Shenzhen sets 2028 targets for AI in classrooms as part of city-wide action plan

Shenzhen has set out a three-year push to weave artificial intelligence through its economy and public services, including a blueprint for human-AI collaborative classrooms. On 4 September 2026, the city’s Industry and Information Technology Bureau, Development and Reform Commission and Science and Technology Innovation Bureau jointly issued the Action Plan for Advancing Artificial Intelligence and Application Development (2026–2028).

By 2028, the plan calls for deep integration of AI across six priority areas — scientific research, industry, consumption, public wellbeing, governance and global cooperation — with adoption of next-generation smart terminals and AI agents exceeding 90%, a core AI industry above RMB 300 billion (roughly US$42 billion), and more than 3,000 AI companies in the city. The document sets out 23 tasks across the six directions.

On education, the plan asks for a change in how students are taught: building a future education model of human-AI collaboration, virtual-real symbiosis and instruction tailored to each learner. It backs AI agents for courseware generation, homework marking and learning analytics to cut teachers’ administrative load, and encourages students to use agents for personalised learning through intelligent tutoring, Q&A support and virtual teaching assistants. It also promotes contextual, self-directed learning and AI-supported, problem-based projects in which students use AI tools to deepen practical skills through hands-on exploration and teamwork.

Education sits inside a much broader municipal agenda. The plan also covers AI in elderly care and childcare — companion robots for emotional support, smart monitoring for infants, and early development assessment — and in consumer technology, where it backs AI phones, AI PCs, AI glasses and humanoid robots alongside “AI-native” services such as digital avatars and AI companionship.

The action plan implements the State Council’s opinion on the “AI+” initiative and Guangdong province’s all-sector application programme, and is framed as part of building Shenzhen into an “AI pioneer city”. It leans on deployment rather than model-building alone: the scenario agenda is paired with support for frontier algorithm research — world models, embodied intelligence, long-term memory and agent reasoning — and for an autonomous, controllable “chip–framework–model” stack, with open sourcing of training frameworks, inference frameworks and model parameters.

Shenzhen also frames the plan globally. It will use the Qianhai and Hetao cooperation zones to deepen international collaboration and extend the “Hong Kong research, Shenzhen commercialisation” model, and aims to showcase Chinese AI solutions — including pushing the country’s AI stack into Latin America, Southeast Asia and Central Asia — around the 33rd APEC Economic Leaders’ Meeting.

For education companies, the signal is that municipal procurement and city-level pilots, not just national policy, will shape demand for classroom agents over the next three years. The 90% adoption target applies to smart terminals and agents across the city rather than to schools alone, but the education provisions make clear that lesson generation, marking and learning analytics are among the use cases the local government intends to fund and scale.