01
Decision rights
Clarify what headquarters controls, what China can decide, and which choices require joint approval.
Choosing the Right AI Operating Model for China
Choosing an AI operating and governance model that balances global and China priorities
AI adoption in China is becoming a management-design challenge, not only a technology or compliance challenge. Multinational companies need to decide which decisions should remain with headquarters, which decisions should sit in China, and when China-specific tools, vendors, data infrastructure, and budgets are justified.
The central question for China executives is: what AI operating model should we propose to headquarters so China can move fast enough for local business needs while staying aligned with global governance?
The answer should not be a binary choice between global control and local autonomy. Most companies need a use-case-based portfolio approach: global standardization where consistency matters, localized delivery where China needs speed, localized technology where China data or compliance requires it, and China-led pilots where local innovation may later scale globally.
02
Technology route
Decide when global tools are sufficient and when China-local cloud, models, vendors, or data infrastructure are justified.
03
Scale pathway
Separate China-only solutions from China pilots that should be documented, modularized, and evaluated for broader rollout.
Contents
1. Why China AI operating models require explicit design
AI in China creates a specific management tension: global companies want consistency, control, and compliance, while China teams often need speed, local ecosystem fit, and user proximity. Without an explicit operating model, companies often drift into either over-centralization or unmanaged local experimentation.
1.1 China speed changes the cost of delay
Chinese competitors and local digital teams often move faster in customer engagement, e-commerce integration, factory automation, AI-assisted sales, and rapid product development. In several industrial and B2B settings, AI is already being used as a practical tool to reduce response time, automate documentation, improve sales conversion, and support local operations.
For multinationals, the key issue is whether the operating model allows China teams to act quickly enough when the business case is clear.
1.2 China use cases often require local ecosystem fit
Many China AI use cases depend on local data sources, local customer touchpoints, local cloud environments, local vendors, or local compliance requirements. This is especially true for customer-facing applications, local customer-engagement workflows, domestic e-commerce, local market intelligence, production data, and China-specific service tools.
Global tools can remain the default, but they cannot be the only path. The operating model must define when local technology is allowed and how it will be governed.
1.3 AI governance must be practical, not theoretical
Companies that over-govern every use case through the same process risk slowing adoption. Companies that under-govern local experimentation risk fragmentation, duplication, data leakage, and shadow AI. A practical model clarifies five decision areas.
| Decision area | What needs to be clear |
|---|---|
| Business ownership | Who owns the use case, adoption, and business value? |
| Budget | Who pays: HQ, China, business unit, function, or shared pool? |
| Technology stack | Which tools, cloud, models, data platforms, and development environments are allowed? |
| Delivery model | Who builds: global IT, China IT, business team, vendor, or mixed team? |
| Governance | Who approves risk, compliance, vendor selection, data use, and scale-up? |
2. Executive implications: the decision is not global versus local
The most common mistake is treating AI governance as a binary choice. A global-versus-China framing is too simple because AI use cases differ sharply by data source, user group, risk level, technology requirement, and scale potential.
Global control and China speed can coexist, but only if management distinguishes between low-risk pilots, operational workflow tools, customer-facing AI, and strategic enterprise systems. The target for many multinationals is a controlled hybrid: global principles, local execution, use-case-based technology decisions, and clear scale pathways.
| Design element | What should be clarified |
|---|---|
| Use-case categories | Which use cases are global, local, localized-tech, China-only, or pilot-to-scale? |
| Decision rights | What can China decide, and what must HQ approve? |
| Budget thresholds | What can be locally funded versus globally approved? |
| Technology boundaries | Which global and local tools are approved, restricted, or prohibited? |
| Data rules | Which data can remain local, move globally, or require review? |
| Vendor rules | When can local vendors be used? |
| Scale rules | When does a China pilot become a regional or global candidate? |
3. The five AI operating models
The five models differ according to who selects technology, who defines use cases, who controls the adoption timeline, and whether China-built solutions are intended only for China or may scale globally. This section provides a portfolio-level view; the compact playbook below provides the detailed decision logic.
M1
Global Standardization
Global defines the AI stack, use cases, approvals, and rollout timing.
M2
Global Technology, Local Adoption
Global chooses the platform; China owns local use-case sequencing and adoption.
M3
Localized Technology, Global Strategy
Global defines the strategic use case; China localizes the technical path.
M4
China-for-China Innovation
China selects technology, vendors, use cases, and timelines for China-specific outcomes.
