Infrastructure, openness and entrepreneurial drive define China’s AI innovation

Infrastructure, openness and entrepreneurial drive define China’s AI innovation

China’s AI momentum stems from physical, digital and institutional infrastructure, open cross-border orientation, and platform-scaled entrepreneurship. During a study tour, delegates saw frameworks operationalising AI, privacy-preserving data pipes, indigenous tech stacks, and compliance-led innovation that scale inclusion, resilience and connectivity.

The success of artificial intelligence (AI) in China cannot be understood without recognising the depth of the country’s infrastructure — physical, digital and institutional — and the entrepreneurial energy that has harnessed it. Delegates travelling from Shanghai to Hangzhou, Shenzhen and Dongguan experienced how high-speed rail networks and world-class airports connect China’s major financial and technology hubs. These systems are not just conveniences for visitors, but part of a deliberate strategy to make the country more open and mobile. China wants the world to come in, but equally wants its own people to travel out, learn from global best practices, and return with new knowledge. This openness underpins the policy essay on cross-border finance, which seeks to strengthen international connectivity by enabling the free flow of ideas, capital and expertise alongside goods and services.

Equally important is the digital and data infrastructure that has been built in parallel. Near-universal mobile internet penetration, ubiquitous quick response (QR) code payment adoption and the scale of domestic data centres provide the foundation for every AI application observed during the tour. But China’s approach goes further, creating integrated platforms for business, financial, trade and taxation data, linked to a national digital identity system. Together, these rails provide unprecedented volumes of validated data for risk management, compliance and financial inclusion. Institutions such as MYbank use satellite imagery to extend credit to farmers, while WeBank leverages blockchain to incorporate third-party data into lending decisions. These capabilities directly advance the essays on technology and digitalisation and inclusive finance, demonstrating how infrastructure is channelled into both innovation and access.

Infrastructure alone, however, does not explain the pace of adoption. China’s innovation story is equally one of entrepreneurial drive and scaling capacity. Leaders such as Jack Ma, Pony Ma and Robin Li turned start-ups into platforms that reshaped finance, commerce and culture. Today, a new wave of firms such as Lexin, PingPong and ZhongAn continue the tradition, building on gaps in consumer demand or regulatory opportunity. Many began as challengers to banks — in some cases threatening their core businesses, as Alipay’s Yu’e Bao did with deposit-taking — but increasingly they act as collaborators, even enablers. OneConnect’s role in providing AI-based compliance and risk tools to smaller banks illustrates this shift, aligning entrepreneurial energy with the regulatory focus, reflecting the five major essays on pension and elderly care finance by equipping institutions to serve vulnerable demographics, and green finance by embedding sustainability metrics into risk frameworks.

China’s research institutions complete the picture. Delegates saw how the elite C9 universities and their affiliated centres provide both basic science and applied innovation. At the Zhejiang University - Hangzhou Innovation Centre, for example, where founders of the six little dragons like DeepSeek and DeepRobotics are alumni, have learned to scale their AI and robotic start-ups into commercially viable companies with global reach. This bridging of academia, entrepreneurship and industry reflects the essay on technology and digitalisation, demonstrating how state-funded research is increasingly translated into market-ready solutions.

In sum, China’s infrastructure story is not just about trains, planes or data centres. It is about how these systems are integrated into a coherent strategy that combines openness, entrepreneurial dynamism and research capacity with clear policy direction. The five essays serve as both a compass and a filter, ensuring that infrastructure supports not only growth but also stability, inclusion, sustainability and global engagement.

The themes examined in the rest of this report are:

  • AI frameworks embed intelligence into processes: How institutions operationalise AI beyond pilots—into credit, fraud, anomaly detection, simulation and code generation—exemplified by OneConnect’s model development workflows, Lexin’s LexinGPT agents and WeBank’s 200+ AI use cases.
  • Data, privacy and trust as enablers of growth: Alternative data pipelines (business, financial, trade, taxation and satellite), blockchain-based third-party data sharing and privacy-preserving/federated learning in AML—seen at MYbank, WeBank and Ant Group.
  • Platforms drive scale through horizontal integration: Ecosystems that integrate content, commerce, payments and finance—Douyin’s AI-enhanced Doubao large language model powering engagement and monetisation; PingPong’s cross-border rails; ZhongAn’s embedded insurance.
  • Compliance becomes a field of innovation: Regtech as competitive capability—SPDB’s AI-enabled risk models; Ant’s end-to-end anti-money laundering (AML) stack with privacy-preserving computing; OneConnect’s codified compliance platforms for banks, especially smaller institutions.
  • Indigenous technology builds resilience: Cloud-native, microservices and distributed databases (WeBank’s Open Hive core system, Ant’s OceanBase) coupled with domestic AI hardware progress (e.g., Cambricon) and Huawei’s end-to-end stack to reduce external dependency.
  • Talent, culture and continuous improvement sustain innovation: Organisational stamina and learning cultures—Douyin’s youthful teams (average ~27 years), MYbank’s average age ~31 reflecting risk maturity, Huawei’s 19-level advancement ladder and emphasis on experience-driven improvement.
  • Inclusion and responsibility anchor growth: Financial access and accessibility at scale—WeBank’s services for people with disabilities; MYbank’s SME finance models; ZhongAn’s micro-protection—aligning commercial innovation with social objectives.

AI frameworks embed intelligence into processes

One of the strongest impressions from the study tour was how Chinese financial institutions have moved beyond pilot projects and isolated use cases of AI to build systematic frameworks that embed intelligence into the very core of their processes. Rather than treating AI as a tool to automate discrete tasks, these institutions are designing end-to-end workflows in which AI models generate, validate and improve outputs across the credit, compliance, risk and customer management cycles.

At OneConnect, Ping An’s financial technology subsidiary, delegates saw how AI has been applied to both operational and strategic areas of banking. Wu Lei, chief AI scientist, described how natural language processing and image recognition are used in customer service and claims processing, while micro-expression analysis supports lending interviews by detecting subtle behavioural cues. More striking was his example of AI-assisted code writing. Here, AI frameworks are not simply used to generate snippets of code but to design the entire process flow for developing stock or portfolio selection models. From requirement preparation, to framework generation, to process validation, the system automates a cycle that traditionally demanded weeks of human labour. This reflects the growing use of agentic AI — systems designed not just to provide outputs but to carry out tasks autonomously through planning, reasoning and multi-step execution.

A parallel was evident at Lexin, the Shenzhen-based digital retail and consumer finance platform. Delegates were introduced to LexinGPT, the firm’s proprietary large language model, and the suite of more than 50 AI agents it powers. Dusty Lui, head of AI, highlighted two in particular: the Riemann Agent, which continuously monitors core business indicators for anomalies and traces their root causes across correlated metrics; and the simulation agent, which models the impact of onboarding new funding partners by tracing transaction flows, funding distributions and customer segments before launch. These applications illustrate the shift towards agentic AI architectures, where agents can trigger other agents, retrieve relevant datasets, and recommend actions. Lexin’s systems employ principles similar to retrieval-augmented generation (RAG), drawing on vast stores of transactional and behavioural data to contextualise large language model outputs, ensuring that recommendations and simulations are grounded in real-world data. Lui also noted that Lexin’s code-generation systems produced hundreds of thousands of lines monthly in 2024, with cumulative AI-reviewed code reaching into the millions of lines, validated through simulation frameworks before deployment.

At WeBank, China’s first digital-only private bank, frameworks were applied both in scale and in breadth. The technology team explained how the bank has built more than 200 AI applications, embedded across marketing, fraud detection, risk assessment and customer service. These are managed within its cloud-native, modular architecture, where microservices are designed to interoperate through a common framework similar to a model context protocol (MCP). This approach allows different AI models and agents to exchange context and resources seamlessly, making them easier to orchestrate at scale. The design supports high-volume processing — up to 50,000 transactions per second — with AI-driven monitoring systems providing continuous oversight. WeBank also applied AI frameworks to accessibility, enabling over 6,000 hearing-impaired customers to use services through adaptive interfaces, a striking example of process-embedded intelligence that combines technical robustness with social impact.

