The day began at the Ping An International Financial Centre, the tallest building in Shenzhen, where delegates toured the group’s artificial intelligence (AI)-powered showroom. Exhibits demonstrated applications across banking, insurance and healthcare, reflecting Ping An’s transition to a “finance + healthcare” model. With more than 247 million retail customers, the group has embedded AI into online consultations, elder care and mental health diagnostics. Delegates observed Ping An’s micro-expression analysis tool, designed to support loan interviews by detecting subtle emotional cues that may indicate higher credit risk. The system captures facial movements via cameras or smartphones, builds a benchmark profile and compares it with applicants’ responses to financial questions. Higher risk cases are sent for human review. Other demonstrations included AI-powered auto damage assessment for claims and intelligent customer service robots, which Ping An said now automate 99% of its customer-facing operations. The briefing then turned to OneConnect, Ping An’s financial technology subsidiary. Matthew Chen, chief executive officer of OneConnect International, with Rena Wang, head of product, and Wu Lei, chief AI scientist, outlined AI-powered banking solutions such as digital know your customer (KYC), risk management and customer engagement platforms, and noted that OneConnect is exporting these technologies to Southeast Asia and Africa. Wu described natural language processing for call centres, image recognition for claims and micro-expression analysis in lending; he said baseline expressions are benchmarked and unusual results sent for review. He also presented a use case of AI for code writing, where the system generates frameworks and process flows for stock or portfolio selection models. Wang added that vehicle identification and image recognition are used to prevent duplicate or fabricated claims. On customer interaction, Wang and Wu said that more than 80% of calls are managed by AI, combining rule-based processing with large language models (LLMs). Chen noted that international projects are underway and require localisation for compliance in overseas markets. Delegates next visited Lexin, a Nasdaq-listed consumer finance and digital retail platform founded in 2013. By 2025, Lexin had grown to serve 236 million users, 20 million merchant partners and nearly 200 financial institutions, with loan origination exceeding RMB 200 billion ($27 billion). Dusty Lui, head of AI, introduced LexinGPT, the firm’s proprietary large language model that underpins more than 50 AI agent applications across risk management, customer service, marketing and code generation. Delegates were shown the Riemann Agent, an anomaly detection system that monitors core business indicators and traces root causes across correlated metrics. Another system simulates the impact of onboarding new funding partners before launch, modelling transaction flows, funding distributions and customer segments to reduce operational risk and speed decision-making. Lui said Lexin’s customer data platform (CDP) integrates behavioural, transactional and partner data to support real-time segmentation by value, risk and life cycle stage for personalised marketing and targeted credit. He also highlighted AI for code generation and process automation, noting monthly output in the hundreds of thousands of lines and cumulative AI-reviewed code in 2024 in the millions of lines, with agents embedded from requirement preparation through coding, testing and deployment and validated by simulation frameworks before implementation. Discussions also covered international expansion to Mexico and Indonesia and potential partnerships to extend multilingual capabilities. The final institutional visit was to WeBank’s new headquarters, opened to underline its role as China’s first digital-only private bank. Founded in 2014 with Tencent’s backing, WeBank has grown to more than 420 million customers, supported by one of the largest transaction-processing infrastructures globally. The technology leadership team described an in-house, cloud-native core system capable of 50,000 transactions per second, built on the bank’s own database and modular microservices, and showed a credit journey in WeChat that completes in under five seconds. Flagship products include Weilidai for unsecured consumer lending in WeChat, Weiyedai for micro and small enterprises, and Weichedai for auto loans serving nearly three million customers. They said around 40% of loans now originate from the WeChat and QQ channels. AI is applied across marketing, fraud detection and customer service, with more than 200 applications developed internally. Accessibility features enable more than 6,000 hearing-impaired customers to access services. The team added that WeBank has exported its technology stack through a wholly owned Hong Kong subsidiary since 2024, supplying core banking and regulatory technology solutions to institutions overseas. The day ended with dinner at Huawei’s Songshanhu Campus, which will host the delegation for the fifth and final day of the tour. Day four combined walk-throughs of flagship facilities with intensive technical briefings. At Ping An and OneConnect, delegates examined AI in customer service, claims processing, fraud prevention and code writing, alongside projects exporting technology abroad. At Lexin, the discussion focused on LexinGPT and its agents for anomaly detection, simulations, requirement preparation and large-scale code generation. At WeBank, the focus was on its cloud-native system, high-volume processing, small and medium enterprise (SME) lending models and AI-driven accessibility. Together, the visits showed how Shenzhen’s leading financial institutions are embedding AI into the foundations of their operations and exporting these models internationally.
