Jobs That Will Survive and Disappear in the AI Era, Kunlun Tech Chairman Declares

“Experience is no longer a defense in the AI era.” This statement by Fang Han, Chairman and CEO of Kunlun Tech, at the “2026 China AIGC Industry Summit” on May 23, 2026, sent ripples through the audience. A seasoned entrepreneur who has witnessed the cycles of the internet and mobile internet, Fang offered a vision of humanity’s role in the AI era that was not alarmist but grounded in a sharp and pragmatic understanding of reality.

Token Consumption as a New “Performance Metric” At the start of his speech, Fang highlighted the growing importance of “token consumption” as a performance indicator. While general users consume millions of tokens per month, technicians engaged in AI coding use billions, and heavy users of AI agents can easily burn through tens of billions of tokens monthly. As a concrete example, Fang referenced reports that Peter Steinberger, founder of “Lobster,” consumes approximately 600 billion tokens monthly on OpenAI’s platform. Fang pointed out that even within the same company, token consumption can differ by several orders of magnitude between general users and heavy users, and this divergence is becoming a new “hardcore performance benchmark” in the AI era.

Experience is No Longer a “Defense” Fang challenged the optimistic belief that “experience will always be valuable in the AI era.” He argued that extensive experience in an industry can sometimes foster resistance to adopting AI. While methods for “distilling” skills from AI are gaining attention, the inverse—trying to resist AI through individual effort—is ultimately not sustainable. Fang proposed a framework for categorizing professions based on two axes: whether the role involves a “closed-loop process” and whether it can tolerate errors. Tasks that are both closed-loop and error-tolerant are more easily replaced by AI. Examples include certain artistic tasks, programming, middle management in white-collar jobs, and junior apprentice-level roles. Conversely, jobs where errors are intolerable—such as surgeries, on-site construction, and blue-collar roles requiring high judgment and responsibility—are harder to replace. Fang also referenced AI researcher Andrej Karpathy’s perspective: “Let AI handle 80% of the work, leaving 20% for humans. The remaining tasks for humans involve judgment and verification.” In AI coding, the critical skill is not how extensively one uses AI but the ability to determine what needs to be done.

Five Types of People AI Will Never Replace One of the most attention-grabbing points of Fang’s speech was his classification of human roles that AI cannot replace. He identified the following five types: 1. Storytellers — The essence of content is storytelling, and the ability to weave meaningful narratives from AI-generated materials is hard to replicate. 2. Innovators — Those who create new concepts from scratch, beyond extracting patterns from existing data. 3. Aesthetes — The human ability to define beauty and judge what is “good” transcends quantifiable measures and remains uniquely human. 4. System Builders — Individuals who can design and integrate entire systems, rather than focus on individual components. 5. Paradigm Shifters — Those who can fundamentally reevaluate existing frameworks and establish new ways of thinking. What these five types have in common is that they operate outside “closed loops” and require high levels of judgment and creativity, making them difficult to automate with AI.

AI is Eliminating the “Middle Rungs” of Careers Fang identified another significant shift caused by AI: the compression of individuals’ career growth paths. Traditionally, career progression within companies followed clear stages, from entry-level to mid-level to senior roles. However, AI is erasing the intermediate stages, creating a binary classification of workers as either novices or experts, with little room in between. General employees might consume millions to tens of millions of tokens per month, while technicians consume billions. This is creating an unprecedented gap between those who can effectively utilize AI and those who cannot. Fang noted that AI is compressing the “career ladder,” removing the middle rungs entirely. How ordinary workers will adapt to this new reality is, according to him, “a phenomenon worth observing.”

A Cost Revolution Reshaping the Content Industry Fang also shared specific figures on how AI has revolutionized costs in content creation. For instance, Kunlun Tech used to spend approximately 100,000 yuan (about 2 million yen) to produce a single piece of music for video games. This involved hiring composers, lyricists, studio musicians, and sound engineers. But with the company’s AI music model, Mureka Agent, the cost of producing a song has dropped to just 1 yuan (about 20 yen). Similar transformations are occurring in short drama production. While a live-action short drama might cost around 1 million yuan (about 20 million yen) to produce, the current cost for an AI-generated short drama stands at 100,000 yuan (about 2 million yen). Fang predicts this could drop to 10,000 yuan next year and 1,000 yuan the year after. As a result, the number of AI-based short drama production teams is rapidly increasing, while live-action productions are declining. While this trend has sparked concerns about actor unemployment, Fang noted that the overall scale of the content industry is likely to expand.

Advocating for the “Second-Mover Strategy” Fang offered intriguing advice on corporate AI adoption strategies: most industries should aim to be the second mover rather than the first. The first movers bear high costs for experimentation and exploration, while third movers risk being left behind and missing out on industry benefits. However, he added a caveat: in the IT sector, “only first movers can compete; this is the harsh reality.” That said, Fang suggested that AI has the potential to change this dynamic. AI levels the playing field, allowing all players to start from the same position and potentially reshuffling traditional hierarchies.

”Overestimated in the Short Term, Underestimated in the Long Term” As he wrapped up his presentation, Fang offered a sober assessment of the current AI boom. “It’s a classic case of being overestimated in the short term and underestimated in the long term,” he said. While AI demonstrations are impressive, significant challenges remain in achieving actual productivity gains. For example, studies show that while AI programming improves efficiency when project requirements and outcomes are clear, productivity does not improve when requirements are ambiguous. Reflecting on the advent of the internet in 1995, Fang noted that while it was predicted to spell the end for newspapers and publishing, the content industry actually expanded significantly, with the decline of these traditional sectors taking more than 20 years to manifest. Each new technological breakthrough brings predictions of obsolescence for older industries, but in reality, the pace of change is slower than expected, although its eventual impact often exceeds expectations.

Skepticism Toward Predictions of AGI Arrival Fang did not hold back in criticizing bold predictions about the arrival of AGI (artificial general intelligence) made by some prominent individuals and companies. “Any statements about AGI ‘coming soon’ made by companies raising funds or preparing for IPOs are entirely fabricated to boost their valuations,” he declared. He pointed out that those who predict earlier AGI arrival dates tend to have a vested interest in increasing their market capitalization. Interestingly, he noted that the only person forecasting a relatively distant timeline for AGI is DeepMind’s leader, Demis Hassabis, who predicts its arrival in 2034. Fang warned against being misled by statements driven by short-term financial incentives.

The Beauty of an Unpredictable Future Fang concluded his talk with a message for the audience: “The greatest charm of the future lies in its unpredictability.” He urged attendees to avoid being swayed by alarmist rhetoric or overly optimistic forecasts. Instead, he encouraged them to calmly assess whether their own roles are susceptible to being replaced by AI and to focus on honing their judgment and creativity as the keys to thriving in the AI era. Drawing from decades of experience in the evolution of the internet, Fang’s words carried the weight of a seasoned practitioner’s balanced perspective.