M5
China Pilot-to-Global Scale
China pilots quickly; successful solutions are evaluated for broader rollout.
How the models map against China autonomy and global standardization
The matrix is intentionally simplified around two management dimensions. The x-axis is China autonomy: how much authority China has over technology selection, use-case priorities, budget, and rollout pace. The y-axis is global standardization: how strongly the model is governed by global technology standards, common use-case definitions, and global scale requirements.
Global standardization
China autonomy
Lower
Higher
Low
High
M1. Global Standardization
High global standardization; low China autonomy.
M2. Global Tech, Local Adoption
Global platform with China-led use-case sequencing.
M3. Localized Tech, Global Strategy
Global objective with China-local technical execution.
M4. China-for-China Innovation
High China autonomy for China-specific outcomes.
M5. China Pilot-to-Global Scale
High pilot autonomy with global scale review.
The matrix should not be read as a maturity ranking. Each model can be appropriate depending on business need, risk tolerance, technology constraints, and China capability.
| Model | China autonomy | Global control | Technology logic | Best-fit situation |
|---|---|---|---|---|
| M1. Global Standardization | Low | High | One global stack; global selects tools and timeline | Global productivity, shared services, global process tools |
| M2. Global Technology, Local Adoption | Medium | High | Global technical platform; local use-case ownership | Operational use cases where global tools work but local delivery needs speed |
| M3. Localized Technology, Global Strategy | Medium | Medium / High | Global strategy with localized China deployment when needed | Global use cases that require China-local technology execution |
| M4. China-for-China Innovation | High | Medium / Low | Local stack, local vendors, local budget for China-only use cases | China-specific speed, local data, local customers, local ecosystem |
| M5. China Pilot-to-Global Scale | High for pilot, medium for scale | Medium / High at scale stage | China-built pilot, modularized or standardized for scale | China as testbed for broader innovation |
4. Compact model playbook
The following model playbook removes repetition by focusing each model on the information executives need to decide: definition, strengths, limitations, decision logic, best-fit use cases, choose-when signals, and observed company pattern.
M1
Global Standardization
Use when consistency, security, vendor control, and enterprise-wide governance matter more than local speed.
+
Global Standardization
Use when consistency, security, vendor control, and enterprise-wide governance matter more than local speed.
This model centralizes technology selection, use-case definition, rollout timing, and approval authority at headquarters. China primarily adopts globally approved AI tools and follows global timelines. Local funding or local demand may influence prioritization, but it does not create independent China decision rights.
Strengths
Provides strong consistency, security, vendor control, and cost discipline. Works well when use cases are standardized across markets, global tools work reliably in China, and local speed is less important than enterprise-wide control.
Limitations
Slow when China has urgent local requirements, customer-facing needs, or local ecosystem dependencies. If global tools are unavailable, delayed, or poorly adapted to China workflows, local teams may defer adoption or find informal workarounds.
Choose when
- Use cases are internal, standardized, and similar across markets.
- Global tools are available and reliable in China.
- Risk appetite is low and local vendors are not yet trusted.
- China does not need major local workflow or tool customization.
| Dimension | Decision logic |
|---|---|
| Governance | Centralized at HQ or regional/global committees. China provides input but does not control the approval path. |
| China decision rights | Limited. China can propose needs, support localization, and drive adoption, but rarely independently selects technology or expands the AI portfolio. |
| Technology stack | Global-approved tools, global internal AI platforms, global cloud, and enterprise systems. Local hosting, local LLMs, or local vendors are exceptions. |
| Vendors | Local vendors are restricted or require global approval. Vendor selection is usually handled through global procurement, security, and architecture review. |
| Data | Global integration is preferred. China data is handled through global policy and enterprise data controls; local data residency is considered only where required. |
| Budget | Global platform budget or business unit budget may fund implementation, but local funding does not create local decision rights. |
| Delivery | Global platform, shared service, or enterprise IT teams lead delivery. China supports adoption, training, translation, and workflow localization. |
| Typical use cases | Internal productivity tools, global knowledge assistants, global reporting, shared services, standardized office AI, and globally consistent process tools. |
Observed company pattern
A large global beauty and consumer products company tightly restricts formal China-for-China AI development. Global governance controls tools, approvals, and rollout timing; China can fund use cases but still needs global approval. The benefit is strong consistency, security, and brand-level control; the trade-off is slower local innovation and reliance on agencies or informal external execution for some creative work.
M2
Global Technology, Local Adoption
Use when global technology works, but China needs more ownership over prioritization and adoption.
