Cai Fang: AI's employment impact has yet to show up in macroeconomic data; we must seize the window to prepare in advance

"In the future, employment will become increasingly flexible and take on new forms, and the share of non-unit employment will grow larger and larger. We must therefore explore new forms so that these labor market institutions, worker rights protection systems, and social security systems are consistent with the new forms of employment," Cai Fang, chairman of the Academic Committee of the China Finance 40 Forum and a member of the Chinese Academy of Social Sciences, said recently at the 2026 Bund Summit.

He also suggested that resolving structural employment contradictions and responding to future AI shocks should be incorporated into the overall strategy of "investing in people."

"How are the potential employment shocks we face different from those in history? The biggest difference is that the employment shocks in history were other people's history, whereas what we face today is our own history and our own reality," he said.

Cai Fang further said, referring to historical situations such as the "Engels' pause" during the First Industrial Revolution, that in history, when new technologies emerged, there was a time gap between job destruction and job creation. "But in history books and in the data, we may have overlooked this, (seeing only) that new technologies will ultimately create more jobs."

He believes that, globally, including in China, the AI employment shock has not yet been reflected in macroeconomic data, but some recruitment data and micro-level links have already shown signs. At present, the impact of AI is in a brief quiet period, which gives us a rare opportunity to prepare response measures in advance.

Considering AI's impact on productivity, Cai Fang mentioned observations at three levels. The first is the unevenness of productivity gains—as with any past technological progress, it is certain that artificial intelligence will raise productivity through multiple channels, but the application of artificial intelligence is uneven across industries, sectors, enterprises, time, and different processes and chains of labor tasks.

"When you reduce labor input at one link in the chain to save costs, job substitution begins. However, productivity will be constrained by overall equilibrium and may not rise immediately, which will produce phenomena such as inconsistency between productivity and employment shocks. This is what we need to address," he said.

Second, he cited relevant research as saying that AI's improvement of labor productivity will follow a J-shaped pattern of change. Considering that labor substitution may already have occurred before productivity has improved, preparations to address employment problems must be made in advance.

"In a particularly bad scenario, perhaps during the 'short leg' stage on the left side of the productivity J-curve, job destruction, slower income growth, and widening income gaps may already occur, thereby damaging the national economic circulation and instead hindering the improvement of labor productivity," Cai Fang said.

Finally, he emphasized that technology itself cannot solve the negative socioeconomic impacts of artificial intelligence, especially the employment impact, and that government intervention, financial participation, and other forms of coordination must be used to solve the problem of AI for good.

"At least for emerging markets, we need government policy intervention and institutional building more than at any previous time," he said, putting forward policy recommendations from several aspects.

First, policy should focus on promoting AI equality. This is the key to ensuring that labor productivity improves and avoiding the so-called Solow paradox.

Among these, the urgent task is to resolve the structural employment contradiction involving "the elderly and the young." "The emergence of new technologies in any era creates gaps and so-called K-shaped divergence. Artificial intelligence may bring an unprecedented gap." Combining this with the characteristics of China's labor market, he said that a large part of the structural employment contradiction is reflected in "the elderly and the young": in the face of AI, the entry-level skills that young workers previously possessed are depreciated, while the technological gap that older workers did not previously face may widen.

Considering the iteration of artificial intelligence technology, changes in its impact, and the expansion of its coverage, employment support policies should be continuously upgraded.

Second, focusing on the issue of AI alignment, policy should guide artificial intelligence well and require AI alignment comprehensively from social, macroeconomic, and industrial dimensions.

"The artificial intelligence we see and talk about now is mainly some foundation models. In the future, what will really play a role and affect our lives is the practical application of these models, and this is likely to have the greatest impact on workers. Therefore, in this process, guidance, incentives, and regulation should be fully upheld," Cai Fang said. From an employment perspective, AI should be made an enhancer of workers' capabilities, not a destroyer of workers' jobs.

Third, industrial policy should focus on cultivating in advance those industries with high demand elasticity but currently low labor productivity, such as services.

Cai Fang said: "We raise overall labor productivity precisely to give those industries with high demand elasticity room to survive. To this end, industrial development, job creation, and people's employment, income, and individual labor productivity should be appropriately decoupled. This is a direction for the future and also a buffer to prevent employment shocks today. From this, we can obtain new demand, new markets, and new jobs, all of which come from people's all-round development."

Fourth, policy should start with addressing structural employment contradictions. Among these, the first thing to consider is how labor market institutions should connect with new forms of employment.

"Because in the past, China's labor market institutions and social security system were closely linked to employment units. As employment becomes increasingly flexible and takes on new forms in the future, and the share of non-unit employment grows larger and larger, new forms must be explored so that these labor market institutions, worker rights protection systems, and social security systems connect well with the new forms of employment. Therefore, there are many urgent institutional building goals here. At the same time, our social security system and social welfare should become more inclusive." He suggested that under the influence of artificial intelligence, it will become increasingly difficult to identify whether an employment shock is an individual problem or a macroeconomic problem, and that maximum coverage of all workers should be achieved. In short, resolving structural employment contradictions and responding to future AI shocks should be incorporated into the overall strategy of "investing in people."