The day travel support stopped being human, and why it came back
The industry spent a decade trying to automate the human out of customer support. The data now shows why the smartest players are putting it back.
At two in the morning, in a city where you don’t speak the language, with a cancelled flight and a meeting in six hours, you do not want a chatbot. You want a person. Someone who picks up, understands what’s wrong, and fixes it. That moment, the one where a trip falls apart in the middle of the night, is where the whole debate about automating customer support gets decided, and the industry spent years getting the answer wrong before the data pulled it back.
The story of travel support over the last decade is the story of a pendulum. It swung hard toward automation, swung to an extreme that didn’t work, and is now settling somewhere more sensible. The interesting part is that the numbers, not the sentiment, are what forced the correction.
The swing toward automation
The logic of automating support was never crazy. It was arithmetic. Human support is expensive, and the math is stark. Industry figures put a chatbot interaction at roughly fifty cents against about six dollars for a human agent, according to data compiled from Zendesk and other sources. When one channel costs a twelfth of the other, the pressure to automate isn’t a trend, it’s gravity.
And the capability caught up to the ambition. The market for AI customer service is projected to reach around fifteen billion dollars in 2026. AI agents now handle up to eighty percent of standard inquiries without escalation, by IBM’s estimate, and Salesforce expects half of all service cases to be resolved by AI by 2027, up from thirty percent in 2025. The returns are real too, with benchmarks pointing to roughly three and a half dollars back for every dollar invested, and the best implementations reaching much higher.
So the swing made sense. Cheaper, faster, available around the clock, which matters when Zendesk’s 2026 research finds that seventy-four percent of consumers now expect service to be available 24/7. For a huge range of support tasks, the machine genuinely is the better answer. If your problem is a password reset or an order status, a good AI agent solves it instantly, at any hour, and you never needed a human at all.
Where the extreme broke down
Then some companies did the thing companies do with a good idea. They took it too far. If automating some support saves money, the reasoning went, automating all of it saves more. And that’s where the data starts to bite back.
The Zendesk CX Trends 2026 report, built on surveys of more than eleven thousand people across twenty-two countries, put a sharp number on the limit. AI-handled tickets average a customer satisfaction score of 4.10 out of 5, against 4.30 for human agents. That gap sounds small until you see where it hides. Break it down by type of problem and the picture gets clear. Structured, simple requests score high on AI, with password resets around 4.41. But sentiment-heavy situations collapse. Complaint handling scores 3.34. Billing disputes, 3.61.
Read that carefully, because it’s the whole argument in a single dataset. AI is excellent at the transactional and poor at the emotional. When someone is calm and their problem is simple, the machine does great. When someone is upset, stranded, or scared, and the problem is tangled, the machine falls down, and satisfaction drops with it. Which is exactly the situation a broken business trip creates at two in the morning.
Why travel is the hardest case
Every industry is working through this, but travel is a particularly unforgiving place to get it wrong, for a reason that’s specific to the product.
Most support happens after the fact and at low stakes. Your order is late, you want a refund, you’re annoyed but safe. Travel support is different because it happens in real time, while the customer is physically out in the world, often far from home, sometimes genuinely vulnerable. A cancelled flight isn’t an inconvenience you handle from your couch. It’s a person stuck in an airport in a foreign country with nowhere to sleep. The stakes and the emotion are both higher than almost any other support context.
That’s why travel exposes the limits of pure automation faster than most industries. The complaint-handling score that drags down the averages elsewhere is, in travel, not an edge case. It’s Tuesday. The hard, emotional, high-stakes interaction is a routine part of the job, not a rare exception, which means a travel program that automated away its humans didn’t trim a cost. It removed the thing customers need most in the moment they need it most.
What “coming back” actually means
Here’s the part that matters, because “put the humans back” is too simple, and it’s not what the data actually recommends. The answer isn’t a return to the old all-human call center. It’s a specific division of labor, and the numbers point right at it.
That same body of research shows the gap between AI and human satisfaction narrows dramatically, from 0.20 points to just 0.05, when there’s a strong escalation path, a fast handoff from machine to person at the right moment. That’s the whole model in one statistic. Let the AI own what it’s good at, the high-volume, structured, unemotional requests that make up most of the tickets and cost a fraction to serve. Route the emotional, complex, high-stakes moments to a human, quickly, before frustration hardens. The failure was never using AI. It was using AI without a fast door to a person when the situation turned human.
Even the industry’s own framing has shifted to match. Zendesk’s 2026 message is no longer that AI is the differentiator. Their line is that AI has become table stakes, and the differentiator is how intelligently you combine it with human judgment. The vendors that spent years selling pure automation are now selling the blend, because the data made the pure version indefensible.
The lesson underneath
I’ve believed for a long time that technology and people together beat either one alone, and it’s satisfying to watch the numbers catch up to that. But the deeper point isn’t about travel, or even about support. It’s about what happens when an industry chases a cost curve past the point where it still serves the customer.
Automating support looked like a pure efficiency play, and for the transactional majority of cases, it was. The mistake was treating the emotional minority as if it were the same kind of problem, just a more expensive version, when it’s actually a different problem entirely. The password reset and the stranded traveler at 2am are not two points on one spectrum. One is a task. The other is a person needing another person. A company that can tell those two apart, and build for both, wins. One that automates them both the same way saves money right up until the moment a customer needed a human and found a script instead.
The pendulum is settling where it probably should have started. Machines for the transactions. People for the moments that are actually human. And the wisdom to know, instantly, which one is on the line.
About me
I am an entrepreneur with over 20 years of experience at the intersection of tourism and technology. I am co-founder and Chief Business Officer of VOLL, the largest mobile-first corporate travel and expense management platform in Latin America, and a recognized reference in the development of the corporate travel industry.
A Marketing specialist from Fundação Dom Cabral, I serve on the Tourism Council of FecomércioSP and on the Executive Council of the Latin American Association of Corporate Events and Travel Management (Alagev). A frequent traveler and close observer of human behavior in motion, I write and speak about innovation, digital transformation, entrepreneurial leadership, and the future of corporate travel.