AI agents poised to replace humans as basic units of a company, Kai-Fu Lee says
AI agents can serve as ‘Lego blocks’, allowing CEOs to build and manage companies like a scalable machine, Lee says at a UBS summit
“The basic unit of a company will evolve from a human being to an AI agent,” Lee said on Thursday at a summit on disruptive technologies hosted by Swiss bank UBS. AI agents are software apps that use AI to autonomously execute tasks and achieve goals on behalf of users.
Lee pointed out that AI agents could operate around the clock, be replicated infinitely, and scale effortlessly – capabilities unmatched by human workers. “If you have a super employee, you can’t replicate [them], right? Human cloning is not legal, but AI agent cloning is perfectly fine, and they will scale,” he said.
“You can completely use agents as Lego blocks,” he said. “So you have a Lego block that’s [human resources], a Lego block that’s legal, a Lego block that is finance, and then a Lego block for customer service, et cetera.”
“Then you can have a huge, giant Lego-created machinery that is your company agent, where the CEO interacts and manages the company, and that’s what [OpenAI CEO] Sam Altman means when he says there will be US$1 billion companies [built by solo entreprenuers armed with AI agents].”
Lee said AI agents would liberate people from repetitive and routine tasks, and “that’s why agents are so critical for both bringing an end to some jobs, but also [fostering] a beginning for our creativity.”
When it comes to developing agents, Lee highlighted the advantages of open-source AI models, in which Chinese companies like DeepSeek play a leading role. Those models enabled companies to fully customise, enhance and secure their AI agents, he said.
“Open source makes a much better ecosystem,” Lee said. “The model is yours; it’s free … You can inspect it, modify it [and] improve it.”
While open-source initiatives originated in the US, most American companies did not open-source their large language models, primarily because firms like OpenAI must justify the high valuations of their businesses to investors, Lee said.
“Now that China’s [embracing] open source, I think it’s really going to be a neck-and-neck competition in regions outside the US and China because Chinese models are free, available and open.”
Chinese companies could start making money from AI agents next year, UBS says
As advanced models like DeepSeek’s R2 become more sophisticated, UBS expects ‘to see more monetisation’
“2026 is the year of agent monetisation in China,” said Sundeep Gantori, an equity strategist at UBS Global Wealth Management CIO, on Thursday. As advanced models like DeepSeek’s R2 become more sophisticated, “we expect to see more monetisation”, he added.
AI agents are widely considered to be the next evolutionary stage of generative AI and some of the largest tech companies in the world, like Google and OpenAI, have been hotly competing in the space. On Wednesday, German translation start-up DeepL said it launched an AI agent that can complete “repetitive, time-intensive tasks across a wide variety of functions”.
