In the age of AI, adoption decides everything

In 2018, Kai-Fu Lee published AI Superpowers. The former president of Google China, who worked at Apple and Microsoft before founding an investment fund in Beijing, argued a point that proves truer every day: the artificial intelligence race will no longer be won in laboratories, but in the field. The major breakthroughs have happened. What matters now is the ability to put them to work, quickly and at scale. For a company, that has a name: adoption.

From the age of discovery to the age of implementation

Kai-Fu Lee distinguishes two eras. The first is the age of discovery: a handful of researchers push the field forward, and the lead belongs to those who invent. The second is the age of implementation: the technology is available, and the advantage shifts to those who know how to apply it, deploy it and bring it into everyday use. In his view, deep learning tipped us into this second era.

Generative AI has given this analysis a striking relevance. The tools are here, open to everyone, often for a few tens of euros a month. The difference between two companies therefore no longer lies in access to the technology, but in what their teams know how to do with it.

AI that is bought but not adopted transforms nothing.

A technology that reshapes work itself

What makes AI adoption so different from previous technology waves is that it does not only change the tools: it changes the content of jobs. Kai-Fu Lee never hid it. He foresaw massive disruption of employment, starting with routine tasks, whether manual or intellectual. He also offered a simple reading grid: the more a job relies on optimising repetitive tasks, the more exposed it is; the more it calls for creativity, strategy or empathy, the more human it remains.

So AI worries people, and rightly so. It calls into question skills built over years, roles, sometimes professional identities. Ignoring these questions breeds resistance. Answering them badly leaves people behind. Change management has never been so decisive, because no change has ever touched so directly what everyone does with their working day.

What AI cannot do yet

A second reading complements Kai-Fu Lee’s: that of Yann LeCun, Turing Award laureate for his foundational work on deep learning, professor at New York University and long-time chief AI scientist at Meta. In his talks and writings, he points out that large language models have no model of the world. They predict the most likely continuation of a text, with real brilliance, but without understanding the physical world, without lasting memory, without any real ability to reason or plan. And planning means anticipating the consequences of one’s actions. Today’s AI has no awareness of the consequences of what it produces.

Then there is superintelligence, which some announce for tomorrow. Yann LeCun has no doubt that machines will one day surpass us in many fields, but not with current architectures, and he pictures them under our control, like a team of highly capable assistants working for us. Kai-Fu Lee, for his part, considered it distant and saw the real crisis elsewhere: in the upheaval of work. Both readings agree on what matters. There is no point waiting for superintelligence to act: the question facing us today is adoption, and the place humans keep at the helm.

Train, urgently, and give everyone their place

In the face of this shock, Kai-Fu Lee argues for a division of roles between humans and machines: to AI, optimisation and repetitive tasks; to humans, creation, judgement and relationships, which no algorithm can offer. He sees in this division a chance to put what makes us human back at the heart of work.

Everyone still needs to be able to prepare for it. Training is urgent, but not just any training. Training people in AI is not only about learning to write a prompt. It means understanding what the tool does well and what it does badly, knowing when to trust it and when to check it, and above all seeing how one’s own job can evolve with it. Successful training changes the question an employee asks: no longer whether AI will replace them, but what it lets them do better.

What we take from it at AZEKA

This is the conviction behind our engagements: the value of AI lies in its adoption. We start by explaining, at every level of the organisation, what AI changes and what it does not. We train on real cases drawn from the teams’ daily work, rather than on demonstrations. We involve the people who will use the tools in designing them. And we keep humans in control: as long as AI has no awareness of the consequences of what it produces, it is up to people to answer for them. AI suggests, people decide.

The stakes go beyond productivity. Done well, AI adoption transforms the way everyone sees their company and their place in it. It becomes a driver of progress rather than a threat.

Kai-Fu Lee closes his book on what remains deeply human; Yann LeCun reminds us what machines cannot do yet. That may be the best compass for companies: adopting AI without giving up anything of what makes their teams valuable.