Two Sessions puts AI inside the work report

The Government Work Report delivered to the National People’s Congress on 10 March contained the expected list of targets, and one sentence that will outlast all of them: a commitment to an ‘AI-plus’ action plan, integrating artificial intelligence across industry, health, education and public services. The phrase is deliberately dry, in the manner of all documents that will be quoted by a thousand procurement departments. Behind the dryness is a decision about where the next cycle of Chinese investment is going, and reading the report as an allocation of money rather than a collection of words changes how it sounds.

The ‘plus’ in ‘AI-plus’ is doing more work than it appears to. It signals that the policy centre of gravity has shifted from building models to deploying them, from the labs that make the technology to the factories, hospitals and government offices that use it. That is a bet that the frontier of the AI economy is no longer in the training run but in the application, and it aligns with what the big Chinese model companies have been saying themselves for a year: that the domestic advantage is not the algorithm but the scale of the deployment surface, the sheer number of industries and institutions that can be wired into an intelligent system.

Reading it from Chengdu, where the technology industry is growing faster than the local universities can supply it, the report feels like a confirmation of something already visible. The policy question was never whether Chinese industry would adopt AI, which was settled years ago, but whether the adoption would be coherent and funded. The work report’s answer is that it will be both, and that the state intends to participate not only as a regulator but as a customer, which is a far more powerful position from which to shape an industry.

I am cautious about the parts of the plan that sound like magic. Every country that writes AI into its central budget discovers within a year that integration is slow, that most applications are mundane, and that the exciting demos are not the same as the boring work of changing how an organisation operates. The report cannot repeal that reality, and the ‘AI-plus’ label will be attached to a great deal of mediocre software by the time 2027 arrives. The waste that follows any industrial push is real, and it is worth naming.

But the direction of travel is harder to argue with. A country of this size wiring AI into its hospitals and factories is not a demonstration project; it is an economic event, whatever the short-term noise. The reports that matter in a command economy are the ones that say where the money goes, and this one says it plainly. The models will keep being built, and now they will be installed. Those are two different industries, and China has just told the market which one it is betting on.