Bairong Inc. (6608) Earnings Call Transcript & Summary

Dear investors, analysts and friends from the media. Good afternoon. My name is Sandy Qin, Director of IR at Bairong. I'd like to thank all the investors online for following the development of Bairong. Bairong has released its 2024 annual results today, March 26, and we'll report and share the company's 2024 annual operating results as well as outlook for 2025 and answer your questions. This earnings call will be divided into 3 parts. Our CEO will present the business progress and outlook. Followed by our CFO, who will interpret the financial performance. And finally, the Q&A session. [Operator Instructions] Management will be happy to answer your questions after the presentation. The earnings announcement contains forward-looking statements that reflect the company's current beliefs and expectations regarding the future. These statements contain words such as anticipate, believe, intent, estimate, expect or words of similar meaning. All statements other than statements of historical facts contained in this meeting are forward-looking statements. These forward-looking statements reflect the views of the company's management as of the date thereof and are not guarantees of future performance. Neither the company, any member of the group nor any of its relevant affiliates nor any of their respective directors, officers, employees, consultants or representatives undertake any obligation or expressly disclaims any obligation or undertaking to disseminate any forward-looking statements that are updated or revised. Now I would like to introduce the company's management team present at the earnings meeting today. They are Mr. Zhang Shaofeng, Founder, Chairman of the Board and CEO of Bairong. Executive Director, Senior VP, Chief Financial Officer and Head of BaaS insurance scenario, Mr. Zheng Wei. Senior Vice President of Bairong and Head of MaaS and BaaS financial scenario, Mr. Duan Ying. Now I'd like to invite Mr. Zhang Shaofeng, Founder, Chairman of the Board and CEO of Bairong to update you on the business progress in the year of 2024.

