Key insights
- Yiren Digital reported a significant net loss in Q2 2026, a reversal from profit a year prior, as it undergoes a strategic transformation from a fintech lender to an AI-native platform. The company is shifting towards asset-light models, reducing balance sheet credit risk, and monetizing technology services. Despite operational scale, shares are trading near a 52-week low amid regulatory challenges in China's online lending sector.
Yiren Digital (NYSE:YRD) presented its September 2026 investor update on September 30, outlining an ambitious transformation from a traditional fintech lender to an AI-native, multi-industry operating platform. The presentation comes as the company navigates significant regulatory headwinds in China’s online lending sector while posting a second-quarter net loss of RMB 449.6 million—a sharp reversal from the RMB 357.5 million profit recorded a year earlier.
Despite the financial pressure, with shares trading near their 52-week low at $0.94 in premarket activity, the company’s presentation emphasized measurable operational improvements driven by artificial intelligence deployment and a deliberate pivot toward asset-light business models. The investor deck highlighted both the pain of transition and the potential of its AI-first strategy across credit, insurance, and new verticals.
The company’s core strategic shift centers on moving away from capital-intensive, balance-sheet lending toward a referral and technology-services model. As illustrated in the presentation’s business model transition framework, Yiren Digital is fundamentally restructuring how it generates revenue and manages risk.
The transformation involves three key changes: shifting credit risk increasingly to funding partners rather than retaining it on the balance sheet, monetizing technology and platform services independently of loan volumes, and expanding omnichannel insurance distribution as a "capital-light second engine." With 14+ million cumulative customers served in credit solutions and RMB 15.2 billion in loans facilitated during the first half of 2026, the company maintains significant operational scale even as it restructures.
The presentation outlined three strategic pillars driving this transformation: an asset-light credit model focused on premium repeat borrowers, an AI-native growth platform leveraging proprietary large language models, and disciplined capital allocation funded from existing cash resources. Each pillar represents a deliberate move away from the high-capital, high-risk lending model that dominated the company’s earlier years.
Perhaps the most compelling section of the presentation detailed tangible results from Yiren Digital’s AI deployment across operations. The company showcased its proprietary technology stack, including the Zhiyu LLM launched in April 2025, the MagiCube 1.0 multi-agent platform from October 2025, and the upgraded MagiCube 2.0 with XuanJi and ZhiNao AgentOS rolled out in 2026.
The AI architecture rollout demonstrates a clear evolution toward enterprise-grade AI governance and execution capability.
MagiCube 2.0 represents four major upgrades: centralized governance with enterprise-wide AI control, autonomous execution replacing task assistance, unified orchestration improving output consistency, and composable modules enabling reusability across functions. The company reports deploying over 10 reusable AI capabilities across six enterprise functions, significantly reducing the cost and time required for incremental AI deployment.
The operational proof points presented are particularly notable given the company’s financial challenges. In customer operations, the Qingniao customer-service system achieved a 98.7% answer rate, with text agent autonomous resolution increasing from 60% to 80%. The system processes approximately 1,500 hours of speech-to-text daily at 97.8% accuracy and covers over 2 million records per day with automated quality inspection. Complaint handling reached 100% resolution within 24 hours, while total complaint volume declined 35.97% year-over-year.
Operating efficiency improvements were equally striking. Human handling rates in asset-recovery operations dropped from 45.0% to 24.9%, while M1 tickets handled per collections staff increased 47%, from 358 to 525. The share of D1 workload handled by AI rose from 50% to 81%. The company’s LingShu content agent completes over 1,700 tasks daily, with average marketing copy generation time of just 0.6 seconds.
Risk control metrics demonstrated the fraud prevention value of AI deployment. The presentation reported RMB 165 million (US$23 million) in potential fraud losses avoided in 2025, with 10,300 fraudulent applicants intercepted and 14,500 fraud cases blocked. The proprietary fraud blacklist contains approximately 800 million records, with the DiTing system screening roughly 30,000 credentials daily.
Recognizing the risks inherent in AI deployment at scale, Yiren Digital presented a comprehensive four-tier defense system governing every agent across lending, insurance, and international markets. The framework provides what the company describes as "controllable invocation, manageable permissions, and auditable behavior."
