Kai‑Fu Lee, former head of Google China, chairman of Sinovation Ventures and founder‑CEO of 01.AI, warns that corporate AI transformation will bring more radical change than most chief executives expect. Lee, whose 2018 book anticipated much of the US–China tech competition, argues that AI could again reshape the economic rivalry between the two countries if companies and leaders do not adapt.
“Move fast, but make sure safety keeps pace”
Lee supports the principle that advances in AI capability must be matched by safety and governance improvements. He stresses that regulatory frameworks inevitably lag technological development, making industry collaboration especially important now. Responding to Anthropic’s Dario Amodei’s call to “pace the frontier,” Lee reframes it as: “Move fast, but make sure safety keeps pace.”
Two clear problems from CEO conversations
From his transpacific discussions with chief executives, Lee draws two clear conclusions: everyone wants an AI transformation, and most are doing it wrong. His new book is presented as a roadmap for doing it right.
Transformation must start at the top
Lee says CEOs face a “Darwinian choice”: embrace AI or suffer the consequences. Firms with largely digital assets — such as entertainment, gaming and internet companies — will be “dead in the water” if they fail to adapt. Companies farther from the AI frontier, like coal mines, farms or factories, may have a three‑to‑five‑year buffer before they feel major disruption.
CEOs should ask whether their AI efforts pass the “earnings call test”: if AI initiatives haven’t improved profits, they are merely “AI theatre.” Lee argues the common mistake is delegating responsibility to the chief information officer. “Only the CEO can define the ambition, concentrate resources, redesign the operating model, remove barriers, and remain accountable for the result,” he writes, so AI transformation “can only be done from the top.”
He also says each CEO must use AI effectively themselves to demonstrate they understand its importance to the company’s bottom line.
Boss AI and radical transparency
Following his own advice, Lee designed a decision assistant composed of 19 agents to help run 01.AI. The company packaged the system as a commercial product called Boss AI. At its core is what Lee calls a “truth engine” that records, organizes and analyzes every internal meeting, sales call and customer service request to reveal what blocks the company from meeting its goals. That provides CEOs visibility across the business and a “promise ledger” to hold people accountable for commitments.
Lee says middle managers often hope the CEO will forget, “and the CEO does forget, but the CEO’s AI does not.” He points to strong resistance at the middle‑management level: there are “antibodies” in companies — people who simply do not want AI transformation. Middle managers feel vulnerable because they are not technology experts and fear the boss will install someone who is; they also fear losing power if direct reports are laid off.
Despite predicting a split between “irreplaceable humans” and “displaceable humans,” Lee cautions that companies should not measure AI success by headcount reductions. “The winners will not measure success by payroll cuts. They will measure it by what their best people, amplified 10 times, can now do that was previously impossible.” Most people will not be fired if they learn new AI skills; Lee adds they may be paid three or five times more than before.
He still advocates flattening hierarchies that protect change‑resisters: with transparency, “there’s no need to hide, nowhere to hide, and everybody talks about the real issues and tries to solve the problems.”
Culture, geography and AI adoption
Lee frames his ideas as exhilarating for employees willing to sign up: improved morale, culture, execution, results and stock price could form a virtuous cycle. He acknowledges radical transparency “takes some getting used to” and that some corporate cultures are reluctant to expose themselves to an all‑seeing boss AI.
Lee argues the Boss AI concept is more acceptable where the CEO commands great respect and power and organizations operate by consensus. “Everything else being equal, Asian companies will tend to adopt it earlier and faster. And we’re seeing that.” 01.AI has been selling Boss AI in China and across Asia and, Lee says, has encountered “no resistance” to recording meetings or capturing employee communications. Privacy is important in his region, but not as overriding as in many Western markets — a “second Asian advantage,” he calls it.
“Asian companies are more collectivist,” Lee says, predicting many Chinese employees will tolerate companies recording “everything all the time” if they perceive business benefits. Western workforces, he predicts, will not accept that level of surveillance, which could leave their CEOs without the vast data available to Asian rivals. He links this to his earlier assertion that “data is the new oil.”
Where the AI superpowers stand now
Lee notes that Chinese firms such as DeepSeek and Moonshot already offer meaningful competition to Anthropic and OpenAI, even if they do not fully match the frontier US models. He warns it would be a mistake to assume China will continue to lag Silicon Valley.
“If you look at the top 1,000 people in AI in the US, they are significantly better than the top 1,000 people in China. However, if you look at the top 100,000, the two countries are very comparable,” Lee says. He therefore cautions against underestimating Chinese companies.
Lee doubts that US export controls on Nvidia’s most powerful chips will secure a lasting advantage. “The export controls do not work,” he says. Chinese engineers’ skill in producing “good enough” solutions has been underestimated, and export controls may spur domestic investment. Lee predicts China will be “completely self‑sufficient” in chips within two to three years.
While some US tech CEOs debate slowing development because of hypothetical future risks, Lee says Chinese leaders are more practically focused on building applications, driving industry adoption and finding ways to monetize AI — though they do not dismiss long‑term risks.
Privacy limits for always‑on devices
Lee expects Western cultures’ greater reluctance to trade privacy for utility to pose a problem for always‑on, listening AI devices such as the product Jony Ive is reportedly developing with OpenAI. “It will not work because people won’t fully accept it,” he predicts.
In sum, Kai‑Fu Lee urges CEOs to take personal charge of AI strategy, implement radical transparency to improve accountability and performance, and recognise that cultural differences and national industrial policies will shape how the AI race unfolds globally.