Dear shareholders, investors, analysts, friends from the media, good afternoon. Welcome to the 2024 annual performance announcement of Bairong. It is my great honor to update you on Bairong's business performance and growth and to share with you our views on Bairong's long-term development. First, I'll briefly introduce Bairong. Our business primarily utilizes AI technology to assist thousands of institutions in 2 main areas. First, to help them manage risks and client assessment, be it financial services, insurance or nonfinancial services. And we call this segment Model-as-a-Service or MaaS. We use technology to model risk and to manage risk. And we also provide marketing support to help our clients generate more income. In other words, we assist our clients in our effort to cross-sell and to increase their revenue through -- so the Model-as-a-Service system, which leverages AI modeling to aid intelligent decision-making receives an average -- on average about 300 million calls a day. And then we, through generative AI and VoiceGPT to support intelligent marketing and operations, known as BaaS or Business-as-a-Service. We are able to deliver services and charge for fees on top of those services. We've made great progress, and we've developed first-class infrastructure. Moreover, based on the AI GPT technology, we've developed AI products such as CybotStar, an Avatar. CybotStar is an AI agent platform that can help institutions solve a practical business problems in a variety of processes and scenarios. To date, we've accumulated over 7,000 institutional clients, including household names such as Baba, Tencent, ByteDnce Xiaomi, ICBC, Bank of Communications, Ping An, SPDB China Life, China Taiping and Taikang among others. Some of the marquee accounts contributing tens of millions in revenue to our company. Second, I'll present the highlights of our operations. Macro environment last year was very challenging. Market interest rates fluctuated and banks were not having an easy time, particularly in terms of net interest margin, there was persistent pressure according to data from National Financial Regulatory Administration. Net interest margin or NIM declined from 1.69% in 2023 in fourth quarter to 1.52% in fourth quarter '24, mainly due to factors such as reduction in low interest rates, competition for deposits and adjustment to existing house mortgage interest rates. By the fourth quarter of 2024, nonperforming loan or NPL ratio dropped to 1.5%, down by 6 basis points from the previous quarters. However, it was still relatively high. And therefore, 2024 was a year of low -- weak consumption and low credit confidence. The external environment was complex and challenging, but Bairong held its ground and achieved revenue growth and sustained its profitability. Our MaaS business resumed its double-digit growth in the second half of the year, which was really not easy because last year and this year, so shrinking client budget. In other words, our growth was achieved on the back of capturing market shares and wallet shares. Our BaaS business also maintained stable growth, especially the BaaS financial scenario revenue, driven by generative AI, which increased by 19% year-over-year. Our gross profit margin remained above 70%, indicating that our investment in AI has not only brought us first mover advantage, but also creating economies of scale. In 2024 our profit margin reached RMB 266 million with a profit margin of 9%, a decline compared to 2023. However, the adjusted net profit reached RMB 376 million, and adjusted NP margin was 13%. In this challenging environment, we still achieved stability and sustainability. It couldn't have been achieved without the long-term support, trust and loyalty of our shareholders, investors and now as on behalf of Bairong, I really wanted to thank everyone for your effort. Now I'd like to go over some of the work we've done in AI. I believe Bairong's greatest opportunity lies in artificial intelligence. AI is one of the most revolutionary and disruptive technologies in history. It is definitely going to transform productivity. And I think there's a consensus forming which is the era of generative AI has arrived in almost all -- every industry will be transformed by this new technology. Fortunately, we're at the front and center of this to technological evolution. We believe the AI industry is -- has a 5-layer structural system. At the bottom layer, we have the general purpose computing power and general purpose models such as infrastructure DeepSeek and NVIDIA chips. Those are infrastructure solutions. They're like utilities that provide basic services and fundamental services. And then the second layer is industry or case specific computing power and models. For example, NVIDIA introduced some specific or specialized chips just recently. We call those players at this layer in the model factory. This is where Bairong has done very well because we've developed models tailored to specific industry fields and scenarios. We would train our model to reduce the size and to improve responsiveness so that we can reduce customer cost when it comes to computing power purchase, or utilization. And then we can also improve accuracy and reduce hallucination, thanks to our specific training efforts. And then the third layer is what we call cyber bot, which is the workflow and intelligent agent construction layer. This layer is a layer of development tools. This layer makes it possible for average employees to use drag and place to very simply create intelligent agents. And then the next layer is what we call agent market, or agent store, where marketing, sales, legal, HR personnel are able to take advantage of applications at this layer. And those applications are task-specific. And these are what we usually know are agents. And we believe there's not going to be much, if any, general purpose agents. Rather these agents are going to be specialized. And then on top of that, you have CRM, ERP and other systems that have been integrated these agents. So some of the features will be replaced by agents will be still traditional software. And Bairong focuses on the second and the third layer. We develop industry-specific LLMs as well as the model factory that we built to make it possible for people to very easily develop agents that they can use. And in the future, we might move up the layer in terms of developing AI applications for the end user. Next, I'd like to go over our business development. Over the last period, we've been fully committed to the R&D of AI technology. And we've developed AI technology infrastructure that has put us in a very competitive place in terms of product capabilities. Our VoiceGPT technology is self-developed, fully self-development, meaning that we're able to develop solutions from end-to-end, from the underlying layer to the application layer. Our LLM an LLP model is also fully self-developed, including the full stack training capabilities such as pretraining post-training, SFT, DPO and others activities. We've completed the training and deployment of multi-scenario and multi-model applications, be it risk management, marketing or loan collection or contract approval and other tasks. We've also achieved -- we also developed AI products such as CybotStar, AI agent platform and digital human All-in- One Machine. In 2024, our VoiceGPT made significant progress. And we've also made great progress elsewhere. Now if present the core technologies are self-built. And therefore, our infrastructure cost is only 20% to 30% versus our peers. We -- in terms of voice synthesis, we support over 20 dialects and are able to express emotion in different scenarios with our model. Our model is able to continue to improve and reinforce itself. It's able -- they're able to learn from real-life interaction, and with the user and to become stronger over time. Now in terms of product portfolio, we introduced a brand-new product, the CybotStar enterprise intelligent agent platform. We know that there's a -- so ByteDance has codes and then Baba has [indiscernible] While we face enterprise customers, while codes is more end-user facing. CybotStar is specifically designed for enterprise customers, and its mission to help clients apply large models to their various business scenarios. And as CybotStar, where the agent is able to take actions. It's not just a brain, obviously, the brain is the LLM, but CybotStar is the brain plus the body. It's able to process text, image, videos, and is able to get integrated into mails, messages, voicing, including WeCom. We cover clients from insurance, financial services as well as retail. In terms of payment models, our clients can choose to pay based on subscription, pay by project. We're based on the revenue, in fact, they're able to generate. So with -- this is a flexible pricing model for our clients. And CybotStar is a very powerful and flexible platform that can meet the needs of different industries and companies of different sizes. We're very excited about the future of CybotStar, and we look forward to working with customers moving forward. Bairong has been widely recognized by all areas of society. In the first half of the year, Bairong's large model, BR-LLM won prestigious awards. We were selected together with other LLMs such as [indiscernible] into the 2023 Annual Large Model VitaLITy list, which was a list put together by the Chinese Academies of Social Sciences. And then BR-LLM also awarded the 2023 annual product by the Chinese Academy of Social Sciences and eNet Research Institute. And then in the second half of the year, we received more honors in the 2024 New Technology Top 100, we were selected