Your lead entry number has been confirmed. You will now be joined to the conference. Excuse me. This is a conference operator. May I have your name, please? Hello. This is an operator. We are speaking privately. May I have your name, please? Your line is open. Hello? Would you please respond? Casey Lee from Waterdrop Investor Relations.

If my player is to welcome everyone to Waterdrop's fourth quarter and official year 2024 Earnings Conference Call. All participants are in listen-only mode in our English line. As a reminder, today's conference call is being recorded. Please note that discussion today will contain forward-looking statements made under the Safe Harbor provision of U.S. Private Securities and the Litigation Reform Act of 1995. Forward-looking statements are subject to risks and uncertainties and may cause actual results to differ materially from our current expectations. Potential risks and uncertainties include, but are not limited to, those outlined in our public filings with the SEC. The company does not undertake any obligation to update any forward-looking statement except as required by law. Also, this call includes discussions of certain non-GAAP measures. Please refer to our earnings release for our reconciliation between non-GAAP and GAAP. Joining us today on the call are Mr. Peng Shen, our Founder, Chairman, and CEO; Mr. Ran Wei, Director and GM of Insurance Business; Mrs. Head of Finance Department; and Mrs. Jasmine Li, our Secretary. We will be happy to take some of the questions in the main line at the end of the conference call. Hello, everyone. This is Peng Shen. In 2024, Waterdrop delivered solid performance across all business units, driven by our integrated neuro ecosystem and the sustainable AI investments. By fully upgrading products and services, we have established new benchmarks. In Q4, our total revenue was about $687 million. The annual revenue reached $2.77 billion, up by 5.4% year-over-year. In terms of profit, the net profit attributable to ordinary shareholders in this quarter was over $99.6 million, and the annual profit results have doubled, achieving our financial guidance to the market. The company has remained profitable for twelve consecutive quarters. I am also pleased to…

It's fully in one capability system.

Integrating scenarios, service, data, and knowledge, ensuring the rapid implementation of Waterdrop's AI solutions. While general-purpose AI struggles with hallucination, our vertical models can already deliver definite answers to our users. Overall, the combination of multiple scenarios and online data has offered irreplaceable strategic value in insurance LLM. Next, I will demonstrate how our AI has delivered measurable results. Over the past two years, we have actively adopted cutting-edge fundamental technologies while developing AI applications tailored to our business needs. I want to share some examples. Firstly, our insurance AI agent has end-to-end self-decision-making capability. It integrates the ability of professional and precise decision execution, leveraging multi-agent collaboration technology. In the medical insurance scenario, it enables fully automatic decision-making up to fifteen-minute conversations. For example, a single call can automatically complete processes like understanding user needs, health risk modeling, and policy interpretation. Our AI agent can support modular agent unit collaboration in real-time and flexible transition across health insurance, auto insurance, and other scenarios. The system rates over 7,000 products, and it can respond within two seconds. Our AI agent can dynamically optimize LLM algorithms by learning from the operation feedback and sales outcomes through over 100,000 daily interactions.

And secondly, we also continue to

optimize our model through two main approaches. The first one is knowledge distillation. This method uses technology like FFT and chain of thought reasoning, impacting millions of insurance-related knowledge to the model, making it comparable to senior human capital tests. The second is the reinforced fine-tuning of the sales strategies. We integrated the deep sync and other open-source frameworks with our self-developed GRPO algorithm. This has increased model accuracy by 20%. Furthermore, our AI agents can make model interactions with our users, such as real-time coordination of voice, text, and image, minimizing user waiting time.

We also upgraded it with

an emotion-adaptive system. It can dynamically adjust its speaking speed and modify its prompts based on semantic analysis. For example, our AI agent will slow down by 20% when it interacts with elderly users. Over the past two years, our 48 LLM-related patents have been granted, with over 100 patents filed and nearly 100 AI innovations. Waterdrop is one of the first insurtech companies capable of enabling AI agents to independently close deals in complex scenarios. Next, I will want to show you our CTO AI application results. Following a development path from simple automation to advanced integration, our LLM application now focuses on two directions: Copilot and AI agent directly serving users. For our Copilot application, we have quality control Copilot and sales Copilot. Our quality control Copilot can replace the manual sampling of service recordings and perform text analysis with an 88% accuracy level of human force across voice and text channels, reducing frontline quality control costs by 40%, significantly improving efficiency.

And our sales Copilot can enhance sales efficiency by enabling rapid summarization of client intent and tailored product recommendations. This tool automates the service quality and conversion rate, empowering our human agents to better meet client needs and achieve sales targets. About our AI agent, we have three different expectations, including AI health insurance agent, AI service assistant, and AI insurance planner.

Well, our AI health insurance agent can directly serve customers via voice-based conversation, with a single session lasting up to fifteen minutes, outperforming junior agents' efficiency. It currently generates millions of insurance premiums per month, projected to reach tens of millions by year-end. Building on this, our AI auto insurance agent has entered daily training with two PNC insurers, with selected projects already in user beta testing. For our AI service assistant, Xiaohui, embedded in our app, this 24/7 AI assistant handles Q&A about user policy and products of interest. By 2025, we expect AI to reduce case transfer rates by 50% and improve efficiency by 50% compared to human agents. With its long-chain reasoning ability, we are developing an AI insurance planner to provide real-time, high-quality family insurance consultation and planning services, enhancing user experience and growth potential. Talking about future outlooks, in the next three to five years, along with AI technology and application trends that I just mentioned, based on our industry know-how, we will further consolidate our leading position in the insurtech industry. Thank you for listening. Ladies and gentlemen, with that, we will conclude today's earnings conference call. We do thank you for joining. Have a good time.