What Is China's 'AI+' Initiative? A 2026 Guide for Thai Business

What "AI+" means, and how it differs from "investing in AI"

"AI+" is built as a grammatical formula: AI plus something else. AI is not treated as a standalone industry. The distinction matters. When the state writes "AI + manufacturing", it means existing factories must apply AI — not that new AI companies should be founded in the hope that factories will buy from them.

MERICS found that "AI" is the single most frequent term across the entire plan document, which signals its weight more clearly than any press statement.

AI + what	What the plan intends	What Thai trading partners will see
Scientific innovation	Use AI to accelerate research and experimentation	Shorter product cycles from competitors
Industrial development	Embed AI in production lines and supply chains	Another layer of unit-cost advantage for competitors
Culture	AI-produced content and media	Cheaper marketing costs behind imported goods
Public welfare	Healthcare, education, elderly care	Fast-growing market for AI-enabled health services
Social governance	Public services and regulatory work	Standards and approval processes change faster

Digital China — the fourth of eighteen sections

This plan elevates digital development to a full section of its own, rather than a subheading under industry. Three concrete elements stand out.

1) A national data-resource ledger (数据资源一本账) — a single account of what data the country holds, which agency holds it, and who may use it. The idea is to treat data as a resource that must be booked like any other asset, rather than something each agency keeps to itself.

2) National AI innovation hubs and applied pilot bases (应用中试基地) — the interesting term here is "applied pilot base", a place to prove AI works on real work before scaling, not a research laboratory.

3) Coordinating algorithms, computing power and data — the plan treats these three as one set, because missing any one of them makes the other two useless.

Three terms China uses to steer its factories

In manufacturing, the plan uses three short terms as its direction marker: 智改 (make it smart), 数转 (make it digital) and 网联 (connect it as a network), backed by smart manufacturing and industrial internet programmes.

The slogan’s word order is not the working order. In practice 数转 (digital) has to come before 网联 (networked), and both have to come before 智改 (smart) means anything. A factory that skips ahead and buys an AI system before its production data lives in one place will get results it cannot measure. This is the same structural problem we described in how to make every ERP system communicate.

The lane China chose — application, not models

The plan targets the commanding heights of AI application in industry, not ownership of the strongest model. That is a deliberately different playing field from the United States, and it explains why Chinese models such as DeepSeek emphasise cost per call over benchmark scores. For the wider context of that competition, see our roundup of the current model war.

For Thai organisations this resolves one common worry: you do not need to build your own model to benefit from AI. A country with vastly more resources than Thailand has chosen to compete at the application layer.

Compared with what Thailand is doing

Aspect	China	Thailand
Where the policy sits	Inside the national development plan, section 4 of 18	National AI Action Plan (2022–2027), a separate plan
Emphasis	Embedding AI in industry and government process	Raising productivity and workforce skills
Latest programme	National AI innovation hubs and applied pilot bases	TH-AI Passport, launched 28 August 2026 by the Office of the National Digital Economy and Society Commission with a THB 1,600 million budget from the DE Fund, targeting 5 million places, with over 1.3 million registered on launch day
How it is measured	Innovation indicators; R&D spending growing 7% a year	Share of working-age Thais using generative AI reached 12.4% in Q1 2026 (Microsoft Global AI Diffusion report)

The fair reading is that the two countries started from different points because they are different sizes. China has a manufacturing base large enough to push AI directly onto production lines. Thailand chose to start with people, which costs less and spreads faster. Giving citizens direct access to the best tools available globally, rather than trying to build everything domestically, is a sensible decision at Thailand's scale.

The limitation is equally direct: employee skill only converts into organisational productivity when the organisation has data for those employees to apply AI to.