again by the organizer, and we were awarded the 2024 Large Model Innovation Application award. We were given the CMMI Level 5 certification, the highest level in -- by global standards. This means that we've reached the world's most advanced level in terms of software development, project management and continuous optimization. CMMI, which stands for software Capability Maturity Model Integration, is the authority in the software development capability assessment. And Level 5 is indeed the most difficult to obtain. We've obtained 319 patents as software copyrights by the end of 2024, covering fields such artificial intelligence, machine learning, privacy, computing, human computer collaboration and multi-modality. When it comes to ESG. As a leading enterprising field of data, we bear significant corporate social responsibility. Today, I would like to share with you a particularly meaningful project, one that Bairong has recently participated in. This project is beyond financial sector. It's called AI heartwarming hut. We worked together with organizations who care about adolescent well-being and developed a program empowered by AI technology. We want to look out for the well-being, especially mental well-being of young teenagers. We know that many of them suffer from mental stress, sometimes there can be dangerous incidents incurred by such mental stress. So that is why we donated AI heartwarming hut to over 100 countries across China. This hut serves an AI-powered emotional support center. It does come in the physical shape of a hut and it features an integrated digital human system designed in the image of [Foreign Language] or the Monkey King. This digital human interacts with children in their daily life, offering companionship and emotional support without the supervision of teachers and parents. So the kids can tell the monkey king avatar, what's making them unhappy or their concerns. The kids don't necessarily want to tell these things to teachers and parents, but they might want to share it with the AI digital human avatar. So this kind of interactions can help them open their mind and address potential psychological pitfalls. And at the core of this digital human is China's first large scale AI model dedicated to adolescent emotional support. We have trained this model extensively with knowledge related to youth psychological care as well as a wealth of general knowledge. This enables our digital human to answer many whimsical and imaginative questions. And moreover, the AI model can analyze speech tones and emotional cures, allowing the synthesized voice to carry a natural emotional words ensuring that the conversations with children are empathetic. Beyond emotional companionship, we have equipped the AI heartwarming hut with additional capabilities, building a powerful intelligent ecosystem to extend the protection of adolescent well-being. For example, our smartwatch features real-time location tracking and heart rate monitoring, helping to detect potential bullying incidents in schools because when kids are bullied, their heart rates go up and automatic messages and notifications will be sent to teachers and school regulators so they can intervene in time. This has attracted a lot of positive attention from across society. We have implemented this in Beijing, Shenzhen and Hubei, among other places. We'll continue to work in this direction by collaborating with charity organizations so that our AI-powered hut can reach out to more adolescents, giving them a brighter, more intelligent future. At the same time, this year, we officially launched our integrated AI Green Finance business solution. This solution is our response to the national dual carbon strategy. It helps companies detect green projects. It can also help them with benefit assessment, ESG risk monitoring thereby enhancing productivity. Our solution aligns with -- closely with the action plan for promoting high-quality development of digital finance jointly issued by the National Financial Regulatory Administration, The People's Bank of China and 5 other departments. The plan encourages financial institutions to leverage carbon accounts, carbon emissions data and ESG scores to explore innovative financial products and service models. There are several highlights. Number one is the integrated system covering underlying information systems, business processes and top-level design. It meets diverse practical needs in green finance. Second, it has 2 AI capabilities. We have embedded our foundational large model, which is BR-LLM and our AI agent application platform, CybotStar, bringing groundbreaking innovations to multiple scenarios and processes. There's also a 3-in-1 management system. Our solution helps financial institutions establish an effective management framework for clients, portfolios and institutions. Fourth, it covers development on 4 front supporting financial institutions in digitalization, informationization, technological innovation and data infrastructure to enhance green finance initiatives, has already been implemented at 2 joint stock banks achieving breakthroughs in environmental and corporate benefits. It was also recognized by the ministry of information, technology and industry. Green finance not only benefits the present but also ensure long-term stability. As the pioneering AI technology provider, Bairong will continue to leverage AI innovation, empowering financial institutions to explore new pathways and models for green finance development. Sixth, last but not least, I want to share with you our work plan and outlook for 2025, we will deepen our focus on vertical industries while expanding our investments in cross-industry fields. We're going to expand to nonfinancial sectors so as to achieve a business layout driven by AGI, structured along both vertical and horizontal dimensions. Our plan consists of four key areas. IT infrastructure, IT -- AI infrastructure. So one is general IT infrastructure. Second, specialized AI infrastructure. Thirdly, AI Internet empowerment for contract review, HR recruitment, our own customer service. And the fourth would be something that we export, which is AI capability products. So that is the intelligent body itself. So let's zoom in on IT infrastructure, we aim to make our Internet systems more flexible by adopting containerized and pooled technologies to enhance scalability. Additionally, we plan to unify multi-cloud and hybrid cloud management to improve operational efficiency. IT resources will be optimized based on the needs of different business units with a goal of achieving 99.999% system reliability while cutting unit business costs. We will continue developing multimodal large models for text and voice with the goal of building a world-class AI voice infrastructure. We'll buy more GPU, more training data, we'll also recruit more high-caliber talent so that our AI infrastructure will go from strength to strength. We aim to build a world-class AI infrastructure. And we plan to integrate AI into various internal operations such as coding, image generation and report generation. Today, AI is used to automate internal processes. About 20% of our codes are generated by AI automatically. But our internal codes are rather complicated. So 20% to 25% is already a very high percentage of automated AI-generated codes for our processes. At the same time, we will bring our CybotStar AI products as a platform to the market while launching digital human products and voice agent solutions. So in 2025, we'll definitely drive business growth and technological innovation through increased investment, enhanced infrastructure, improved internal efficiency and the development on the commercialization of AI products, both in terms of financial segment and nonfinancial business. We believe what we invest this year will pay off in the year after next or maybe next year. If we don't invest today, then we will meet more obstacles in the future. So that is why we have to do that now. In order to better seize market opportunities, Bairong has also undergone a comprehensive strategic upgrade. We're going to significantly enhance our R&D and business investments. We're going to occupy more market share regardless of cost, especially in an pan-AI industries and more generalized AI use cases. In the general finance AI segment, we want to focus on customer service and payment collection. Previously, this kind of business used to be limited by number of headcounts you can hire, but our service breaks that boundary and allows us to increase the supply indefinitely. Now for internal operations, AI helps us reduce costs and improve efficiency in contract review, customer service, HR recruitment. It has already helped to cut our cost significantly where 80% to 90% of the workload is carried out by AI and the cost is only 10% of what it used to be. This includes text-based and image-based intelligent systems. And our text base, the customer service will be launched soon. This not only boosts efficiency but also reduces labor cost. Bairong's AI voice customer service agent is about to be launched. Industries such as retail, FMCG and education are already benefiting from our AI solutions. We will continuously expand our business frontiers, leveraging AI to empower more industries step by step. With large-scale investment, this may put pressure on Bairong's profitability in 2024 and '25, but I firmly believe that these efforts will deliver longer-term value and returns in the years to come. This concludes my brief remarks for Bairong's 2024 performance and future outlook. Thank you all for taking time to attend the 2024 Bairong annual performance conference. Thank you.

