공시 • May 16
Yiren Digital Ltd. Rolls Out Enterprise AI Architecture and Advances Transition to AI-Native Multi-Industry Operating Model
Yiren Digital Ltd. announced that AI is increasingly executing core workflows across its credit and insurance businesses. This shift is improving operating efficiency and unit economics, particularly in customer acquisition and servicing processes, and marks its transition toward an AI-native, multi-industry operating model. This evolution from AI-assisted operations to agent-driven execution is enabled by Yiren Digital's proprietary AI architecture, which is now deployed across its credit and insurance operations and extending into selected AI-native verticals beyond financial services. The architecture spans a foundation model, a multi-agent platform, an agent execution layer, an enterprise AgentOS, and a workspace-level employee co-pilot. Under its All-in-AI strategy, the Company is advancing its transition from a technology-enabled financial services platform toward an AI-native, multi-industry operating company. The strategy builds on over 5 years of the Company's purpose-built AI infrastructure and R&D investments, demonstrated operating impact across the Company's credit and insurance businesses, and an AI-executed internet insurance line that has emerged as a second growth engine. Following the 2025 launch of its proprietary multi-agent platform, MagiCube, the Company's 2026 rollout of MagiCube 2.0 marks a pivotal operational shift. AI agents now function as autonomous executors rather than mere task assistants, delivering end-to-end outcomes across the enterprise lifecycle. Key milestones underpinning this architectural evolution include: Zhiyu - the Company's proprietary large language model (LLM), received regulatory filing approval in April 2025 and serves as the foundational intelligence layer; MagiCube - the Company's multi-agent integration platform, launched in October 2025 as the connective infrastructure for large-scale agent deployment across sales, risk, capital, compliance, and customer service; MagiCube 2.0 - introduced in 2026, restructures the platform around two specialized layers: XuanJi, for human-and-organization workflow execution, and ZhiNao, the enterprise-wide AgentOS for LLM coordination and multi-agent orchestration. XuanJi is the Company's AI-driven workflow execution layer, delivering business outcomes directly across high-volume, repetitive processes including outbound customer service, telesales, insurance proposal generation, lending operations, and post-sale engagement. By moving these functions from human-assisted tasks to agent-delivered results, XuanJi structurally improves cost-to-serve, response speed, and consistency. ZhiNao is the Company's centralized orchestration and governance layer, an AgentOS that coordinates LLMs and agents, routes workflows, manages permissions, integrates enterprise knowledge, and enforces AI governance across the organization. ZhiNao serves as the single coordination hub for every agent the Company deploys, internal or third-party, and scales agent-driven operations across knowledge-intensive functions such as advisory services, strategic analysis, and management decision support. Workspace AI is the Company's enterprise-wide operating platform connecting employees, AI agents, knowledge systems, and intelligent workflows in a unified execution environment. Its employee-facing interface, AI Buddy, provides every colleague with a personalized co-pilot routed through ZhiNao, a single interface giving frontline staff and management direct access to enterprise data, agentic workflows, and approved third-party AI tools. All agent-driven workflows within the platform operate under a centralized governance framework, ensuring compliance, auditability and risk control across regulated business lines. In fiscal year 2025, integrating proprietary AI agents and LLMs generated meaningful cost efficiencies, and the continued upgrade of the Company's enterprise AI architecture is expected to further accelerate these benefits. Most notably, AI-powered precision marketing has reduced sales & marketing and customer service expenses as a percentage of total revenue. This migration toward an agent-led execution model has fundamentally bolstered unit economics and positioned the Company for sustainable, profitable growth. By repurposing the core AI architecture engineered for credit operations, the internet insurance distribution business achieved rapid, consecutive-quarter premium growth, becoming an increasingly meaningful revenue contributor. Capitalizing on this momentum, the Company is advancing an autonomous AI insurance agent assistant, currently in alpha-stage for internal testing. The platform is designed to automate standardized advisory tasks across the insurance customer journey. It will structurally lower customer acquisition costs while freeing up human agents' time to focus more on higher-value client relationship development activities. Yiren Digital is expanding its All-in-AI strategy through strategic minority investments in high-growth AI startups, further extending its ecosystem. The Company is leveraging its proprietary AI infrastructure and operational experience gained from serving over 14 million cumulative credit customers and more than 2 million insurance customers to support ecosystem partners and emerging AI-native initiatives. By focusing on AI-native verticals where results and benefits drive value, such as AI-enabled learning, interactive entertainment, and autonomous services, this capital-light approach enables the Company to capture structural AI growth while maintaining disciplined capital allocation and limiting execution risk. By unifying intelligence, decision-making, and execution across its proprietary stack, from the Zhiyu foundation model to the ZhiNao AgentOS, Yiren Digital has built a highly scalable architecture. Paired with disciplined capital allocation and a stabilizing credit environment in 2026, this cohesive system is positioned to drive renewed revenue and margin momentum in the periods ahead, supported by continued deployment of agent-driven workflows across its core businesses.
