作者 | 周星星 编辑 | 大模型之心Tech 原文链接： https://zhuanlan.zhihu.com/p/1983512173549483912 点击下方 卡片 ，关注" 大模型之心Tech "公众号 戳我-> 领取大模型巨卷干货 本文只做学术分享，已获转载授权 ，欢迎添加小助理微信AIDriver004做进一步咨询 前言 随着 2025 年即将画上句号，我想对"Agent 元年"根据个人这一年的实践和认知进行一次收敛。 技术观点：Agent 架构之争已定，收敛至以 Claude Code 和 Deep Agent 为代表的「通用型 Agent」形态。 Claude Code 虽然在 2025 年 3 月作为"智能终端编程助手"推出，但其不止于编程。 行业认知: 2025 年作为 Agent 元年，既没有悲观者眼中的"名不副实"，也未完全达到乐观者预期的"全面替代"，而是处于稳步落地的中间态。 作为一线从业者，我的评价是： 技术已就绪，爆发在局部 。 基于以上背景，本文将从 Deep Agent 为切入点，分享我作为一线开发者在 2025 年的实战感悟。 主要参考资料： Anthropic、Lan ...

Core Viewpoint - Deep Intelligence, a leader in AI applications for marketing and sales decision-making in China, has shifted its focus to the Hong Kong Stock Exchange after previously planning to list on the New Third Board and the ChiNext, submitting an updated prospectus recently. The company reported a revenue of 540 million yuan last year, holding a market share of 2.6% [1]. Revenue Structure - The revenue structure of Deep Intelligence remains largely unchanged since its initial application to the ChiNext three years ago, with revenue sources primarily from intelligent advertising and data management. The company increasingly relies on its early business of intelligent advertising, which poses challenges for its path to the capital market due to heavy dependence on a single business and fierce competition for future growth [1]. - For the years 2022 to 2024, Deep Intelligence's revenues were 543 million yuan, 611 million yuan, and 538 million yuan, with 82.1%, 80.5%, and 85.5% of revenue coming from intelligent advertising, respectively. In the first half of this year, the revenue from this segment increased to 93.3% [2]. Client and Supplier Overlap - Deep Intelligence boasts a client list that includes 89 Fortune Global 500 companies across various sectors such as fast-moving consumer goods, automotive, and retail. However, there exists a complex interest structure where some clients are also suppliers, leading to potential conflicts of interest [4]. - The company holds a leading position in the marketing and sales decision-making AI application market with a 2.6% market share, while its closest competitors hold shares of 2.4%, 2.1%, 1.9%, and 1.8% [4]. Customer Dependency - Deep Intelligence has been criticized for its reliance on a small number of major clients. The revenue from the top five clients accounted for 51.1%, 50.2%, and 54.6% of total revenue from 2022 to 2024, with 70.2% of revenue coming from these clients in the first half of this year [7]. Competitive Landscape - The competitive landscape for Deep Intelligence includes intelligent advertising service providers, enterprise data management service providers, and other emerging players in the industry. The company emphasizes the intense market competition it faces [6].

Core Viewpoint - 深演智能科技股份有限公司 has submitted its listing application to the Hong Kong Stock Exchange for the second time, focusing on AI applications in marketing and sales, amidst a competitive and rapidly changing market [1][2]. Group 1: Company Overview - 深演智能 is a decision-making AI technology company that emphasizes AI applications in marketing and sales scenarios, ranking first in China's marketing and sales decision-making AI application market with a market share of 2.6% as of 2024 [2][3]. - The company has developed a product matrix centered around two flagship platforms, AlphaDesk and AlphaData, with the latest addition being Deep Agent [3]. Group 2: Financial Performance - The company's revenue figures for the years 2022 to 2025 (first half) are approximately 543 million, 611 million, 538 million, and 277 million RMB respectively, with net profits of about 59 million, 61 million, 22 million, and 3.6 million RMB for the same periods [3][4]. - The revenue from the top five clients accounted for 51.1%, 50.2%, 54.6%, and 70.2% of total revenue in 2022, 2023, 2024, and the first half of 2025 respectively, indicating a high dependency on a limited client base [5]. Group 3: Market Competition - The market in which 深演智能 operates is described as highly competitive and rapidly changing, with significant risks if the company fails to compete effectively against existing or emerging competitors [6][7]. - Competitors may have advantages such as longer operating histories, stronger brand recognition, and better financial, technical, and marketing resources, which could lead to price pressures and reduced profit margins for 深演智能 [7].