Thank you, Chairman, Zhang Shaofeng, for your insightful and detailed presentation. Now let's invite company partner, CFO, Senior Vice President of BaaS Insurance Industry Cloud, Mr. Zheng Wei to present the 2024 performance and financial highlights.

Thank you to our shareholders, investors, analysts and media friends for joining us today. I will now walk you through Bairong's 2024 financial performance and highlights. First, financial overview. In the past year, our revenue reached RMB 2.929 billion, a 9% increase from last year's RMB 2.681 billion, despite a challenging year, Bairong successfully achieved stable revenue growth and maintained the sustained profitability. Gross profit reached RMB 2.142 billion, a 10% increase from last year's RMB 1.955 billion. Gross profit margin remained stable at 73%, primarily benefiting from the scalable AI cloud business model, which further realized the benefits of scale. Operating profit reached RMB 285 million with an operating profit margin of 10%. Net profit reached RMB 266 million with a net profit margin of 9%. Now IFRS net profit reached RMB 376 million with a non-IFRS net profit margin of 13%. Non-IFRS EBITDA reached RMB 486 million with a non-IFRS EBITDA margin up 17%. Now before we dive into different business lines, let's take a look at revenue structure. In 2024, our revenue saw steady growth, driven primarily by BaaS revenue powered by generative AI, which grew 12% year-on-year to RMB 1.997 billion, accounting for 68% of total revenue. Within the BaaS business, the Financial Industry Cloud revenue grew 19% to RMB 1.411 billion, accounting for 71% of BaaS segment's revenue and 48% of total revenue. Insurance Industry Cloud actively responded to regulatory changes and witnessed a slight decline of 3%, heading RMB 586 million, accounting for 29% of BaaS business revenue and 20% of total revenue. Our other business, MaaS saw a 5% increase in revenue, reaching RMB 932 million, accounting for 32% of total revenue. Now let's zoom in on MaaS business line. Our MaaS business aims to assist institutional clients in decision-making by providing models and evaluation results with diverse application scenarios. In 2024, the RMB 932 million in MaaS revenue came predominantly from core clients who contributed RMB 300 million annually -- RMB 300,000 annually. These core clients accounted for 76% percent of MaaS revenue. The number of core clients reached 211 with the average contribution per core client hitting 3.37 million. Our BaaS business aims to enhance the efficiency of assets operation in vertical industries by leveraging generative AI and large models and allows for large-scale reuse in various subsegments such as banking, insurance, wealth management and inclusive finance. Our BaaS financial industry cloud primarily helps institutions achieve more efficient intelligent marketing and intelligent operations covering business scenarios such as retail lending, small and micro lending, wealth management and leasing e-commerce. The BaaS Financial Industry cloud charges a technical service fee based on the size of asset transactions facilitated in 2024 as a leader in AIGC application implementation, our BaaS Financial Industry Cloud revenue grew 19% to RMB 1.411 billion. Our insurance industry cloud deployed nationwide through Dong Insurance Brokerage, our decision-making AI provides comprehensive customer insights, accurately recommends insurance products and operates high-value policies with the help of generative AI and offline insurance brokers. We charge a commission based on the premiums of the policies sold. In 2024, our BaaS Insurance Industry Cloud generated RMB 586 million in revenue in terms of the revenue structure. New policy revenue was RMB 487 million, down by 4% from last year's RMB 508 million. Life insurance renewal rate continued to exceed 90%, ranking among the top in the industry. Renewal revenue reached RMB 99 million, up by 2% from last year's RMB 97 million. Total premiums reached RMB 5.442 billion, up by 63% from last year's RMB 3.331 billion, significantly outpacing the 5.7% industry growth rate reported by the National Financial Regulatory Administration. Now let's look at profit and loss. In 2024, our gross profit margin remained stable at 73%, primarily benefiting from the scalable AI cloud business model, which further realized the benefits of scale. R&D expenses reached RMB 509 million, up by 34% from RMB 379 million in 2023, accounting for 17% of total revenue. The company continues to increase investments in AIGC and large models to support product supply and technological development. During the reporting period, our VoiceGPT Intelligent Voice product made significant progress in both technology and application, establishing top-tier infrastructure with all technologies being self-developed effectively lowering costs. In addition, we successfully launched the CybotStar enterprise-level intelligent platform and have integrated large models such as DeepSeek aiming to solve the application of large models in customer business scenarios. Sales and marketing expenses amounted to RMB 1.119 billion, accounting for 38% of total revenue, down by 2% compared to last year's or the previous year's total revenue. General and administrative expenses was RMB 328 million accounting for 11% of revenue remaining stable. Net profit reached RMB 266 million, down by -- down from last year's RMB 335 million with a net profit margin of 9%. Non-IFRS net profit reached RMB 376 million with a non-IFRS net profit margin of 13%. In 2024, our cash, cash equivalents and similar financial assets totaled RMB 3.657 billion, a 9% decrease compared to RMB 4.032 billion at the end of last year. This decrease was primarily due to Bairong enhancing the efficiency of fund usage, increasing strategic investments and repurchase efforts. The RMB 3.657 billion include cash and cash equivalents, RMB 739 million, large certificates of deposits, RMB 2.437 billion and RMB 480 million in current financial assets measured at fair value through profit and loss. The company remains confident about its long-term future. And in 2024, the company repurchased a total of 25.49 million B shares from the open market totaling HKD 238 million as one of the few AI companies in China to have successfully implemented generative AI in practical scenarios and achieve the profitability, Bairong has maintained a stable revenue growth and sustained profitability in 2024. Looking ahead, Bairong will continue to increase investments in R&D and marketing while empowering industry through AI. Thank you, everyone.