Across institutions, what stood out to delegates was the process orientation of AI deployment. Rather than being bolted onto existing systems, AI is integrated into development lifecycles and business workflows, from requirement gathering to testing to deployment. By applying agentic AI, retrieval-augmented generation and model context protocols, these firms are operationalising intelligence in ways that allow it to scale across diverse functions, adapt to new use cases and reduce risk through validation at every stage. This shared approach across OneConnect, Lexin and WeBank demonstrates that a new model of AI operationalisation has emerged in China: one that embeds intelligence into frameworks, not just tasks.

Data, privacy and trust as enablers of growth

If AI is the engine of China’s digital finance model, then data is its fuel. Throughout the tour, delegates observed how Chinese institutions have moved far beyond conventional data sources to create comprehensive infrastructures of business, financial, trade and taxation data, combined with national digital identity systems. This breadth of validated information underpins the scale at which financial services can be delivered, while also serving as a foundation for compliance and trust.

At MYbank in Hangzhou, the digital bank under Ant Group, delegates were introduced to the “bird systems” that exemplify the creative use of alternative data. The Tomtit System uses satellite imagery to help farmers map their plots and estimate yields, allowing them to secure loans without traditional collateral. The Lark System digitises small enterprise documents for credit assessment, while the Goose System extends credit to suppliers and distributors often overlooked by larger banks. The Cuckoo System forecasts cashflows to match idle funds with wealth products. Each of these demonstrates how AI-driven models can draw on new forms of data to extend credit responsibly. Such systems reflect the operationalisation of RAG, where models retrieve structured information from multiple, heterogeneous sources to contextualise outputs — in this case, credit decisions grounded in verified data streams rather than model assumptions.

The focus on privacy and security was equally strong at Ant Group’s AML Centre. Experts showcased how privacy-preserving computing and federated learning enable multiple parties to share insights from sensitive datasets without exposing raw information. Suspicious activity monitoring models are trained on distributed datasets, with alerts escalated to a central case management system. This balances regulatory compliance with data security, while maintaining effectiveness at scale. Here, the principle of MCP is visible in practice — ensuring interoperability across different models, data sources and systems, so that information can flow securely while context is preserved.

WeBank in Shenzhen offered another perspective by demonstrating its blockchain-based integration of third-party data. By using distributed ledgers, WeBank validates external inputs such as merchant sales or logistics flows without requiring full disclosure of underlying records. The bank’s technology team emphasised that this infrastructure supports over 420 million customers, processing up to 50,000 transactions per second. AI models are embedded throughout, monitoring fraud patterns, assessing creditworthiness and personalising services. More than 200 applications work together in a framework that combines agentic AI with blockchain, enabling autonomous agents to request, validate and integrate external datasets while minimising risk. This design allows trust to be established even when datasets are fragmented or proprietary.

Delegates noted that this approach to data is not only about improving models but also about building systemic trust. By combining alternative data sources, privacy-preserving techniques, federated learning and blockchain verification, Chinese institutions have created an environment where AI can be scaled without undermining privacy or regulatory compliance. In practical terms, this allows for inclusion — bringing farmers, small enterprises and underserved populations into the financial system — while ensuring the resilience and stability demanded by regulators.

Across MYbank, Ant Group and WeBank, what became clear was that data governance and innovation are inseparable. Trust is not treated as an afterthought but as a design principle, embedded into every stage of the data pipeline. By deploying agentic AI, RAG and MCP-enabled architectures, these institutions ensure that data can be harnessed intelligently, securely and responsibly, making growth both scalable and sustainable.

Platforms drive scale through horizontal integration

A defining feature of China’s digital economy is the role of platforms as the organising model for both commerce and finance. Unlike in many other markets where services evolve vertically within sectors, Chinese institutions have built ecosystems that span multiple industries, drawing users into horizontal integration across content, commerce, payments and financial services. Delegates on the tour observed how this approach has enabled platforms to scale to hundreds of millions of users, extend into entire value chains and, increasingly, align with national development priorities.