+
Global Technology, Local Adoption
Use when global technology works, but China needs more ownership over prioritization and adoption.
This model uses a global technical platform, such as an approved cloud, model layer, data platform, low-code tool, or enterprise AI service, while giving China or local business units responsibility for use-case selection, adoption sequencing, data preparation, configuration, and rollout. In short: global chooses the technology, while China chooses where and how to apply it locally.
Strengths
Accelerates local adoption without creating a separate China stack. It is effective when global tools work in China, but central AI teams cannot deliver every local operational use case. It also preserves enterprise controls while giving China more ownership of business value realization.
Limitations
Fails when local business teams lack AI delivery capability, tool access, or clear ownership. It can create a delivery gap: global provides infrastructure, while business teams are expected to execute without enough technical support.
Choose when
- Global AI tools work in China but adoption is too slow.
- China has many operational use cases that central teams will not deliver.
- Business units can fund or sponsor use cases.
- The main bottleneck is delivery ownership, not technology availability.
| Dimension | Decision logic |
|---|---|
| Governance | Global sets guardrails, risk rules, security standards, and approved platform boundaries. China manages use-case prioritization and adoption within those rules. |
| China decision rights | Medium. China can define local use cases, adoption sequence, configuration needs, and business rollout, but technology selection remains global-led. |
| Technology stack | Global AI platform, global cloud, enterprise data platform, low-code tools, agent builders, and approved model services. |
| Vendors | Local vendors are limited. They may support configuration, data preparation, or implementation, but are usually secondary to the global platform. |
| Data | Data usually remains in global or enterprise-governed platforms. China may onboard, clean, and connect local data if compliant with global controls. |
| Budget | The platform is typically globally funded. Individual use cases may be funded by business units, functions, or China/local budgets. |
| Delivery | China business, digital, IT, or BU teams configure tools, prepare data, build lightweight agents/workflows, and drive adoption. Global teams provide platform support and enablement. |
| Typical use cases | Local productivity agents, workflow automation, internal knowledge search, sales/marketing support tools, and business-specific analytics that can run on global technology. |
Observed company pattern
A large global chemicals and coatings business uses a global enterprise AI platform and data stack, while China and business teams identify many operational use cases, such as lightweight agents and workflow support. The benefit is reuse of compliant global technology without creating a separate local stack; the trade-off is a delivery ownership gap when central teams provide infrastructure but local teams lack enough technical support to execute.
M3
Localized Technology, Global Strategy
Use when global strategy is valid, but China requires local deployment to make it work.
+
Localized Technology, Global Strategy
Use when global strategy is valid, but China requires local deployment to make it work.
This model preserves global use-case priorities and adoption logic, but allows China-specific technology deployment when the global stack is not suitable. The business objective remains globally defined, while China may use local cloud, local models, local vendors, or a separate deployment environment to execute the same strategic use case in China.
Strengths
Preserves global strategic control while solving China-specific technology constraints such as data residency, latency, access, content expectations, or local platform integration. Useful when the global use case is valid but the global stack is not sufficient.
Limitations
Requires strong architecture and governance discipline. Without clear routing rules, it can become a patchwork of exceptions, duplicated technology choices, and unclear accountability between global and China teams.
Choose when
- Global strategy is clear, but China requires local deployment.
- Use cases involve China data, local external data, or public-facing interaction.
- Global tools are strong but insufficient for China-specific constraints.
- The company can manage two technology stacks with clear rules.
| Dimension | Decision logic |
|---|---|
| Governance | Global defines the strategic use case, business objective, or required capability. China defines the local technical path needed to execute it. |
| China decision rights | Medium. China has meaningful authority over local deployment architecture, cloud, model, integration, and vendor choices, but global still defines the use case and adoption requirement. |
| Technology stack | Global and China stacks coexist. The same business process or use case may run on a global stack outside China and a localized stack inside China. |
| Vendors | Both global and local vendors may be used. Local vendors are justified by compliance, latency, local data, local language, public-facing exposure, or local platform integration. |
| Data | Data residency and access rules drive routing. China data may remain in China even when the overall use case is globally defined. |
| Budget | Funding can be global, local, BU-based, or shared. China localization cost must be justified by compliance, performance, or business value. |
| Delivery | Usually joint delivery: global product/process owner defines the target capability; China technology and business teams localize and implement. |
| Typical use cases | Global use cases that require China-local execution, such as public-facing AI, China customer data use, local market intelligence, local data-provider integration, or localized AI agents. |
Observed company pattern
A global industrial adhesive manufacturer keeps global strategy as the anchor, but assesses local versus global technology by compliance, architecture, speed, talent, and coordination. China-local deployment is selected when China data, latency, or public-facing requirements make global execution impractical. The benefit is strategic alignment with China-fit execution; the trade-off is greater architecture complexity and the need for disciplined routing to avoid duplicated stacks.