Thank you so much management for the wonderful presentation. Now we open the line for questions.

[Operator Instructions] So we have a few analysts who are waiting in line.

My name is Xu Ran from Morgan Stanley. This is Xu Ran speaking from Morgan Stanley. Thank you. So my first question is about AI. There's been a really rapid development of AI technologies. My question is regarding management's view on DeepSeek. And in particular, what are the core advantages of CybotStar versus DeepSeek? If there's room for cooperation? And also perhaps we can compare CybotStar with Claude and Rimba as well. We are a leader in the financial vertical. Just wondering what are the key differentiators you have in the AI field?

So Shaofeng will answer this question.

Thank you so much for your question. This is indeed a very relevant question for many investors, and we were asked the same question many times since the emergence of DeepSeek. So I think the emergence of DeepSeek is definitely a net positive for the industry. And because it's definitely made AI household technology in almost overnight. And I think that awareness -- improvement has been very, very positive. And then we can also leverage DeepSeek to develop agents. So that's a positive for us as well. In terms of the cost in the short to medium term, there's some negative impact as well because DeepSeek is freely available and it's also open source. And so some large institutions might have this perception, which is perhaps they can follow-up LLMs by themselves, or they can deploy LLMs by themselves. And so some financial -- large financial institutions had budgets still for AI models by the end of last year. But in Q1, they put the budget on pause because now they wanted to give it a try themselves first. So that's, I think, in the short term, a negative impact. Also in a short period of time, you might -- or rather in the short term, you might start to see a lot of small new entrants. Small startups who are technology service providers just like us. And so I think that's a negative as well. And these two negative factors are mostly impacting our top line. But I think in the medium to long term, it's going to be positive for us because similar to Q1 as well as some of the other LLMs, right? We're talking about general purpose LLMs that are not specialized for specific tasks, and cost will be high. And then the response tends to be still and hallucinations are bound. And the other thing is even with LLMs, you still need to build agents by yourself. Agents and LLMs are different. Many people have this misconception -- misperception that AI equals LLMs equals agents. But actually, that's not -- they are not the same thing. So with AI, you have generative AI, but then there's also decision-making AI. So AI's come in different flavors. And then AI is just the brain. For you to have a functional agent, right, there's a lot of other things that need to be added. So I think customer education is definitely needed. But in the meantime, I think we -- for us, it's just really be patient. Over time, these institutional clients will realize that AI is not the same as agent. And they still need to rely on specialists such as us to help them build agents and to deploy AI models. And I think DeepSeek is almost a symbol with political backing and with patriotism or patriotic sentiment attached to it. And you've seen so many announcements from SOEs as well as local governments regarding their deployment of the DeepSeek model because nobody wants to be left behind in this wave. And I think in the short term, there is negative impact. And I'm going to be honest about it, but that's going to be short term. And so short term, we expect some pressure on the top line, especially with our larger institutional clients. However, DeepSeek is not the only flavor of AI and AI is not agent. And also AI's can be generative AI or it can be decision-making AI's. And that -- which means the impact for us is going to be short term. And secondly, DeepSeek is general purpose AI, while we specialize in industry-specific AI. And then cyberbot is able to be very important in terms of actual applications. We are different from Claude in that Claude is more a consumer-facing, media-facing type of technology. But when it comes to enterprise AI, I think you need to align values, you need to make improvements, you need to also be cost effective. Enterprise applications need to have attractive ROIs unlike applications that are facing consumers. We -- and not to mention hallucinations. And so we develop our technologies with business clients in mind. While Rimba is also very different. Rimba is more like an agent that's a competitor to Baba's Arca. While CybotStar is more of a agent builder, which is different. So we may be compared with Claude but we're definitely not the same as Rimba. Rimba and Doubao, and Grok they are consumer-facing AI's or AI models, while CybotStar is something that's completely different. So in answer to your question, I think perhaps there's going to be some pressure on the top line, especially from large institutions. But then with small or SME clients, we don't think there's going to be any impact because they don't have the ability to deploy AI models by themselves. And I think even with large institutional clients, it's really just a matter of time. And by the second half of this year, they will have realized their misperception and they will have come back again to our services.

Thank you so much for your answer management. That's great.

Yes. So I wanted to add that the emergence of DeepSeek has been a net positive. It also increased the watermark for the industry. And the reason why we will stop at no -- we will not stop at any cost in terms of development, in terms of achieving differentiation and it's because even with open source or free solutions, right, we already have this level of certification. And therefore, we will continue to increase our R&D to the effort.

I am Omar from Citic Gentle -- CSC Financial. And in particular, I know that the company has been expanding into nonfinancial services. However, I know such expansion is not easy. I'm just wondering how should we think about the company's potential making breakthroughs in those non-financial sectors?

