Ping An Insurance is expanding AI across financial services, healthcare and internal operations, reporting more than 300 billion tokens of average daily consumption and 1,500 PF of computing capacity. At the 2026 Apsara Conference, CTO Ray Wang outlined how the insurer is moving from AI experimentation toward large-scale, domain-specific applications designed to produce measurable operational value.

Ping An is positioning artificial intelligence as an operating layer across its financial and healthcare businesses rather than as a collection of standalone tools.

At the 2026 Apsara Conference, Ping An Chief Technology Officer Ray Wang described the group’s approach in a presentation titled “From Tokens to Value: Ping An’s Approach,” highlighting AI deployments across customer service, healthcare, software development, claims and fraud detection.

The strategy is built around Ping An’s “AI in ALL” approach and its broader integrated finance and health and senior-care businesses. The company says the objective is to move AI beyond answering questions and into workflows where systems can perform tasks, reduce processing time and support business decisions.

That focus reflects a broader challenge for financial institutions adopting generative and agentic AI. AI usage is spreading rapidly across enterprises, but scaling individual experiments into production systems remains difficult. McKinsey’s 2025 global AI research found that 88% of respondents said their organizations were using AI in at least one business function, while only 7% said AI had been fully scaled across their organizations.

Ping An’s latest figures illustrate what large-scale deployment can look like in practice.

Its AI-powered Express Service allows customers to interact with the company through natural-language prompts to complete tasks. Ping An says the service now supports more than 300 commonly used services spanning transactions, financing and claims, and has generated more than 100 million user interactions in its first four months.

The company reports a 92% end-to-end resolution rate and says average processing time has fallen by more than 50% compared with traditional service models.

The approach is closer to an AI-enabled service layer than a conventional chatbot. Instead of simply retrieving information, the system is designed to complete customer requests across Ping An’s existing financial-services workflows.

Healthcare represents another major deployment area.

Ping An’s AI Doctor is being developed to support multiple stages of the healthcare journey, including consultations, multidisciplinary diagnosis and treatment, and ongoing health management. The company says its AI assessments in multidisciplinary diagnosis and treatment for breast and gastric cancer were consistent with expert conclusions in 91% of cases, while its reported hallucination rate remained below 0.3%.

Ping An also says AI Doctor currently serves as the first point of contact for 25% of in-hospital services and expects annual service volume to exceed 100 million by the end of 2026.

The healthcare deployment highlights why domain-specific AI is becoming increasingly important in regulated industries. Financial services and healthcare cannot rely solely on general-purpose language models; systems need specialized data, workflows, validation processes and controls.

Ping An says it is addressing this through an integrated framework combining proprietary data, expert-defined standards, production quality controls and feedback from real-world business scenarios.

The infrastructure required to support those applications is also expanding rapidly.

According to Ping An, its computing capacity increased from 800 PF to 1,500 PF over the past year. At the same time, average daily token consumption increased from 30 billion at the end of 2025 to more than 300 billion.

The company says full-stack optimization across foundation models, inference engines and computing-resource scheduling increased token output per unit of computing power by 368% while reducing cost per token by 76%.

That economics question is becoming increasingly important as enterprises move AI from pilots into production. Token consumption can grow rapidly as organizations embed models into customer service, software development and operational workflows, making inference efficiency an increasingly important component of enterprise AI infrastructure.

Ping An also reports that AI-generated code now represents approximately 80% of its total code output, while around 20% of routine development tasks can be completed end to end by AI.

The company says AI service representatives handled 81% of total customer-service volume during the first half of the year. In insurance, quick claims represented 59% of life-insurance claims, while Ping An P&C generated RMB7.11 billion in claims savings through AI-powered fraud detection during the first half of 2026.

These figures illustrate a shift from AI as a productivity assistant toward AI as an embedded operational capability. They also mirror a broader enterprise trend identified by McKinsey: organizations are increasingly redesigning workflows around AI, but most have yet to capture significant enterprise-level financial impact.

For financial technology providers, the implication is that model performance alone is becoming a less useful measure of AI maturity. The harder problem is connecting models to proprietary data, business rules, applications and human oversight while maintaining acceptable economics.

Ping An’s strategy also places it within a growing ecosystem of AI infrastructure providers. Microsoft, Google, Amazon and NVIDIA are building the cloud, model and computing layers used by enterprises, while financial institutions increasingly develop specialized applications on top of those foundations.

The distinction between horizontal AI platforms and domain-specific intelligence is particularly relevant for banks and insurers. A general-purpose model can provide broad reasoning capabilities, but financial services applications require specialized knowledge of products, regulations, customer workflows and risk controls.

Ping An’s presentation suggests that the next phase of its AI program will focus on that specialization. Its stated objective is to combine increasingly efficient model infrastructure with domain knowledge and production feedback to create AI systems capable of delivering measurable value at scale.

The company’s reported deployment figures are its own claims and have not been independently audited in the announcement. Even so, the direction is notable for the fintech sector: AI investment is increasingly being measured not simply by model capabilities or token consumption, but by whether those technologies can operate reliably inside high-volume financial and healthcare workflows.

Market Landscape

Enterprise AI is moving from experimentation toward broader operational deployment, but scaling remains a challenge. McKinsey’s 2025 research found that 88% of surveyed organizations were using AI in at least one business function, while only 7% reported fully scaling AI across their organizations.

Financial services are among the industries where domain-specific deployment can have significant operational implications. McKinsey’s research shows organizations increasingly using AI across multiple functions and redesigning workflows around the technology, while governance and risk management remain important barriers to capturing enterprise value.

Ping An’s approach combines AI infrastructure, domain-specific intelligence, financial services automation and healthcare AI, reflecting a wider shift toward vertically integrated enterprise AI systems.

Top Insights

• Ping An reports more than 300 billion tokens of average daily consumption after increasing computing capacity from 800 PF to 1,500 PF.
• Its AI-powered Express Service supports more than 300 services and has recorded over 100 million customer interactions since launch.
• AI Doctor is being deployed across consultations, diagnosis, treatment and health management, including multidisciplinary cancer-care applications.
• Ping An says AI-generated code represents approximately 80% of its total code output, reflecting deeper AI integration into enterprise operations.
• McKinsey found that 88% of surveyed organizations used AI in at least one business function in 2025, while only 7% had fully scaled it.

Get in touch with our fintech expert