M4
China-for-China Innovation
Use when the use case is clearly China-specific and speed or local ecosystem fit is critical.
+
China-for-China Innovation
Use when the use case is clearly China-specific and speed or local ecosystem fit is critical.
This model gives China significant autonomy to identify, fund, build, and iterate AI solutions for China-specific business needs. Headquarters may define broad guardrails, but China controls the portfolio logic, local vendor engagement, budget allocation, and execution path for use cases that are clearly local.
Strengths
Strong for speed and local relevance. It allows China teams to work close to users, data, vendors, and customer workflows. Particularly useful when local business teams need fast iteration and cannot wait for global platforms or global roadmaps.
Limitations
Can create fragmented architecture, duplicated tools, inconsistent data standards, and weak global visibility. Local vendor dependence can also create IP, cybersecurity, compliance, and maintainability risks.
Choose when
- Use cases are clearly China-specific and close to local users/customers.
- Speed directly affects competitiveness or customer response.
- China-based data, local vendors, or local customer platforms are central to the use case.
- China has the capability and budget to build under defined guardrails.
| Dimension | Decision logic |
|---|---|
| Governance | Global sets red lines, risk boundaries, and visibility requirements. China manages local governance for China-only use cases. |
| China decision rights | High. China selects technology, defines use cases, sets the adoption timeline, and manages the local AI portfolio for China-specific outcomes. |
| Technology stack | Local cloud, local models, local vendors, and China-specific integration architecture are allowed. Global enterprise tools may still be used where they fit. |
| Vendors | Local vendor ecosystem is actively used for development, implementation, compliance interpretation, and rapid iteration. |
| Data | China data typically stays in China. Architecture is optimized for local systems, local data sources, and China user experience. |
| Budget | Funding usually comes from China P&L, local BU/function budget, local digital budget, or a China innovation portfolio. |
| Delivery | China digital, IT, business, and vendor teams build and iterate close to local users and customers. |
| Typical use cases | China-only sales enablement, customer service, AI knowledge base, product recommendation, contract automation, factory AI, and local R&D/product development tools. |
Observed company pattern
A global B2B ingredients company runs an independent digital team for China-for-China use cases, with local budget flexibility and the ability to use global or local LLMs. The benefit is speed, strong business fit, and local vendor leverage; the trade-off is weaker global scalability unless China-built solutions are documented, governed, and aligned with global standards.
M5
China Pilot-to-Global Scale
Use when China can pilot faster and the solution has potential relevance beyond China.
+
China Pilot-to-Global Scale
Use when China can pilot faster and the solution has potential relevance beyond China.
This model uses China as a testbed for AI innovation. China leads pilot development because it has speed, cost, talent, vendor access, market pressure, or user proximity. Once a use case proves value, it is evaluated for rollout to APAC, Europe, or global operations.
Strengths
Captures China’s speed advantage while avoiding purely local isolation. It allows companies to prove value quickly before committing global resources and helps headquarters see China as a contributor to global AI transformation.
Limitations
Only works if global stakeholders are engaged early enough. If China builds something locally and headquarters later rejects the architecture, vendor, data model, or governance logic, the use case may not scale.
Choose when
- China can pilot faster than HQ or other regions.
- The use case has possible relevance beyond China.
- Headquarters is open to adopting China-developed solutions.
- The use case can be documented, modularized, and adapted for other markets.
| Dimension | Decision logic |
|---|---|
| Governance | China has authority to pilot quickly. Global or regional governance reviews whether the solution should scale beyond China. |
| China decision rights | High during pilot. Shared decision rights during scale-up, when architecture, vendor, data, and operating model must be validated for broader rollout. |
| Technology stack | May start with local technology for speed. Must be modularized, documented, or standardized if the solution is expected to scale. |
| Vendors | Local vendors may support pilot development. Scale-up may require enterprise vendor review, contracting, security review, or replacement. |
| Data | Pilot may rely on China data. Scale requires review of data portability, data model, process fit, and compliance across markets. |
| Budget | Pilot budget can be local or jointly funded. Scale funding usually requires regional/global business case approval. |
| Delivery | China leads pilot design, testing, and user validation. Global or regional teams support standardization, rollout planning, and broader adoption. |
| Typical use cases | Smart manufacturing, R&D acceleration, customer collaboration, process automation, predictive maintenance, product development support, and operational excellence. |
Observed company pattern
A global science and technology company established a technology capability platform to leverage local AI, robotics, and digital innovation for both China needs and global opportunities. The benefit is access to China speed, talent, and vendor ecosystem; the trade-off is that scale depends on strong business-case discipline, clear handover mechanisms, and global approval once a pilot moves beyond China.