Thank you, Mr. Omar for your question. We will continue to double down on financial services, but we will continue to invest in nonfinancial sectors as well. In terms of what we do with financial services, there is MaaS, which is risk assessment, risk management. And then there's also marketing, which is about profit sharing. We also now help our clients to -- on compliance, on marketing material development as well as on, for example, research note development. I think pricing model will be challenging. We all know that software subscription is a challenging model in China. With generative AI, we're able to achieve many great things. We can either charge by project. For example, if you've successfully recruited a person or you've successfully completed a project where we have completed a research note, you can charge based on that. Or you could for example, charge by FTE, for example, before a research -- an article would take about 55 minutes to compete, now with our solutions, you're able to finish it within, let's say, 35 minutes. So you can perhaps arrive at the time you can save on a -- for completing a contract. And so you can charge for the difference for the cost saved. Obviously, this type of pricing, I think, requires Bairong as well as its peers in the industry to come to consensus with the end user. I think there's going to be a lag which is understandable. But I think the important thing for us for Bairong is to explore different avenues to identify legitimate needs and to develop solutions to address those needs. We need to continue to communicate with clients to change the way they pay for services and change the way they develop budget. Even in the financial services sector, we still face challenges because there are many clients who are charged by, for example, projects, which is already the preferred way because every project can be CNY 0.5 million or CNY 1 million or it could be by FTE, which is only, let's say, CNY 3,000 or CNY 5,000 a month. That for us would not be able to make much, if any, profit. Obviously, we need to develop new -- we need to work with clients and to change their mindset. So I think for us, there are two tasks. First is to identify legitimate needs. Second to change client pricing perception. Obviously, when we expand into nonfinancial services, right, it's going to be more difficult because we are already a specialist in financial services, not to mention that clients in financial services tend to be -- tend to have deeper pockets. But I think there are a lot of solutions, for example, VoiceGPT, et cetera, industry agnostic, which can be deployed in financial services, but also nonfinancial services such as call centers. So I think we are definitely able to expand into nonfinancial services, but there's going to be challenges in terms of how we price our products. So as the industry leader, we have a responsibility to educate the market to define the market from scratch. BaaS and MaaS, both of these products were defined by ourselves. Nobody created such products in these categories. So we used our own models to create a whole segment. So obviously, as spearhead or as somebody who spearheaded and created something from scratch you have to pay for a lot of cost and then you will see a lot of challenges.

Thank you both for the question and the answer. Let's take the next question from the phone. From Great Wall Securities, we have [indiscernible] online.

I'm from Great Wall Securities, the Chief Technology Analyst. My name is [indiscernible] very glad to be here to learn from you. My question has to do with your AI product portfolio. You mentioned we're going to invest a lot this year. Now where do you see yourself in 5 to 10 years or by your own? If you look at BaaS or AI models, where do you want Bairong be in terms of the size, volume or standard? Who are you benchmarking against? That's the first question.

Mr. [indiscernible], why don't you just finish all the questions to let the management answer.

I actually have a lot of questions. Do you need me to...

All right. Let me take your questions. Yes, Mr. [indiscernible], there is no benchmark. Let me just tell you clear cut, no benchmark for us. Short and clear, we are carving out a new path on our own. Maybe someday, we will become the benchmark for somebody else.

Question, in the area of AI you mentioned we are trying to craft vertical models. But be it overseas or domestic models, the general AI models are trying to make inroads into vertical AI models. So we started with verticals. Maybe you have specialized models, but it's kind of challenging for us to expand into general AI models in terms of the fine-tuning this whole operational methodology, like the R&D process, which is going to take the same amount of resources. So we are trying to shift from vertical to general. Is that the case?

Let me clarify, it's not general, it's more like a pan-industry AI model. When you see general AI, it is like a general intelligent agent or the general AI model. We don't do large models, but we could provide a general intelligent agent for, let's say, HR recruitment or customer service, it's not really general. In a general sense, as kind of cross-industry specific modules, they are specific to the different use cases of the industry. So if I understand you correctly, you're talking about the specific verticals within the industry, right? When we use the term vertical, there are several meanings attached to it. Number one, is industry specific. Second, it's use case specific. Or put in other words, for example, recruiting use case. Every industry has recruitment needs, but recruitment is different from legal review. So it's still a vertical in a sense, like it's specific to a specific task, like Office would be like contrary example because every industry uses Microsoft Office.

So basically, you're saying it can be cloud, video or text. I'm trying to ask for computers, the data and processes, like the use case it handles, like every industry has recruitment needs, but the glossary they use are mostly general, but when it comes to voice recognition, specific industries will use specific jargons in the recruitment process. So when it comes to that specific terms, your models recognition and ability to understand these different jargon will be of a different level. Is that a case? And how do you see by rolling 5 to 10 years? Do you want to tackle these verticals one by one and gradually becoming a more well-rounded general AI model because if you can cover all industries, it's not like general in a true sense. But if you have all industry covered, then maybe the aggregated whole is also kind of like a general model. Well, every industry has recruitment needs if your model can cover 60% to 70% of different industries recruitment needs, does that make you general?

I guess, it is general recruitment. And then same for training for like contract review, it's all under HR that it's use case specific. Recruitment is different from contract review or training. So it's pan-industry. By definition, you are just saying that in specific areas, in specific use cases, it spans multiple industries at the same time. But sometimes an industry can be very technical. You have to understand the jargons, the glossaries. You have to make decision beforehand as to whether it's worth diving into. If they are a high-value use case, purely recruitment price is RMB 1,500 per headcount. Now if they can pay RMB 15,000 that obviously is going to worth it. But you also have to look at the volume if they pay RMB 2,000 only slightly higher than RMB 1,500, but your volume is much smaller. So that's not enough output for the input we invest. So when I say pan-industry, what I really mean by that is that we span certain industries. It can be applied to maybe 30%. If it's only applicable to 4 to 5 different industries, only 4% to 5% of the industry is too narrow, I wouldn't call it pan-industry.