Core Viewpoint - Deep AI has submitted an updated prospectus to the Hong Kong Stock Exchange, showcasing strong financial performance with a revenue of approximately 277 million yuan in the first half of 2025, reflecting a year-on-year growth in both revenue and net profit, with net profit increasing by 134.3% [1][6]. Group 1: Financial Performance - In the first half of 2025, Deep AI reported a significant turnaround in cash flow management, with net cash flow from operating activities improving from a negative 20.67 million yuan to 33.05 million yuan, an increase of 53.72 million yuan year-on-year [6]. - The company's asset structure has improved, with cash and cash equivalents increasing by 32.88% compared to the previous period, while accounts receivable decreased by 19.58% [6]. - Deep AI has maintained high customer retention rates, with AlphaDesk and AlphaData showing net revenue retention rates of over 85% and 80% respectively, and an overall customer net revenue retention rate of 95.5% in the first half of 2025 [6]. Group 2: Strategic Direction - Deep AI has established a clear strategic focus with its "All in AI Agent" positioning, aiming to enhance marketing efficiency and effectiveness through intelligent agents [2][5]. - The company plans to launch its enterprise AI agent system, Deep Agent, in February 2025, followed by an upgraded version, DeepAgent Neo, in June 2025 [2][5]. - Deep AI has developed over twenty intelligent agents tailored for various business functions, including sales, customer service, and product research, creating a comprehensive capability matrix for enterprise-level intelligent agents [5]. Group 3: Market Position and Industry Outlook - Deep AI has solidified its position as a leading player in the decision-making AI sector, ranking fourth in China's overall decision-making AI application market and first in the marketing and sales decision-making AI application market [7]. - The decision-making AI market is expected to experience explosive growth, with the market size projected to reach 34.5 billion yuan in 2024 and grow to 161.5 billion yuan by 2029, reflecting a compound annual growth rate of 36.2% [7]. - The marketing and sales segment is identified as a core growth driver, with an expected market size of 20.3 billion yuan in 2024, projected to reach 94.4 billion yuan by 2029, with a compound annual growth rate of 36.5% [7].

Core Insights - Beijing Deep Intelligence Technology Co., Ltd. (Deep Intelligence) has submitted its IPO application to the Hong Kong Stock Exchange amid the transformative wave of generative AI in various industries [1] - The company holds a leading position in China's marketing and sales decision-making AI application market, with a revenue scale of RMB 538 million and a market share of 2.6% in 2024 [3] Financial Performance - Deep Intelligence's revenue increased from RMB 543 million in 2022 to RMB 611 million in 2023, but fell to RMB 538 million in 2024, a year-on-year decline of approximately 12.0% due to macroeconomic factors [3][4] - The net profit experienced a significant drop from RMB 60.7 million in 2023 to RMB 21.5 million in 2024, a decrease of 64.5% [4][5] - The company reported a revenue of RMB 277 million in the first half of 2025, showing a growth of 5.8% compared to RMB 262 million in the same period of 2024 [3][4] Business Model and Strategy - Deep Intelligence operates two main business lines: AlphaDesk for intelligent advertising and AlphaData for intelligent data management, with AlphaDesk contributing 85.5% of total revenue in 2024 [7][9] - The company has launched a new product, Deep Agent, aimed at integrating AI technology into business decision-making processes, with 17 contracts signed worth approximately RMB 11.9 million as of mid-2025 [9][10] - The company serves around 530 clients, including 89 Fortune Global 500 companies, indicating strong customer retention with a net revenue retention rate exceeding 85% for AlphaDesk [9][10] Customer Dependency and Risks - The revenue concentration from the top five clients increased to 70.2% in the first half of 2025, up from 51.1% in 2022, highlighting a significant dependency on major clients [10] - The company has a cash position of RMB 95.8 million as of mid-2025, indicating short-term financial stability, but requires IPO funding for large-scale technology investments [5][6] Shareholder Structure - The company is primarily controlled by its founders, with CEO Huang Xiaonan holding 20.96% and co-founder Xie Peng holding 14.77% of the shares [11] - Notable institutional investors include China Mobile's fund, which holds 10.92%, and other significant shareholders [11] Market Outlook - Deep Intelligence aims to leverage its strengths in data management and cross-platform advertising to expand internationally, with overseas client revenue accounting for 24.9% in the first half of 2025 [14] - The IPO is seen as a strategic move to alleviate operational cash flow pressures and position the company for growth in the AI 2.0 era, focusing on decision intelligence [14]