At Douyin, ByteDance’s Chinese platform, the story of horizontal integration was vividly illustrated. What began as a short-video app has become one of the world’s largest commerce ecosystems, generating RMB 2.7 trillion (about $380 billion) in gross merchandise value (GMV) in 2023. AI recommendation engines sustain user engagement, while embedded payments and logistics turn content into transactions. Delegates were introduced to Doubao, ByteDance’s proprietary large language model, which powers search, recommendations, advertising and merchant tools within the platform. Doubao reflects how platforms now embed agentic AI within their ecosystems, enabling autonomous agents to respond to user intent, retrieve relevant content or product information using RAG, and coordinate across multiple services. This transforms Douyin from an entertainment app into a multi-layered infrastructure for culture, commerce and finance.

PingPong, the Hangzhou-based payments company, offered another lens on platform economics. Founded in 2015, it has grown into one of the largest global business-to-business (B2B) payment providers, processing more than $100 billion annually for over one million merchants across 200 markets. By serving as an infrastructure layer for merchants — with virtual accounts, foreign exchange (FX) risk management, cross-border collections and payouts — PingPong extends platform reach beyond consumer payments to global value chains. Its partnerships with Amazon, Shopee, Rakuten and other e-commerce leaders underline how platforms become embedded in international commerce, enabling horizontal integration across geographies. Delegates noted that PingPong’s strategy aligns closely with the national essay on cross-border finance, building the rails that support China’s merchants as they expand globally.

At ZhongAn Insurance, delegates saw how platforms extend into embedded finance. Established as China’s first online-only insurer, ZhongAn now serves more than 500 million customers, issuing billions of policies annually. Its products are distributed directly within e-commerce, mobility and entertainment platforms, providing micro-protection in real time — from travel insurance during ticket booking to health coverage embedded in ride-hailing apps. By embedding insurance in platforms where customers already transact, ZhongAn captures scale efficiently while reducing distribution costs. This embedded model reflects how ecosystems allow financial products to move horizontally across industries, rather than being restricted to financial verticals.

The strategic significance of platforms lies not only in scale but also in their ability to shape value chains and ecosystems. Platforms like Douyin create demand by curating cultural and commercial content; PingPong provides the financial rails for merchants to participate in global trade; ZhongAn embeds insurance at the point of need. Together, they demonstrate how horizontal integration enables institutions to orchestrate entire ecosystems, aligning commerce, finance and services into seamless user journeys.
For delegates, a key takeaway was that these platforms are not simply business models but instruments of policy alignment. Whether enabling cross-border finance, expanding inclusion through embedded services, or deploying domestic AI models like Doubao to reduce reliance on foreign technology, they reflect a fusion of entrepreneurial dynamism with national strategic priorities. The result is a model of platform economics that scales rapidly, integrates horizontally and sustains growth across sectors.

Compliance becomes a field of innovation

In many jurisdictions, regulation is often perceived as a drag on innovation. China’s experience shows a different trajectory — one where compliance and supervision become fields of innovation in themselves. Throughout the tour, delegates saw how financial institutions, technology companies and regulators have converged on the principle that responsible AI development requires not only technical breakthroughs but also robust compliance systems.

At SPDB, executives explained how compliance is embedded into their AI and data strategies. Li Lin, Vice President, noted that the five major policy essays on finance — covering areas from inclusive finance to green development and intelligent finance — guide SPDB’s adoption of AI across risk management, payments and customer engagement. SPDB has developed AI-driven systems that integrate taxation, customs and trade data into credit assessment, enabling more precise supervision of supply chain finance. Delegates highlighted how this reflects China’s broader approach: rather than loosening rules to allow experimentation, regulators create structured spaces where AI can be deployed safely and at scale.

The visit to Ant Group reinforced this point. Delegates toured its AML centre, where Ant has applied privacy-preserving computing and federated learning to monitor suspicious activity without compromising client confidentiality. By integrating alternative data sources with compliance requirements, Ant enables real-time monitoring across millions of transactions daily. Rather than treating compliance as a cost, Ant leverages it as a competitive differentiator, assuring regulators, customers and partners of its reliability.
OneConnect, the fintech arm of Ping An, extended this perspective by showcasing how compliance frameworks are themselves becoming AI-enabled service platforms. Delegates were shown how OneConnect’s anomaly detection systems and simulation frameworks are deployed to identify regulatory breaches and model compliance scenarios in advance. OneConnect’s AI also automates the drafting of regulatory reports and powers “AI code-writing” for portfolio and risk models, reducing manual workload while improving accuracy.