So if I understand correctly, your models, you wanted to span like 30-plus different industries. And then in specific use cases such as recruitment, marketing, legal those functional tasks, you want your models to deliver better coverage across several industries, right?

In selective industries, yes. So basically, we wanted to apply to the same use case across different industries. The use case itself might be slightly general, but by no means a general AI in on itself.

We also have questions raised online through the chat function. Let me read the questions. The first one is from CICC, Li Yada. I am CICC analyst Li Yada. I would like to ask a question about your 2025 income and profit guidance, where are you going to spend most of the R&D budget in 2025? How will the investment in new business areas affect the company's profitability? Now I'm going to give the floor to the CFO.

Thank you. As I just reported during the financial highlights, 2024 is going to be a year whereby it showcases our flexibility and resilience. In the second half of the year, we saw a 22% increase. Our MaaS business has also witnessed a positive recovery. The fundamental is sound for profit, cash flow have all maintained a high level. As you can see, in order to further deepen and widen our technical moat, our total R&D spend increased to RMB 509 million, up by 34%. We also know that CybotStar platform, digital humans and VoiceGPT are launched one after another, getting great market feedbacks. So for our AI application capabilities has been growing from strength to strength. In 2025, we expect annual revenue to achieve double-digit growth. At the same time, like the Chairman has said, we believe in create long-term value for shareholders. We'll continue to invest more in R&D expenses, building our technology pipeline so as to prepare ourselves for future competition, at the same time, will improve our operational efficiency, optimize our asset allocation so as to deliver on the growth target. We believe AI is the core strategy of our company's growth, will stay faithful to investing in AI business by estimates. In 2025, we're going to add RMB 300 million of spending through continuous innovation and resource allocation will keep our leading position in the market so as to create long-term value for shareholders.

We have another question on the line. [indiscernible] from Zhongtai Securities, please. My name is [indiscernible] from Zhongtai Securities. My question is regarding something that Shaofeng just mentioned. On new technologies, right, we mentioned CybotStar and also new progress with regard to CybotStar. Can we get an update from management in the areas of application, please?

This is Shaofeng. So maybe I can talk about VoiceGPT first. Our free and most abundant source of data will be interaction with real humans. So we have AI who have a synthetic voice, but then we also have users who are real human. Their interaction and their emotions and responses are real. And those data are not available anywhere online because what you can find online tend to be written text, which is very, very different from anything that you see online because even with Weixin messages, right, those messages are different because they're written and you can potentially withdraw a message on Weixin, right, you can rephrase your messages, but with verbal interaction, it's, I think, the most authentic way of interaction. And we also would purchase copious amount of text from third-party providers. Also, we have other data sources such as voice, such as text. So again, we have our own chat history, verbal chat history, and second is from third-party providers and sometimes we also receive data from our clients. Of course, with our CybotStar, we definitely need to differentiate ourselves from peers such as Claude. What I heard is Claude is not necessarily receiving all that attention in terms of the 2B business, perhaps they're more used to -- or focused on the consumer-facing business. And second, even if they pay attention to the 2B business, they don't necessarily have the expertise to serve customers well, institutional clients, well, because it's difficult to align values and ensure that confidentiality is protected, no data leakage and then good performance as well. A lot of our clients are large institutions who have high requirements for our solutions. And we also believe that the agents we create need to be multimodality. They need to understand voice, text, video and also needs to be able to incorporate internal data sources such as MySQL. And there's also unstructured data that perhaps lie in clients' data lakes that can potentially be leveraged as well. So CybotStar is an agent that is inside of an organization, and it needs to take full advantage of that positioning. And it needs to take full advantage of the various -- for example, resources that are available. But to make that possible, we need to just continue to invest in our pipeline, R&D pipeline.

So we have a few questions from online. This question is from [indiscernible] Securities. The question is regarding our MaaS segment if the management can perhaps update us on signs of recovery in the second half of 2024? And if those signs can point to improving pricing strength and the growth in 2025. Following a weak year of 2024, what are our plans for 2025 in terms of expansion and in terms of further investment?

Thank you for the question. So MaaS is obviously based on the output that it generates, which can help our clients make smart decisions. Bairong has built very rich client profiles over the course of its history, and it has a very close partnership with many clients. We understand our clients' preferences, their business needs and their qualifications, and among other -- a lot of other data that we've collected, which allow us to, for example, develop proprietary solutions for credit scoring and for other scoring solutions. We indeed had a year of weakness last year because of the macro environment, but Bairong is at a position of strength. It has an analyst team that's multiple times in size. It also has equally a much larger sales rep size, a much larger sales team and as well as engineering team. This is Bairong's core advantage. In 2025, Bairong will continue to expand its R&D effort to develop its CybotStar products and solutions. It will also continue to develop industry-specific solutions that are up to date. Also Bairong will continue to capture both new market shares and new wallet shares. It is our belief that our MaaS competitiveness will continue to improve, which will put us in even more advantageous market position in the market. We've been offering MaaS for over -- for close to a decade, and we are definitely a leader. And so from user acquisition to conversion to paying users as well as high-value users who bill us -- who pay us over CNY 300,000 a year, right, we've had a very good tier of customer base. So as we continue to introduce new products and solutions and as we continue to expand our client base, we believe that there's going to be a lot of upside for us in terms of ASP and in terms of, I think, further expansion of the client base.

