Automatic for the People

By early March, 31 provinces and municipalities across China had incorporated AI models in some aspects of governance. These “AI+civil services” have streamlined public services like applying for a taxi license. The next step is to explore AI’s role in more complex decision-making processes.　

“AI employees are assistants to human staff. They help manage public affairs and provide public services,” said Luo Gengbiao, director of the digital governance bureau in Shenzhen’s Futian District.　

However, both experts and the public remain cautious about how data should be used and the boundaries of AI’s role.

According to Luo, their AI employees work well gathering public feedback and requests from hotlines and government platforms. “AI employees can quickly generate a daily report of all public requests from the past 24 hours and mark which have been resolved,” Luo said.　

In Nanchang, Jiangxi Province, residents can now easily follow AI-guided steps to apply for a taxi license. Zhu Xian, head of the city’s digital governance department, told NewsChina that their AI use is mainly answering queries and search services related to government affairs.　

Tian Lei, a district-level prosecutor official in Wuhan, Hubei Province, said his office is piloting DeepSeek-based applications, such as working out rulings for similar cases, drafting statements for simple cases, searching for laws and regulations and analyzing evidence.　

According to Li Junmin, digital governance department director of Shanwei, Guangdong, the purpose of using AI employees is improving efficiency. Despite the daily volume of hotline calls doubling from 400-500 several years ago to over 1,000, the Shanwei government has not needed to hire more staff. “AI answers the public’s questions faster by searching its massive database,” he said.　

Li Xiaoming, data management director for the digital governance department of Shenzhen’s Futian District, added that their AI assistants achieve over 95 percent accuracy in processing paperwork, reducing review time by 90 percent. Li noted that an AI assistant once generated a PowerPoint presentation for him in just 20 seconds, only needing some light revisions.　

Ma Liang, a professor at Peking University’s School of Government, believes AI will play an even greater administrative role, eventually supporting more complex decision-making.　

“Intelligent decision-making may be the most difficult area for AI in government,” said Liang Zheng, deputy director of Tsinghua University’s Institute for AI International Governance. “In the past, decisions were generally made based on the decision maker’s own experience and judgement. But given information limits, decisions were sometimes imprecise. If AI can help the decision-making process with its massive datasets, outcomes could be quicker and better, though achieving that is quite hard,” Liang said.

In 2021, Nanchang had over 180 employees handling government affairs. That number has since dropped two-thirds to 60. “As AI technology is applied to government affairs, staffing will continue to shrink,” Zhu Xian said, “Some simple jobs, such as answering questions at a reception desk or sorting information from the public and businesses, will quickly be handed to AI.”　

But Zhu also stressed that not all roles can be replaced quickly. “It takes time for people to adapt to new things. Elderly residents may be unwilling to deal with a cold machine. Meanwhile, humans are needed to deal with complex issues,” Zhu said.　

Li Xiaoming sees AI as a tool to free human employees to do more creative and face-to-face work. “For example, if AI can efficiently generate legal rulings, human arbitrators can spend more time focusing on the emotional needs of applicants and resolving social disputes more effectively,” he said.　

However, Liang Zheng warned that even if new jobs are created, they may not be in the same departments or suited to existing employees. “What will happen to people who have spent years mastering jobs that can be done by AI?” he asked, adding that even if employment numbers do not change, structural shifts could affect salaries and social status, and it will take time for people to adapt to these changes.　

According to Ma Liang, future recruitment will prioritize applicants qualified to work with or manage AI systems.

So far, most local governments operate their AI systems offline, separated from external networks. In Shenzhen, for example, encryption technologies have been adopted to protect data from leaks during transmission or storage.　

To avoid offloading responsibility entirely onto AI, Shenzhen has assigned each AI employee a human supervisor, who makes the final decision on each case.　

Experts warned that special precautions should be taken for AI applications in certain areas, particularly in resource allocation. “For example, when granting subsidies to the elderly or children, human judgment is essential, as any loophole or mistake could have serious consequences,” Liang said.　

Tian pointed out that AI still falls short in sentencing recommendations, often making decisions that do not align with the actual situation. “This is because their training data is not enough and also because machines can’t grasp social context, like the reasons behind a crime or whether the defendant is a minor,” Tian said.　

In September 2024, Shenzhen’s Futian District introduced a framework for managing AI in government affairs, covering ethical guidelines, technical standards, range of applications and safety protocols.　

It proposes classifying AI use cases by risk level to guide management, prohibits AI systems from designing or managing other AI employees, and emphasizes that AI employees must always act in the best interest of humans and under human command. In addition, it calls for setting up an ethics commission to review, supervise and manage the scientific and moral integrity of each application scenario.　

Once refined, this framework should be adopted nationwide, Ma said.