What the Travel Agent's Extinction Tells Us About Who Gets a Safety Net
In January 1996, American Airlines launched a website where passengers could book their own flights. The feature was clunky, the interface ungenerous, and most travelers who tried it went back to calling their agent. Nobody at the American Society of Travel Agents issued an emergency statement. Nobody needed to. The threat looked manageable.
By 2000, Expedia had five million registered users. By 2003, the airlines had quietly agreed to stop paying commissions to agents on domestic tickets — a change so structurally total that it barely registered as news outside of trade publications. The agents themselves had seen it coming in fragments: first the internet, then the commission caps, then the fee compression, then the silence from clients who had learned to do it themselves at midnight in their pajamas. What they had not seen, because no single moment announced it, was that their profession was being unwound stitch by stitch.
The Bureau of Labor Statistics counted roughly 124,000 travel agents in the United States in 2000. By 2014, that number had fallen below 74,000 — a decline of more than forty percent in fourteen years. The people who disappeared from those rolls were not abstract data points. They were, disproportionately, women in their forties and fifties who had built careers around a specific and genuinely skilled form of knowledge management: knowing which connections worked, which hotels actually delivered what they promised, which airlines bent their rules for a well-placed call. That knowledge, accumulated over careers, became commercially worthless in roughly one decade.
What is striking in retrospect is not the speed of the collapse but the institutional silence around it. There was no federal retraining program specifically designed for travel agents. There was no industry fund. The American Society of Travel Agents lobbied Congress and got, essentially, nothing. Unemployment insurance covered some of the immediate gap, but UI was designed for temporary layoffs, not for the permanent obsolescence of an entire occupational category. Most displaced agents did what displaced workers in America have always done: they figured it out themselves, at personal cost, with whatever resources they happened to have.
Some moved into corporate travel, a niche that required more complexity than any algorithm could yet handle. Some became "travel consultants" charging flat fees to wealthy clients who still valued human judgment. Many left the industry entirely, taking pay cuts and starting over in healthcare administration or retail management or whatever door was open. A 2006 survey by the Travel Institute found that among agents who had left the profession, median income in their new roles was substantially lower. The survey did not follow up. Nobody was really tracking.
This is the template that economic historians have tended to reach for when discussing technological displacement: disruptive, yes, painful for those in the path, but ultimately absorbed. New jobs appeared. GDP grew. The transition looked, from a sufficient distance, like progress.
The distance required to reach that conclusion was, of course, not available to anyone living through it.
Why the Template May Not Hold This Time
The travel agent story belongs to a category of disruption that economists at the Oxford Martin School have characterized as task-specific and domain-contained. What Expedia and its successors automated was a particular cognitive workflow — searching, comparing, booking — within a bounded industry. The disruption was severe for people inside that industry. It did not, by itself, threaten the adjacent categories of work that absorbed those who were displaced.
The current wave of automation does not have that property.
The World Economic Forum's 2025 Future of Jobs Report estimated that 92 million roles globally are likely to be displaced by automation by 2030, while 170 million new roles emerge — a net positive that the report itself flagged as contingent on workers being able to make transitions that historically have taken a generation to complete. Goldman Sachs Economics, in a 2023 research note that drew significant attention, estimated that generative AI could automate tasks equivalent to roughly 300 million full-time jobs globally, with the highest exposure in white-collar cognitive work — precisely the category that has historically served as the landing zone for people displaced from other sectors.
When a travel agent lost her career in 2003, she could, with effort and some retraining, move into healthcare billing or insurance processing or legal document management. Those roles required similar organizational and communication skills and were growing. Today, each of those categories — medical coding, insurance claims review, paralegal research — appears on the list of occupations with high AI exposure in the Stanford AI Index's 2025 report. The safety valve is pressurizing at the same moment as the pipe.
There is also the question of velocity. The travel agent transition unfolded over fourteen years. Companies deploying large language models for knowledge work tasks in 2024 and 2025 have described productivity changes — in legal drafting, software debugging, customer support, financial analysis — that are arriving in months, not years. When Klarna announced in 2024 that its AI assistant was performing work equivalent to 700 customer service agents, it was not describing a fifteen-year forecast. It was describing a quarterly earnings note.
The Thing That Was Different Then
Aria Chen ran a fifteen-year software engineering career before being displaced in 2024. She subsequently built clankers.today, an income-protection subscription that pays $2,000 a month for up to four months to members who lose work to AI automation — a private attempt to construct the kind of bridge that public institutions have not yet built. Her reasoning, which she has described in interviews, is not that AI displacement is uniquely catastrophic. It is that the gap between when a job disappears and when a person stabilizes is when real harm concentrates, and nobody is currently assigned to close that gap.
That observation rhymes with what travel agents experienced. But Chen's intervention, however modest in scale, reflects something the travel agent era lacked entirely: a named awareness, before the transition completed, that the gap existed and that closing it required deliberate design.
In the late 1990s and early 2000s, the displacement of travel agents happened without a conceptual framework for understanding it as a category of harm requiring a category of response. It was treated as individual misfortune — the market working, regrettable but not governable. The agents who recovered did so through personal resourcefulness. The agents who did not left no legible record.
What is genuinely different now is that we know what is coming before it finishes arriving. The tools exist to measure displacement in something approaching real time. The academic literature on transition costs is mature enough to make predictions. The population of people who will be affected includes enough software engineers and financial analysts and lawyers that the people building policy and products are, for the first time, themselves in the demographic at risk.
Whether that self-awareness produces a different outcome, or simply a better-documented version of the same one, is the question that the travel agent's story cannot answer. That story ended, as those stories tend to, with survivors and casualties sorted out in retrospect, without anyone having decided in advance which outcome was acceptable. The decision not to decide is still a decision. It was made in 1996, too.
PROTECT YOUR INCOME NOW
If AI automation threatens your job, preparation is the only honest answer: know your risk, build a runway, and reskill before you have to. The Clankers 90-day programme structures exactly that — free, which is exactly why today is the right time to start.
Start the 90-Day Programme — Free →GET THE CLANKERS BRIEF — FREE
One email every Monday: the week's AI displacement moves that matter, and what to do about them. No spam. Unsubscribe anytime.