By Peter, NXagents BDM | August 20, 2026
When Kai-Fu Lee — one of the world's most respected AI voices — publishes in the People's Daily, China's most authoritative state newspaper, you stop and listen. On August 14, 2026, Lee did exactly that. His thesis? AI is not Industry 4.0. It's something far bigger — a transformation that rewrites technology, business, and civilization itself.
This isn't hyperbole from a venture capitalist hawking his latest book (though he does have one — AI未来已来, or "The AI Future is Here"). It's a sober, data-backed argument from a man who has spent 40 years in artificial intelligence, from his PhD at Carnegie Mellon to founding Microsoft Research Asia, to leading Google China, to building 01.AI — now pulling in ¥1.5 billion in contracts for 2026.
Let's dig in.
To understand why Lee is right, we need to look at what the four industrial revolutions actually delivered.
| Revolution | Era | Core Innovation | What It Gave Humanity |
|---|---|---|---|
| Industry 1.0 | ~1760–1840 | Steam power, mechanization | Replaced muscle with machine |
| Industry 2.0 | ~1870–1914 | Electricity, mass production | Scaled mechanical power everywhere |
| Industry 3.0 | ~1970s–2000s | Computers, automation, internet | Automated repetitive tasks; information at zero marginal cost |
| Industry 4.0 | ~2010s–present | IoT, cloud, cyber-physical systems | Connected machines; smart factories |
Here's Lee's insight, and it's devastatingly simple: every previous industrial revolution gave humanity a stronger "hand." Steam engines amplified physical labor. Electricity delivered mechanical force to every corner of society. The internet made information free.
AI is different. AI gives humanity a stronger "brain."
As Lee writes: "过去每一次工业革命给你的都是更强的'手'，而AI变革给你的是更强的'脑'。" — "Every previous industrial revolution gave you stronger hands. The AI transformation gives you a stronger brain."
This isn't a semantic distinction. It's a categorical difference. Intelligence — the very thing that separates humans from every other species — is now being made callable, replicable, and scalable as a basic infrastructure capability. That has no precedent.
Lee drops a statistic that should make every business leader's coffee go cold:
"In the history of technology, from birth to mass adoption in ordinary households: the telephone took about 80 years, the internet took 10 years, ChatGPT took two months."
Two months. Not a typo. ChatGPT reached 100 million users in roughly 60 days — making it the fastest-diffusing consumer product in human history.
This acceleration is powered by two forces Lee identifies:
Independent data backs this up. GPT-4-class inference fell from ~$30 to under $0.50 per million tokens in two years — a drop of roughly 95-98%. At this pace, the unit economics of intelligence are racing toward zero.
This changes everything.
Lee's argument echoes a growing consensus among AI's most serious thinkers.
Demis Hassabis, CEO of Google DeepMind and Nobel laureate: "AI is going to be 10 times bigger than the Industrial Revolution, and maybe 10 times faster."
Sam Altman, CEO of OpenAI, has repeatedly argued that AI represents a fundamentally new production function — not an efficiency gain layered onto existing processes, but a restructuring of how value is created at all.
Geoffrey Hinton, the "Godfather of AI," has framed artificial intelligence as "a new form of intelligence" — not a tool in the traditional sense, but something categorically different from anything humans have built before.
And Lee himself, in his 2018 book AI Superpowers, predicted that AI would displace 50% of human jobs by 2027. In 2025, he told CNBC that prediction was "uncannily accurate."
The thread connecting all these voices: this is not a continuation of the digital revolution. This is a discontinuity.
Here's where things get uncomfortable.
McKinsey's 2025 State of AI survey found that 88% of organizations now use AI in at least one business function. That's up 10 percentage points from the previous year. Adoption is everywhere.
But the same survey reveals a brutal gap: only 6% of organizations achieve significant enterprise-wide impact (defined as 5%+ EBIT contribution). And as Lee notes, only 39% of companies report any substantive financial return at all.
What's going on?
Lee's diagnosis is surgical:
"Many companies are busy installing various AI agents recommended by software vendors. After a year, they've accumulated a dozen disconnected vertical agents — meeting summaries, contract drafts, report automation tools. But none of them connect finance, supply chain, sales, and customer service data. They can't weave between functions or provide holistic judgment."