Our next question is from Huatai Securities [indiscernible].

This is [indiscernible] from Huatai Securities. My question is regarding negative impact from the insurance sector. Wondering how Bairong intends to respond to the declining -- to the issue of declining fees?

Yes. Thank you for your question. The insurance industry indeed faces multiple challenges. There is the adjustment of product pricing rates and the correction of customers' purchasing power and willingness to pay. All these things have put pressure on the demand. And then the insurtech and intermediary services are undergoing reshuffling with a new policy environment. Despite the short-term pressure, we are still bullish on the long-term value of the sector. We believe demand driven by demographic changes is in elastic and it's accelerating. The demand from older demographics and greater awareness of health, all point to irreplaceability of insurance products in terms of risk management. And also the market is transforming from price competition to value creation, I think, which for insurers, this means higher requirements for professional capabilities. And this has created, I think, upside for Bairong to provide its specialized services. In 2024, we remain committed to customer-centric approach, deepening the development of specialized services and enhancing capabilities for our clients. Our insurance cloud platform integrates AI to provide intelligent support for thousands of advisers. And we have a network that covers over 20 provincial level regions, and we have over 100 local outlets to support operations of our clients. And then we have accelerated the application of generative AI and decision-making AI, and we've developed system solutions for sales optimization and customer value enhancement. On the other hand, we've accelerated the development of our support solutions for more efficient business management, which aligns with industry trends. For year 2024, the scale of premium and facility exceeded CNY 5 billion, a year-over-year increase. That's much higher than the growth rate of the industry average. Our insurance business demonstrated strong resilience in a challenging environment. Looking ahead, we will continue to uphold the customer-centric philosophy, deepen application of AI, enhance service experience and improve our operational efficiency. We believe that with greater professional expertise and digital capabilities, we'll be able to take a leading position in the market and create more value for shareholders and our clients.

We have Mr. [XXXXXXXXXXXXXX. The question is, in recent years, with technological innovation, financial security has also become a hot topic. There have been negative public opinions regarding issues such as personnel information leakage and the marketing calls harassment. How does the management feel these issues from the perspective of data security and privacy protection, what measures have been taken to ensure compliance with relevant laws, regulations and industry standards. I'm going to give the floor to Mr. Zhang.

As a technology company focused on serving corporate institutions, we always view data security and user privacy protection as a core responsibility. The company has already passed the National Information Security level Protection Level 3 certification also known as the Ministry of Public Security Certification. We have also received international software industry CMMI Level 5 certification, ISO 27701 privacy information management systems certification, The People's Bank of China Enterprise Credit Business Operating Record certificate and China Quality Certification Centers, Information Technology, Service Management System and Business Continuity Management System certification, as well as certifications from the HTAi Certification Center in Privacy Information Management, Information Security Management System and Quality Management System. Additionally, we are a member of the Privacy Computing Alliance of the China Academy of Information and Communication Technology. Moreover, we have continuously maintained a AAA enterprise credit rating in Beijing for several consecutive years. From technology to management process, we fully meet both domestic and international high standards of full privacy protection. We have made continuous investments in privacy computing federated learning, blockchain and other fields. We also have developed the Bairong privacy computing platform, which places us at the industry forefront in terms of data security. Our self-developed VoiceGPT has powerful intelligent voice capabilities in order to reduce customer complaints and improve the accuracy of product matching, we combine decision-making AI and generative AI at a commercial application level. Our MaaS business system with its rich user profiles and product understanding can significantly reduce customer complaint risks and deliver suitable products to the right audience through precise MaaS matching so that we make sure that the right products are given to the people who really need it and then this really helps to drive down customer complaints ratio. From the perspective of data security and privacy protection, we have taken various measures to ensure the safety of customer information during voice interactions. First, the conversation content of VoiceGPT. Such content undergoes strict review both internally and at Bairong and at institutional client level, furthermore, during the pre-training and inference phases of the large model, VoiceGPT sets up a dedicated safety barrier mechanism where additional checking agents oversee the data. These measures help ensure data security and compliance. At the same time, Bairong CybotStar provides full process model generation control, making content generation safer. We use robust technologies and strategies to construct three layers of defense, pre-interception, generation, intervention and post processing. This ensures the protection of content generation throughout all stages from input to output. This mechanism not only immediately blocks the generation of harmful or sensitive information, but also allows us to trace the content generation process, ensuring that the generated content is secure, reliable and trustworthy.

Thank you so much, Mr. [indiscernible]. In the interest of time, I still see a lot of questions from investors and analysts. Feel free to schedule another meeting to seek your answers. Now please scan the QR code on the screen to schedule a meeting with our AI-backed assistant. And also we are going to update you on the latest developments at Bairong through our regular announcements. And if you have other questions, feel free to reach out to our IR website at ir.brgroup.com or e-mail us at [email protected]. Thank you so much for your participation in our 2024 performance conference. We look forward to seeing you again in our future conferences. Thank you. [Statements in English on this transcript were spoken by an interpreter present on the live call.]