MIIT Issues "AI+Software" Action Plan, Pushing Software Firms Toward AI Application Service Providers
Category:NewsPublished on:2026-09-11Views:14976Comments:0#Policy#Software#AI+
In September 2026, the Ministry of Industry and Information Technology (MIIT) issued the "AI+Software" Special Action Plan (MIIT Xin Fa [2026] No. 209), injecting AI capability into the full chain of software R&D, delivery, and service.
Public-info highlights
- Goal: by 2028 cultivate high-level intelligent programming tools and platforms, promote adoption covering 20,000+ scale software firms, organize 100 software-firm intelligence retrofits, build 100 agent-software benchmark applications, incubate 5+ quality open-source projects; by 2030 key software achieves comprehensive intelligent upgrade.
- Path: improve the intelligent-compute cloud-service system, develop integrated compute-model-software-tool services, and use inclusive policies like compute vouchers to lower software firms' compute costs.
- Orientation: push software firms toward AI application service providers, and implement an AI SME startup-support plan.
Background: why point at "software-firm transformation"
In the past, software firms delivered by selling licenses, doing projects, and charging maintenance. When AI capability can be called directly, clients buy not "software features" but "the degree to which business problems are solved". This requires the deliverer to have business understanding, data processing, system integration, and continuous-operations capability; the delivery center of gravity shifts from writing code to running effect.
What it means to us
"Software firms toward application service providers" is a clear policy direction: delivery shifts from selling licenses to effect-based delivery and continuous operations; capabilityfocus shifts from writing code to "business understanding + data + integration + operations". This aligns with the position we set for ourselves — not building general LLMs, but building the layer that makes AI usable in real business.
Inclusive policies like compute vouchers also lower the trial-and-error cost for small teams to do AI applications. For regional teams, this means model selection can be more measured: validate scenario value with smaller compute first, then decide whether to increase investment.
What enterprises can do
- Sort out their software assets and assess which modules can first introduce intelligent capability (e.g., assisted form-filling, intelligent retrieval, report generation);
- Watch local compute-voucher and other inclusive-policy application conditions and processes;
- Include "intelligence retrofit" in retrofit-project planning, keeping good before-afterinvestment and effect contrast data.
Policy details are subject to the MIIT's official release; application conditions and amounts are subject to competent-authority notices.
Comments(0)
No comments yet — be the first to share
