our market; we exported a piece of that complexity in the form of competence in its founders. No inferiority complex here: this is genuine pride.

Brazil, because of its complexities, did not only get to learn over all those years: it now had to be able to teach. In payments, as the executives we interviewed pointed out, the pain in our market was severe and the urgency just as strong. We were more anchored in a scarcity of resources and technology than the Americans were, and precisely for that reason we held them up as the great reference. They still are, in specific verticals and subjects. But AI changed a structural variable in the game: for the first time, frontier technology is available on the same terms and at the same price to every market. And when the tool is the same, what differentiates is the repertoire of whoever uses it. Perhaps markets accustomed to scarcity will fare better. Creativity sharpens in settings where there is less resource than the task requires. That Brazilian knack for making do, always a diagnosis, has turned into a competitive advantage.

Pride does not excuse us from honesty, and the counterpoint came from the interviewees themselves. On strategic maturity the tables turn: in the United States, whoever owns the category holds an executive seat; in Brazil, travel is still a cost line at most companies. On AI governance, Carolina Gaete saw policies and finished products there that are still missing here. And there is a gap that is not technical but cultural: Brazilian executives still hear "I don't know that market" at the booths. Brazil's deficit at GBTA is not one of repertoire, it is one of recognition, and recognition is built by showing up. The official data helps build it: Brazil growing 13.8%, the fastest pace among the world's 15 largest markets, in a year when global spend reaches US$1.71 trillion.