Taken together, these examples demonstrate how compliance has shifted from a reactive, check-the-box function into a field of continuous AI-driven innovation. By building privacy-enhancing technologies, federated learning systems and AI monitoring tools, Chinese institutions are developing compliance infrastructures that not only satisfy regulators but also create strategic advantages in risk management and customer trust.

Delegates reflected that this approach addresses a fundamental tension in global AI development: the balance between innovation and oversight. In China’s model, compliance is not a constraint to be minimised but an arena of applied AI development, where breakthroughs in privacy-preserving computing, anomaly detection and real-time monitoring are accelerating institutional resilience.
Indigenous technology builds resilience

One of the strongest themes throughout the study tour was the drive to build indigenous technology stacks to reduce reliance on foreign infrastructure and ensure resilience in the face of geopolitical and supply chain risks. Delegates repeatedly observed how Chinese financial institutions and technology companies are investing in cloud-native platforms, microservices, distributed databases, chips and hardware, creating an ecosystem that is both self-sustaining and globally competitive.

At WeBank, China’s first digital-only bank, executives explained how their operations are fully cloud-native, relying on microservices architecture and distributed database systems that the bank developed in-house. These systems enable real-time processing of high-volume, low-value transactions while maintaining low costs and high efficiency. Delegates noted that WeBank’s approach provides a benchmark for how financial services can operate at scale without depending on legacy information technology or overseas vendors.

Huawei offered perhaps the clearest symbol of this indigenous resilience. Its Dongguan campus showcased the company’s end-to-end capabilities — from servers and telecom equipment to chip design and cloud infrastructure. Despite global restrictions on access to advanced semiconductors, Huawei has developed alternatives, producing its own chipsets and hardware through domestic innovation. As Professor John Gong of the University of International Business and Economics remarked during the closing session, Huawei represents China’s ability to “stand on its own feet” in critical technology sectors.For delegates, Huawei’s case exemplified how state policy and entrepreneurial energy combine to secure strategic independence.

At Ant Group, the role of indigenous technology was highlighted through OceanBase, a distributed relational database originally developed to handle the massive transaction volumes of Singles’ Day sales. OceanBase now powers financial institutions and enterprises across China and abroad, demonstrating that home-grown solutions can rival — and in some cases outperform — established global products.

The tour also revealed how China’s AI ambitions extend beyond software to hardware innovation. Delegates discussed the rise of domestic AI chipmakers such as Huawei and Cambricon, whose processors are optimised for AI workloads and are increasingly deployed in financial and industrial applications. This reflects a coordinated effort to close the gap with global leaders such as Nvidia, ensuring that Chinese AI ecosystems are not vulnerable to external supply shocks.

These examples reinforced how resilience is not only about risk management but also about competitiveness. By building indigenous technology stacks — from cloud-native banking systems to AI chips and distributed databases — Chinese institutions are creating infrastructures that are difficult to replicate elsewhere. This combination of necessity and innovation has turned resilience into a strategic advantage, enabling scalability and cost efficiency while safeguarding sovereignty over critical technologies.

For delegates, the lesson was clear: China’s ability to embed resilience into its financial and technology systems is not accidental but the result of deliberate investment in indigenous innovation. This model shows how other markets grappling with geopolitical risks and supply chain dependence may seek to follow similar paths, though replicating the scale and coordination achieved in China remains a formidable challenge.

Talent, culture and continuous improvement sustain innovation

Behind China’s rapid progress in AI lies not only infrastructure and technology, but also the people and culture that sustain innovation. A consistent theme across the institutions visited during the tour was the deliberate focus on talent development, workforce composition and cultural practices that drive both resilience and adaptability.

At Douyin, the short-video and e-commerce giant, executives highlighted that the average age of employees is just 27. This demographic profile reflects an organisation built on youth, energy and stamina. Such a workforce brings agility, creativity and the willingness to experiment at speed, fuelling Douyin’s constant release of new features and services. The company leverages this dynamism to test and refine innovations in real time, matching the platform’s fast-moving consumer base.