The L1-L4 defense structure covers data security (PII detection, masking, input logging), model security (prompt constraints, output rules, validation), business security (compliance control, risk management, closed-loop response), and organizational governance (standards, norms, operational coordination). The system enables full-chain audit with end-to-end agent traceability, efficient closed-loop intervention, and 100% automated quality assurance of high-risk behaviors.
This governance emphasis appears particularly important given regulatory scrutiny of AI deployment in China’s financial services sector and the company’s stated ambition to expand AI applications beyond its core lending business.
While the presentation emphasized operational progress, the actual second-quarter financial results revealed the significant cost of transformation. Revenue declined 46% year-over-year to RMB 890 million, driven by a 29% sequential decline in loan facilitation volume to RMB 6.3 billion. The outstanding balance of performing loans fell 30% to RMB 15.1 billion.
The quarterly revenue mix illustrated the ongoing shift in business composition, with lending-driven revenue declining from 77% in Q2 2025 to 67% in Q2 2026, while technology-driven revenue increased from 14% to 27% and insurance brokerage grew from 4% to 10%.
However, credit quality metrics showed meaningful improvement. The presentation highlighted that repeat borrowers now account for 82% of total loan facilitation volume, up from 78% in the prior quarter and representing a strategic focus on established, lower-risk customers. First payment default rates (FPD30+) declined 10.7% quarter-over-quarter, while delinquency rates across all buckets improved significantly.
The asset recovery case study demonstrated AI’s impact on collections efficiency. Human handling rates dropped by nearly half, while tickets handled per AI agent increased substantially. The D30 recovery rate improved 10 percentage points to 88.5% between October 2025 and June 2026, even as the M1 cost rate declined to 3.24%. Concurrently, collector headcount was reduced by 53% while workload per collector increased 38%, demonstrating what the company calls "scalable operating leverage."
Capital allocation optimization through AI agents also showed measurable results. The YiQ agent for capital analysis and monitoring, combined with the ZhuQue bot for capital deployment, increased fund matching rates from 60% to 66% while reducing asset costs by 24 basis points. Processing speed increased fivefold, and the capital idle rate decreased from 0.76% to 0.53%, generating approximately RMB 54 million in additional cash flow.
The presentation positioned insurance brokerage as an increasingly important revenue diversification strategy. The segment showed strong growth momentum, with cumulative insurance clients reaching 2.71 million by Q2 2026, up from 1.41 million a year earlier. New policies issued exceeded 918,000, up 177% year-over-year, while gross written premiums reached RMB 838.9 million.
The company’s Yixiaobao platform represents its AI-native insurance sales infrastructure, designed to empower online-offline hybrid services. The system has launched knowledge bases for 42 insurance products across six categories, with dedicated databases covering medical insurance, annuity and whole life insurance, and other critical lines.
Yixiaobao integrates intelligent calculation functions, automated content generation, intelligent recommendation and interactive agent optimization, and channel connection with general data tools. The company expects to launch additional independent products in the second half of 2026 as it works toward full commercialization of its AI-enabled insurance solutions.
The national offline brokers network comprises over 500 agents and 30+ branches nationwide, partnering with 100+ insurance companies and offering 1,000+ insurance products. Meanwhile, the fast-growing online channel focuses on short-term insurance with innovative, online-adaptive products, with its contribution to total segment gross written premiums continuing to grow.
Beyond its core financial services, Yiren Digital outlined an expansion strategy into AI-native verticals, including AI entertainment/ACG, AI-assisted language learning, AI research productivity, and AI companionship and emotional wellness. The companionship platform already serves over 3 million users with annual recurring revenue exceeding US$10 million.
The investment structure employs a staged approach: taking minority stakes in high-potential AI application companies upfront based on strategic alliances, with options to purchase additional shares to gain majority stakes at pre-agreed prices. This approach aims to optimize capital allocation through optionality and upside potential while limiting initial capital commitments.
The company’s rationale for expanding into these verticals centers on capability transfer rather than just capital deployment. Yiren Digital argues that its enterprise AI stack—including proprietary LLMs, the MagiCube 2.0 multi-agent platform, and governance frameworks—represents reusable foundational capabi