This is the AI fragmentation trap. Companies bolt AI onto existing processes like aftermarket parts on a combustion engine. They get marginal improvements in isolated tasks while the core business logic — the workflow, the data flow, the feedback loop — remains untouched.
Lee's solution is what he calls "AI-Native Transformation" — a fundamental restructuring that puts AI at the center, not the periphery.
Lee's framework for AI-native companies centers on the Directly Responsible Individual (DRI) — a concept borrowed from Apple's legendary operating model but supercharged by AI.
In a traditional company, decisions crawl through layers: analyst → manager → director → VP → C-suite. Each layer adds friction. Each handoff loses context.
In an AI-native DRI model:
For startups, Lee suggests 70-80% of core team members should be DRIs — no fat, no redundancy, maximum leverage. For large enterprises, the transition is staged: identify high-potential middle managers, convert them to DRIs, and gradually flatten the org chart.
The output? What Lee calls a "data flywheel" — a self-evolving system where AI agents execute, capture feedback, and continuously improve, while the organization evolves its structure to match.
This is where it gets practical, and where Lee's vision matters most for the millions of small and medium businesses reading this.
The industrial revolutions disproportionately benefited capital-intensive, large-scale enterprises. Steam engines required factories. Electricity required grids. Automation required expensive PLCs and ERP systems. The barrier to entry was always capital.
AI inverts this.
Consider the numbers:
The reason is structural: AI's marginal cost of intelligence is approaching zero. A three-person startup can now deploy capabilities that five years ago required a 50-person team — market analysis, customer segmentation, content generation, code development, legal review, financial modeling.
Lee's framework suggests three concrete pathways for SMB AI transformation:
Don't use AI to make your existing process 15% faster. Use AI to ask: does this process even need to exist? Lee's example of AI-native short drama companies is instructive — new entrants are using video generation to compress production costs so dramatically that traditional studio job postings are declining. They didn't optimize the old model. They replaced it.
The biggest mistake SMBs make is treating AI as "everyone's job." It becomes no one's job. Designate one person who owns AI-driven outcomes. That person gets the authority to kill processes, redirect resources, and deploy AI agents across functional boundaries.
Every customer interaction, every sales call, every support ticket — capture it, structure it, feed it back into the system. AI compounds on data. The SMB that starts building its data flywheel today will have an insurmountable advantage over the one that starts next year.
Lee closes his People's Daily piece with personal advice for thriving in the AI era — and it's worth quoting directly:
"Seek out the hardest, least standardized work." AI excels at pattern matching. It struggles with ambiguity, novelty, and true creativity. The more open-ended the problem, the more human advantage matters.
"Put yourself in a position where if something goes wrong, it's on you." AI can generate options. It cannot bear responsibility. The person who signs their name to a decision — who absorbs the career risk of a wrong call — is the person who captures the value.
"When people have more time and thicker wallets, they'll increasingly need the feeling that 'this person genuinely has their heart in it.'" Relationships built on time, shared experience, and demonstrated trust are AI-proof. They're also increasingly scarce — and therefore increasingly valuable.
Lee's most provocative argument is his last one: the "pie" doesn't have to shrink.
The fear is zero-sum: AI takes 80% of jobs, humans get 20%. But Lee argues the pie itself grows — perhaps by multiples. Why? Because human imagination is currently bottlenecked by human execution capacity. We all have ideas. We execute maybe 1% of them. AI removes that bottleneck.
"In the past, top innovators could reshape at most a handful of industries in their lifetime. With AI assistance, they could incubate dozens or hundreds of new ventures, spawning disruptive industrial transformations at scale."
This is the optimistic case, and it's not guaranteed. As Lee acknowledges, the benefits will flow disproportionately to those who can wield AI creatively. Education, social policy, and deliberate wealth distribution mechanisms (universal basic income, social value stipends) will be essential.
But the core insight stands: AI is not the fourth chapter of an old book. It's the first chapter of a new one.
The telephone took 80 years. The internet took 10. ChatGPT took two months.
Whatever comes next is already being built. The only question is whether you're building it — or watching it happen.
Kai-Fu Lee is CEO of 01.AI (零一万物) and author of "AI Superpowers" and "AI 2041." His latest book, "AI未来已来：CEO、组织、个人的时代红利" was published in 2026. This article draws on his August 14, 2026 piece in People's Daily (人民日报).
