Jiayin (JFIN) AI Customer Service Cuts Inference Cost 90%

On August 28, 2026, Jiayin Group Inc. (JFIN) said its AI customer service now provides end-to-end coverage in key operating scenarios and completely replaces human agents in select functions, but the company did not identify the tasks, workload, staffing effect, or financial contribution.[1]

Jiayin Group Inc., also known as Jiayin Technology, is a fintech company that connects borrowers with financial institutions, with customer service supporting customer communications and service delivery. Its AI Customer Service Assistance is used by service agents and operating teams for transcription, summaries, intent recognition, and knowledge prompts before, during, and after conversations. It also handles some service tasks directly, making it a core operating application.

From agent assistance to workflow automation

AI Customer Service Assistance has gradually expanded from an agent tool into a customer-service workflow application. On November 22, 2023, its first-phase CRM functions included real-time transcription and tags for customer intent and agent behavior.[5] In 2025, after Jiayin replaced some commercial large language models with self-optimized models, the cost of conversation summaries fell by about 80% year over year. The customer-service AI product matrix later expanded into training, real-time conversation support, and post-event analysis while reaching all agents.[4][3] In 2026, proprietary models were fully deployed across customer-service operations, and the August 28 disclosure advanced the same application to end-to-end coverage and autonomous handling in select functions.[2][1]

Accuracy improved as technical unit cost fell

The clearest current improvements are in recognition performance and technical unit cost. Jiayin reported in June 2026 that customer-intent recognition accuracy increased from 78% to 93% while model inference cost fell 90%.[2] The first metric indicates better recognition of customer needs, while the second indicates a lower technical cost to run the model for the relevant tasks. Jiayin did not disclose the test sample, measurement period, absolute cost, or service-quality outcome, so these figures cannot be converted directly into total customer-service cost or an overall efficiency gain.

The financial link remains unproven

Lower inference cost could reduce the cost of each AI-handled task, while replacing people in select functions could reduce related staffing or outsourcing needs and ultimately affect customer-service operating expenses and operating margin. However, Jiayin did not disclose task volume, staffing or outsourcing changes, absolute savings, or the relevant financial-statement expense category. Business volume and technology investment could also affect expenses. The 90% decline in inference cost therefore does not mean total customer-service cost fell by the same percentage or that margins have already improved.

What can be confirmed is that Jiayin's customer-service AI has expanded from an agent tool into a workflow-level application, with better recognition, lower technical cost, and limited human replacement. Verifying a company-level profit contribution still requires data connecting customer-service task volume, staffing or outsourcing expense, unit service cost, and operating expenses.

Application assessment

• AI Customer Service Assistance | Business position: core operations | Application stage: limited production | Scope: single business unit | Value type: cost reduction

Sources

[1] Drillr · Jiayin Group Inc. (JFIN) · 2026-08-28 · Earnings call

Original: End-to-end AI coverage has now been implemented in key operational scenarios such as customer service and loan application intake, completely replacing human agents in select functions.

[2] Drillr · Jiayin Group Inc. (JFIN) · 2026-06-23 · Earnings call

Original: Intent recognition accuracy improved from 78% to 93%, significantly enhancing service efficiency while reducing model inference costs by 90%.

[3] Drillr · Jiayin Group Inc. (JFIN) · 2025-11-25 · Earnings call

Original: With 100% agent coverage and over 90% accuracy, it significantly boosted staff efficiency and service quality.

[4] Drillr · Jiayin Group Inc. (JFIN) · 2025-08-20 · Earnings call

Original: For example, in agent assistance scenarios, the cost of AI-generated conversation summaries decreased by approximately 80% year-on-year.

[5] Drillr · Jiayin Group Inc. (JFIN) · 2023-11-22 · Earnings call

Original: Secondly, our Phase One auxiliary functions for customer service agents have been launched in the customer service CRM system, including real-time text transcription, customer intention tagging and agent behavior tagging.