By contrast, MYbank reported an average employee age of 31. While still young by global banking standards, this figure carries a different significance. MYbank emphasised that staff need sufficient maturity and experience to properly understand credit cycles, risk management and regulatory obligations. This balance of youthful digital fluency with risk-aware decision-making allows MYbank to innovate responsibly while managing the large-scale credit risks inherent in serving small and micro enterprises.

At Huawei, the emphasis was on continuous improvement and professional development. The company has established a rigorous 19-level system for staff development, which provides clear progression pathways and embeds a culture of advancement. Employees are encouraged to learn from practical experience and to translate those lessons into incremental but consistent improvements. This structure reflects Huawei’s long-standing culture of discipline and technical mastery, which underpins its ability to compete globally in advanced technologies such as semiconductors, fifth generation (5G) internet and cloud computing.

These institutional examples illustrate a broader cultural ethos in China that prizes hard work, experimentation and iterative progress. Long hours, high levels of commitment and an acceptance of fast cycles of trial and error are not seen as exceptions but as norms. The result is a workforce that is both resilient and adaptive, capable of sustaining the rapid pace of technological and business model innovation that characterises China’s digital economy.

Talent strategies also reinforce institutional goals. Douyin’s youthful workforce drives consumer engagement, MYbank’s mix of youth and maturity supports responsible financial innovation, and Huawei’s structured development framework ensures deep technical competence. Each case shows how human capital has been aligned to organisational mission, ensuring that innovation is not only achieved but also sustained over time.

The lesson for global peers is that talent and culture are as important as technology and capital in driving AI-enabled transformation. Investing in human capital, building structures for continuous improvement and aligning workforce composition to institutional objectives have been central to China’s AI innovation journey — and will remain so as its digital economy evolves.

Inclusion and responsibility anchor growth

China’s AI innovation journey is also marked by a growing recognition that technological progress must be anchored in inclusion, responsibility and broader social purpose. Across the institutions visited, participants observed how leading firms frame financial and technological innovation not simply as a path to profit, but as a means to extend access, empower disadvantaged groups and embed responsibility into their operating models.

WeBank stood out as a case study in inclusive finance. Beyond being China’s first digital-only bank, WeBank has built systems designed to enable those often excluded from mainstream financial services. One of the most striking examples is its work on accessibility for the handicapped, where AI-driven interfaces and customised digital tools lower barriers for visually impaired users and other groups with special needs. In addition, WeBank has leveraged blockchain technology for third-party data integration, enabling secure credit assessments for small and micro enterprises that may lack traditional collateral or credit histories. These initiatives highlight how inclusion and cutting-edge technology can reinforce one another.

At Ant Group, the focus on responsibility was evident in the deployment of privacy-preserving computing for AML. By using federated learning and other advanced techniques, Ant is able to collaborate with other financial institutions and regulators to detect suspicious activity without compromising individual or institutional privacy. This underscores a broader theme of compliance and risk management as fields of innovation in themselves, where responsibility is not a regulatory burden but an opportunity to apply new AI and data technologies.

In Ping An’s OneConnect, responsible AI practices are embedded in the development of AI frameworks for financial institutions. Executives emphasised that AI models must be transparent, explainable and governed by ethical standards if they are to be trusted by banks and regulators. This reflects the broader reality that in China, as elsewhere, adoption of AI in financial services will only scale if it is accompanied by strong governance frameworks.

Responsibility also takes the form of corporate social responsibility (CSR) and cultural commitments. Several institutions, including Huawei, described how their innovation programmes are not solely commercial but are aligned with national goals of sustainability, digital inclusion and technological self-reliance. Huawei in particular has framed its indigenous chip and hardware development not only as a business necessity but also as a contribution to national resilience and global digital equity.

The delegation also noted that inclusion is strategic, not peripheral. Whether it is expanding credit access to underserved businesses through MYbank’s AI-powered lending systems, enabling disabled users at WeBank, or embedding privacy and compliance by design at Ant Group, inclusion and responsibility are being treated as competitive advantages. They allow institutions to scale sustainably while building trust with regulators, customers and society at large.

The lesson for global peers is that in China’s digital finance and AI landscape, responsibility is not an afterthought. It is increasingly seen as an essential enabler of growth, ensuring that innovation delivers broad-based benefits and sustains legitimacy in the eyes of stakeholders.

Comments (0)
Cancel