The human touch is becoming a paid tier

In June 2025, on a stage at SXSW London, a technology and culture conference, the chief executive of Klarna said the quiet part into a microphone. Offering human customer service, he told the audience, "is always going to be a VIP thing", and he compared it to paying more for hand-stitched clothing when a machine could make the shirt for less.

Klarna is the Swedish buy now, pay later company that announced in early 2024 that its artificial intelligence (AI) assistant was doing the work of 700 customer service agents. By May 2025 the same chief executive was admitting the company had focused too much on efficiency and cost, that quality had suffered, and that humans were being brought back, this time as gig workers rather than staff. The company cut the people, watched the quality fall, and is now preparing to sell the people back at the top of the price list, all in under eighteen months.

Last week I wrote that not using AI is a luxury: opting out of these tools is comfortable when you have money and expensive when you don't. This essay is the other half of that argument. While we argue about whether people should use AI, the market has already decided what to do with people. Human attention is quietly being repriced as the premium tier.

Kali Holloway put it plainly in The Nation this April: "As automation and AI become ubiquitous, the human touch has become a luxury good." She adds, fairly, that the two-tier dynamic is older than AI: the rich always got white-glove service while everyone else pressed 1 and 2 into a phone tree. What has changed is the speed, and the honesty. Companies used to hide the tiering behind hold music. Now they print it on the pricing page.

I watch this from inside travel, because I build Sola, an app for women who travel solo. Everything self-serve in travel is being automated as fast as the tools allow, and I am one of the people doing the automating. Meanwhile the human travel advisor, a job the internet supposedly killed twenty years ago, is having a comeback at the top of the market. Virtuoso, the luxury advisor network, reported sales up 12 percent in the first half of 2025, a 76 percent rise in consumers seeking out advisors through its website, and future bookings above $50,000 up 35 percent year on year. That is the network's own press release measuring demand for itself, so read it as direction rather than audit. But the direction keeps repeating. A Chase Travel survey of just over 4,000 travellers, run in January 2026 and reported by trade press, found that 60 percent of Gen Z and millennial travellers want to use a travel advisor. And consumer research from ASTA, the American Society of Travel Advisors, found in 2023 that half of travellers were more likely to use an advisor than before, a double-digit jump year on year. All three are intent and sentiment numbers from parties with something to sell. Still, notice which way every arrow points, and who the comeback serves.

Nobody is assigning a human advisor to a $400 booking.

Banking shows both tiers inside a single industry. Bank of America's virtual assistant Erica has passed 3 billion client interactions across nearly 50 million clients and now handles more than 58 million interactions a month. That is the mass-market tier, and the bank is proud of it. Morgan Stanley, on the wealth side, gave its advisors an assistant built with OpenAI, and its head of wealth management promised in the announcement that "trust-based relationships and human advice will always be valued by clients" and that the financial advisor will remain "the center of our wealth management universe." The same technology sits in both banks, on opposite sides of the counter. For the mass market the AI is the interface. For the wealthy the AI is the back office, and the human stays out front.

Then there is the version where you simply pay. Meta Verified launched in 2023 at up to $15 a month with live account support included, and Delta routes its best phone lines to its top frequent-flyer tier. Customers appear willing: data from Qualtrics' XM Institute, reported by the trade publication CX Dive, suggests roughly three quarters of consumers will spend extra for premium experiences and more than two thirds will pay for better customer service. The question was about better service, not specifically a human, and Meta is the warning label here: suspended Meta Verified subscribers reportedly got the same canned responses as everyone else. Paying for the human tier does not guarantee a competent human. Sometimes it just buys you a nicer queue.

The economics underneath are plain. David Autor, the economist at MIT, writes that expertise commands a market premium "if it is both necessary for accomplishing an objective and relatively scarce". Make human attention scarce inside your product and you have manufactured a premium, whether or not the humans got any better. Autor himself is optimistic about AI spreading expertise to more workers, so I will not enlist him for gloom, but scarcity logic runs in both directions. And buyers already sense it: a 2024 Washington State University study found that describing a product as powered by artificial intelligence lowered purchase intention in every product and service tested, compared with identical items labelled "high tech." In those experiments the AI label behaved like a discount sticker, which tells you how the human label behaves.

I am not watching this from the sidelines. I spend most of my working day inside an AI coding agent, I automate things for a living, and earlier this week I wrote about my first hire, a part-time chief of staff, made partly because working alone all day with an AI, however capable, was not good for me. I now pay a person for exactly the things the AI does not give me, starting with another human in the room. My own payroll agrees with Klarna's pricing page, and I am still deciding how I feel about that.

Plenty of AI-run service is genuinely good, and I have already argued that refusing it outright is a privilege. I am not asking anyone to pick a side between the bot and the person. I want the bot at 2am when a flight moves, and I want a person when something has gone properly wrong and the script has run out. The question that matters is whether reaching that person stays a default or becomes a line item, and right now the market is answering quietly while the public argument stays stuck on whether AI is good or bad. The version worth fighting for gives you both without a platinum card: the tool when speed matters, the human when judgment does, and no pricing page between you and either one. The day "talk to a person" shows up as an add-on next to extra storage, you will know exactly what your attention was worth all along.