China's "AI+" 5-year strategy: 90% economic integration goal by 2030, 4 things companies should prepare now

Publication date: 2026-03-06 | Category: AI News

1) Problem definition

The key issue of today's AI Times is that the Chinese government proposed a goal of integrating AI into 90% of the economy by 2030in the new five-year plan. The working readers of this news are the strategy, product, and data organizations of Korean companies directly or indirectly connected to the Chinese market and supply chain. The problem is not identifying technology trends, but whether our action plan is keeping up with when regulation, procurement, and local partner structures are reorganized around AI. This article is not a news summary, but rather an actionable review framework within 90 days. However, decision-making in sensitive areas for national defense and surveillance purposes is excluded from the scope of application of this article.

2) Evidence and comparison

According to a report by AI Times, AI was mentioned more than 50 times in China's new plan document (page 141) and presented as a core national strategy. Looking at the SCMP and Chinese government public data together, the direction is not simple model performance competition, but Increasing industry-wide penetration

• Cost: Pilot priority over outright conversion lowers the cost of failure.
• Time: KPI must be started from units that can be verified within 90 days (e.g. defect detection, demand forecasting).
• Accuracy: Data quality and field labeling system, rather than the model itself, determine success or failure.
• Difficulty: Legal/security/procurement process alignment is a bigger bottleneck than technology.

3) Step-by-step execution method

1. D+1~7: Exposure diagnosis — Quantify dependence on Chinese sales, supply chain, and partners and select three tasks vulnerable to AI policy changes.
2. D+8~21: Priority pilot design — Choose one of defect detection, demand forecasting, and customer service and define the target KPI (accuracy, processing time, cost reduction).
3. D+22~45: Data/governance alignment — Confirm data export path, access rights, and log retention policy with the legal and security team.
4. D+46~70: A/B operation — Measures false positives/omissions/processing time differences by operating existing and AI methods in parallel.
5. D+71~90: Investment decision making — Decide to expand, hold, or stop depending on whether KPIs are achieved, and fix the budget for the next quarter.

#90 Day Pilot Go/No-Go Pseudocode
if accuracy_gain >= 0.12 and cycle_time_reduction >= 0.20 and compliance_issue == 0:
    decision = "scale"
elif compliance_issue > 0:
    decision = "hold_and_fix_governance"
else:
    decision = "narrow_scope_and_retest"

4) Mistakes/Pitfalls

1. Patch: Immediately misunderstand national announcements as on-site performance
   Prevention: Inside the company Separate verification with KPI
   Recovery: If performance is not achieved every 4 weeks, reduce scope and re-experiment
2. Pitfall: Focusing only on model selection and neglecting data quality
   Prevention: Label error rate/missing rate first Check
   Recover: Retrain after data cleansing sprint
3. Pitfall: Pilot expansion without legal/security agreement
   Prevention: Export/Access Control Checklist Dictionary Approve
   Recover: Immediately stop high-risk path and then override approval process

5) Execution Checklist

• Have you defined the proportion of sales/supply chain connected to China as a number?
• Have you documented your 90-day pilot KPI (accuracy, time, cost) targets?
• Did you block the data path without legal/security approval?
• Have you agreed on the failure criteria (abortion conditions) and retry criteria?
• Do you track the top 3 causes of false positives/omissions in your weekly review?

Definition of Done: In 90-day pilot, accuracy improved by more than 12%, processing time reduced by more than 20%, expansion to the next step when two or more out of zero compliance issues are achieved.

6) Reference

• China, "Integrate AI into 90% of economy by 2030" through '5-year plan' (AI Times, 2026-03-06)
• China's five-year plan emphasises orderly AI development (SCMP, 2026-03-05)
• China unveils AI Plus action plan milestones (State Council, 2025-08-27, confirmation date 2026-03-06)
• China's 2026 growth target and policy priorities (Reuters, 2026-03-05, confirmation date 2026-03-06)

7) Author Viewpoint

I view this announcement as a signal of expansion of industrial policy-based AI, including procurement, governance, and operation KPI, rather than the general argument that “China is good at AI.” From the perspective of Korean companies, we recommend the strategy of quickly running 1-2 90-day pilots to resolve data and compliance bottlenecks first rather than making a full investment. Conversely, the approach of “introducing one large model will automatically produce results” is not recommended. The winner in 2026 will be determined by the density of execution systems rather than model performance.