What Thai businesses should do first

Step	What to do	Question it lets you answer
1	Build a data ledger — what data exists, where it sits, who owns it	What usable data do we actually have?
2	Consolidate cost, inventory and accounting into one system	What is the real unit cost of each product?
3	Pick one repetitive, measurable task and pilot AI there	How many hours a month does this save?
4	Train the people who do that task, not the whole company at once	Who benefited, and why?

Note that this sequence mirrors China's 数转 → 网联 → 智改, scaled down to a single organisation.

Which tasks are ready, and which are not

Type of work	Start here?	Reason
Summarising documents, drafting letters, translation	Strongly suitable	Errors are visible immediately, a human can verify, damage is contained
Searching across large internal document sets	Suitable	Clear time saving, measurable as search time per request
Reconciling documents with repeating formats	Suitable if data is in one system	Needs a correct reference figure to compare against, or the answer looks right but is wrong
Costing or closing the books	Not yet, as a decision-maker	Every line must be auditable under accounting and audit requirements
Approving items with legal or financial commitment	Not suitable	Requires a named accountable person and a verifiable approval trail

The working boundary: tasks where AI drafts and a person decides can start today. Tasks that require an audit trail and legal accountability belong to the system of record, with AI used only to prepare the inputs.

Data risks worth checking before you start

Data readiness comes before tooling

A pattern repeats in organisations starting to invest in AI: the tool is chosen before anyone can say where the organisation's data actually lives. A question like "how can AI help our accounting team" is often followed by the discovery that nobody can state the unit cost of a given product, because production, purchasing and payment records sit in separate files owned by separate departments. When the source figures disagree, the AI's answer inherits every one of those disagreements — and reads more confidently than the raw numbers did, which is the dangerous part.

Note that this is the same order China's plan follows: the data-resource ledger comes before AI application. Know what data exists, who holds it and how far it can be trusted — then lay the tooling on top. At organisational scale, the same principle reduces to four questions you can put to one workflow at a time.

One: does this number have an owner? Who confirms it is correct? With no name attached, nobody is accountable when it is wrong.

Two: can you pull it yourself? Or does someone have to build a report each time? This is what decides whether the work can actually be automated.

Three: can it be traced to a source document? A figure that cannot explain where it came from cannot be used to decide anything about money — whether a person or an AI produced it.

Four: when two systems disagree, which one wins? Organisations that cannot answer this tend to spend more time in reconciliation meetings than in decision-making.

Any question you cannot answer marks a workflow that is not ready for AI, and the fix is in the data rather than in a smarter tool. That work is unglamorous — making systems that hold fragments of the same record talk to each other (covered separately in how to make every ERP system communicate), and agreeing who owns which number.

It also pays back longer. Clean data works with every tool that arrives later, while a tool laid over unsettled data has to be rebuilt each time the tool changes.

Conclusion

China's AI+ is not a technology policy. It is a production policy that uses technology as the instrument. What Thai businesses can adopt immediately is the sequence: get data into one system, connect it end to end, then apply AI where results can be measured.

The next part examines China's technology self-reliance strategy and what it means for technology transfer to Thailand — read part three, or return to part one for the full picture of the plan.

China is not competing to own the best AI model. It is competing to deploy AI across the most real work — a contest any size of organisation can enter, provided the data is ready.

- Paitoon Butri · Network & Server Security Specialist, Grand Linux Solution Co., Ltd.

References

• Cyberspace Administration of China — A clear route for Digital China (17 March 2026)
• NPC of China — Draft Outline of the 15th Five-Year Plan, summary
• Xinhua — From industry to livelihoods: the 15th Five-Year Plan
• MERICS — Deciphering the 15th Five-Year Plan
• NSTDA — National Artificial Intelligence Action Plan for Thailand (2022–2027)
• The Bangkok Insight — TH-AI Passport launch (28 August 2026)
• THE STANDARD — TH-AI Passport launch, over 1.3 million registrations (28 August 2026)
• Thansettakij — Microsoft Global AI Diffusion: Thai AI usage at 12.4% (15 July 2026)