5. Decision criteria by category
Managers should not start by asking how much autonomy China should have. They should assess a small set of criteria across technology architecture, use-case requirements, organizational capability, and business demand. The more checks a model receives, the stronger the fit. The result should guide discussion rather than replace management judgment.
Model-fit assessment
Use the criteria below to identify which operating model best fits a use case or portfolio.
This section keeps the original decision criteria but presents them as an executive assessment. Each table uses the same structure: Decision criterion | M1 | M2 | M3 | M4 | M5. A model with more checks is a stronger candidate for that category.
1. Classify
Is the use case global, local, localized-tech, China-only, or pilot-to-scale?
2. Test feasibility
Can the global stack work in China, or is local deployment required?
3. Assign rights
Clarify ownership for value, budget, vendor approval, data use, and scale-up.
4. Route governance
Separate low-risk pilots from customer-facing, data-sensitive, and enterprise-scale use cases.
Technology architecture
Use these criteria to determine whether global tools are sufficient, whether China-local technology is required, and whether the scope touches core enterprise systems.
| Decision criterion | M1 | M2 | M3 | M4 | M5 |
|---|---|---|---|---|---|
| Global AI tools work reliably in China for most needs. | ✓ | ✓ | |||
| Global strategy is defined centrally, but China technology must localize for compliance, latency, or local data access. | ✓ | ||||
| Important use cases depend on China data, local cloud, local vendors, or local customer platforms. | ✓ | ✓ | |||
| Core ERP, master data, global finance, or global compliance processes are in scope. | ✓ | ✓ |
Use-case requirements
Use these criteria to distinguish standardized global use cases, locally delivered use cases, China-only needs, and China pilots with broader scale potential.
| Decision criterion | M1 | M2 | M3 | M4 | M5 |
|---|---|---|---|---|---|
| Internal productivity or standardized global process use cases dominate. | ✓ | ||||
| Local workflows need faster adoption, but can run on global technology. | ✓ | ||||
| China-only use cases have urgent business needs and local budget. | ✓ | ||||
| China pilots have potential relevance for other regions. | ✓ |
Organization and capability
Use these criteria to test whether the company is ready for local autonomy, local vendor engagement, local pilot delivery, or stronger central control.
| Decision criterion | M1 | M2 | M3 | M4 | M5 |
|---|---|---|---|---|---|
| HQ needs strong control and is not ready to approve local vendors. | ✓ | ✓ | |||
| China has strong local digital, IT, vendor-management, or business translator capability. | ✓ | ✓ | |||
| China can fund and deliver pilots locally under defined guardrails. | ✓ | ✓ | |||
| Central teams will not deliver the volume of operational AI use cases China needs. | ✓ |
Business demand and scale ambition
Use these criteria to assess whether China needs local competitiveness, whether China speed can benefit global business, and whether similar AI solutions are being duplicated across regions.
| Decision criterion | M1 | M2 | M3 | M4 | M5 |
|---|---|---|---|---|---|
| China needs customer-facing or operational AI to respond to local competitors. | ✓ | ✓ | |||
| China speed, cost, talent, or user proximity can benefit global business if scaled. | ✓ | ||||
| Multiple regions are duplicating similar AI solutions. | ✓ | ✓ | |||
| China is mainly a rollout market, not a development hub. | ✓ | ✓ |
6. Conclusion
AI operating model design is becoming a strategic priority for multinational companies in China. The winners will not simply be the companies with the most permissive local policy or the strictest global controls. They will be the companies that can match each AI use case to the right decision rights, technology stack, budget mechanism, and governance path.
For executives, the practical question is not whether AI should be global or local. The better question is: which use cases should be globally standardized, which should use global technology with local adoption, which require localized technology to execute global strategy, which should be built for China only, and which China pilots could become global solutions?
A clear answer gives headquarters the visibility it needs and gives China the speed it requires.
