I Watched AI Destroy a $10,000 Industry. Here’s What Survived.

In March 2026, a man died. His name was Zhang Xuefeng, and he was the most famous college admissions consultant in China. Forty-one years old. Sudden cardiac arrest. During the peak college application season, his livestreams were packed with desperate parents. He spoke fast, with a thick northeastern accent, and he never sugarcoated. ‘Don’t pick that major,’ he’d say. ‘It’s a trap.’ He didn’t sell information — he sold the nerve to act on it.

Three months later, the 2026 college entrance exam ended. He was gone, but the industry he dominated was still alive — and being quietly gutted by AI.

Every major tech company in China now offers free AI college-admission tools. Kuaike. Baidu. Tencent. Alibaba. Doubao. Zuoyebang. Even the Ministry of Education upgraded its system with a 24/7 AI assistant. Enter your scores, your rank, your subject choices, and within seconds you get a detailed ‘safe, stable, stretch’ plan. The data is official. The algorithm is transparent. The price is zero.

This means the entire business model of ‘I know something you don’t’ is collapsing. There are 3,000 universities and 2,000 majors in China. That’s a lot of information — but it’s structured, numerical, perfectly suited for algorithms. AI never gets tired. It never forgets a cutoff score. It never gets emotional. For pure data retrieval and probability calculation, AI has already won.

But here’s the dark irony. While AI was tearing down information barriers, a new kind of scam was building them back up.

One $10,000 ‘expert consultation’ turned out to be nothing more than a human reading ChatGPT’s output aloud.

China’s state broadcaster CCTV exposed the scam: agencies charging 4,980 to 12,980 yuan for ‘personalized master guidance’ — when the master was a college intern who had been hired two weeks earlier. One consultant admitted, ‘I just ask AI to write the plan and read it. I have no idea what I’m talking about.’ Parents paid for the illusion of judgment. They got a printed version of a free AI query, wrapped in a professional-sounding voice.

The worst of these agencies weren’t even in the advice business. Their real profit came from funneling students into for-profit degree programs — kickbacks of 20,000 yuan per head. They were using AI as a loss leader to bait anxious families into much more expensive traps.

So AI, the tool designed to eliminate information asymmetry, became the new engine for creating it. Free information was repackaged as scarce expertise. Sound familiar? Ticket scalpers did the same thing. Only this time, the tickets are college admission plans generated in two seconds by an algorithm.

But that’s not the real problem. The real problem is deeper, and far more uncomfortable.

When information is no longer scarce, what becomes scarce?

The answer: the willingness to take responsibility for someone else’s life-altering choice.

AI can’t give you certainty. It tells you, ‘Based on historical data, the probability of admission to this university is 67%.’ It tells you, ‘For reference only. Final decision should be based on your specific circumstances.’ It never says, ‘Trust me, pick this. You’ll be fine.’ It can’t. It has no license to. AI has no reputation to lose, no empathy to offer, no story to share.

Zhang Xuefeng’s real value was never his data. It was his ability to look a terrified parent in the eye and say, ‘Law, Chinese literature, political science. Keep your options open for civil service exams.’ No hesitation. No ‘for reference only.’ Just a decision. That certainty — even if it was imperfect — was a relief. It absorbed the anxiety of choice.

AI doesn’t absorb anxiety. It reflects it back at you, multiplied by probability.

I started thinking: is college admission really just an information problem? Or is it something else?

The consultants were never selling data. They were selling insurance — insurance against the fear of making the wrong call. They were selling the comfort of shared responsibility. ‘If it goes wrong, at least I had help deciding.’ They were selling companionship in a moment of overwhelming uncertainty.

AI can’t do that. Not because of technological limits, but because AI has no personhood, no biography, no ‘I’ve been through this too.’ It has no scars.

The real divide isn’t between AI and humans. It’s between humans who offer judgment and humans who merely pass information.

The ones who offer genuine judgment — based on deep experience, nuanced understanding of a child’s temperament, and the courage to make a call — those professionals will survive. AI can process data, but it can’t figure out if a quiet, anxious kid should be pushed into a competitive field or protected from it.

But the ones who just shuffle information — collating public data, summarizing admission brochures, listing historical cutoffs — they’re toast. AI does that better, cheaper, and with zero attitude. Their ‘expertise’ is evaporating before their eyes.

This pattern isn’t unique to college admissions. It’s happening everywhere.

Real estate agents used to charge for knowing which neighborhood was about to appreciate. Now Beike and Lianjia publish everything online. Good agents survive by being transaction facilitators and trust builders, not information keepers. Financial advisors used to get paid for stock tips. Now platforms like Xueqiu and AI robo-advisors give everyone real-time data. Real value has shifted to portfolio strategy and emotional hand-holding — keeping clients from panic-selling during a crash. Lawyers used to bill for knowing obscure clauses. Now ChatGPT can recite them in seconds. Value now lies in litigation strategy, negotiation skill, and the ability to make a judge believe your story.

The common thread: AI demolishes the information wall. What remains is the judgment wall. The first is easy to climb. The second is impossible to scale with an algorithm.

We are living through a transfer of judgment. For centuries, judgment sat with experts. They had the data, the experience, the authority. Ordinary people paid them to decide. Now judgment is shifting to every individual. AI gives everyone access to the same raw material. But the act of deciding — that still belongs to humans. And it’s terrifying.

This transfer is both liberating and paralyzing. Liberating because we finally have control. Paralyzing because control means responsibility, and responsibility means anxiety.

The best professionals in the new world will not be information authorities. They will be decision partners — people who help you bear the weight of choice.

They won’t just tell you what to do. They’ll help you understand why. They’ll lay out the risks. They’ll have a Plan B ready. They’ll sit with you in the discomfort of uncertainty. That kind of deep, personalized, human-heavy work cannot be automated. At least not in the next decade.

But let’s be honest: many current ‘experts’ won’t make this transition. It requires giving up the old model of ‘I know, you don’t.’ It requires becoming a partner, not a guru. It means abandoning the high-margin, low-effort information-delivery business for a lower-margin, high-touch judgment-support business. Most will cling to the old model until it dies.

On the day Zhang Xuefeng died, a friend posted: ‘The man who made a living promising to help people beat the system finally learned that some fates can’t be beaten.’

I disagree. Zhang wasn’t selling promises. He was selling reduced blindness. He never claimed the right major would guarantee a good life. He just said, ‘Don’t step into that obvious trap.’ That’s a huge difference. One is creating false hope. The other is removing unnecessary risk.

The truth is, life has never offered certainty. AI can’t give it. Experts can’t give it. What they can offer is the courage to decide anyway, and the wisdom to choose well under uncertainty.

AI can’t teach that. And honestly, most people don’t have it either.

That’s the real opportunity. Not to outsmart AI — but to help humans outgrow their own fear.