Core Viewpoint - DeepMind Intelligence is attempting to go public on the Hong Kong Stock Exchange after two previous unsuccessful attempts on the New Third Board and the ChiNext [1][2][3] Company Overview - Founded in 2009, DeepMind Intelligence focuses on decision-making AI applications, primarily in marketing and sales, with flagship platforms AlphaDesk and AlphaData [4][9] - The company ranks fourth in China's decision-making AI application market, holding a 1.6% market share [4] Market Context - The Chinese decision-making AI application market is projected to grow from RMB 10.6 billion in 2020 to RMB 34.5 billion by 2024, with a compound annual growth rate (CAGR) of 34.3% [4] - The market is characterized by a highly fragmented competitive landscape, with major players like DeepMind Intelligence and numerous specialized service providers [4] Financial Performance - Revenue for DeepMind Intelligence from 2022 to 2024 was RMB 542.9 million, RMB 611.9 million, and RMB 537.9 million, respectively, with profits of RMB 59.4 million, RMB 60.7 million, and RMB 21.5 million [6][7] - In 2024, revenue is expected to decline by approximately 11.9%, and profits are projected to drop by about 66% [6][7] - The company's two core business segments, intelligent advertising and intelligent data management, have both seen revenue declines [7] Recent Developments - In February 2023, DeepMind Intelligence launched Deep Agent to further integrate AI technology into its existing platforms, although it has not yet generated significant revenue [8] - In the first half of 2025, the company reported total revenue of RMB 277.3 million, with a profit of RMB 3.6 million, marking a 134% increase year-on-year [8] Investment and Funding - DeepMind Intelligence has received backing from various investors, including Shenchuang Investment, Fosun International, and China Mobile's innovation fund [3][9] - The company aims to use the funds raised from the IPO for ongoing R&D, expanding its sales network, and pursuing strategic acquisitions to enhance its product offerings [10]

Group 1 - The core viewpoint of the articles highlights the significant growth and market position of DeepAI, a decision AI technology company in China, which ranks first in the marketing and sales decision AI application market with a market share of 2.6% as of 2024 [1][2] - As of June 30, 2025, DeepAI reported revenues of approximately 277 million yuan, with profits increasing by 134.3% compared to the same period in 2024, indicating the effectiveness of the company's business model and development strategy [1] - The company is actively embracing AI and focusing on technology development, transforming AI breakthroughs into commercially scalable products, thereby solidifying its position as a pioneer in the emerging field of decision AI [2] Group 2 - The product matrix of DeepAI includes AlphaDesk, which focuses on public domain traffic acquisition and maximizes advertising effectiveness, and AlphaData, which enhances private customer management through data integration and lifecycle value operations [2] - The recently launched enterprise AI intelligent system, Deep Agent, combines intelligent agent technology with industry expertise to strengthen enterprise-level decision-making capabilities [2] - The Chinese decision AI application market is experiencing rapid transformation, with a projected market size of 34.5 billion yuan in 2024, expected to grow to 161.5 billion yuan by 2029, reflecting a compound annual growth rate of 36.2% from 2024 to 2029 [2]

The Stock Exchange of Hong Kong Limited and the Securities and Futures Commission take no responsibility for the contents of this Application Proof, make no representation as to its accuracy or completeness and expressly disclaim any liability whatsoever for any loss howsoever arising from or in reliance upon the whole or any part of the contents of this Application Proof. Application Proof of Beijing DeepZero Technology Co., Ltd. 北京深演智能科技股份有限公司 (A joint stock company incorporated in the People's Republic o ...

香港聯合交易所有限公司與證券及期貨事務監察委員會對本申請版本的內容概不負責，對其準確性 或完整性亦不發表任何聲明，並明確表示概不就因本申請版本全部或任何部分內容而產生或因倚賴 該等內容而引致的任何損失承擔任何責任。 Beijing DeepZero Technology Co., Ltd. 北京深演智能科技股份有限公司 （於中華人民共和國註冊成立的股份有限公司） 的申請版本 警告 本申請版本乃根據香港聯合交易所有限公司（「聯交所」）與證券及期貨事務監察委員會（「證監會」）的 要求而刊發，僅用作向香港公眾人士提供資料。 本申請版本為草擬本，其內所載資料並不完整，亦可能會作出重大變動。 閣下閱覽本文件，即代 表 閣下知悉、接納並向北京深演智能科技股份有限公司（「本公司」）、其獨家保薦人、整體協調 人、顧問及承銷團成員表示同意： 本公司招股章程根據香港法例第32章《公司（清盤及雜項條文）條例》向香港公司註冊處處長登記前， 本公司不會向香港公眾人士提出要約或邀請。倘於適當時候向香港公眾人士提出要約或邀請，有意 投資者務請僅依據送呈香港公司註冊處處長登記的本公司招股章程作出投資決定；有關文本將於發 售期內向公眾刊發。 ...

Core Insights - The article addresses the paradox of increasing AI tool adoption while still relying on human input for growth, highlighting a common confusion among businesses [1][2]. Group 1: AI Tool Adoption and Effectiveness - Over the past year, numerous AI tools have emerged, but the return on investment (ROI) has not met expectations, with many tools failing to deliver significant results [2][3]. - The issue lies not in the tools themselves but in their implementation; efficiency must translate into effectiveness for AI to be valuable [3][4]. Group 2: Stages of AI Implementation - Companies typically go through four stages in AI application: initial excitement, widespread tool acquisition, realization of limited effectiveness, and finally, focused and deep implementation [4][9]. - The focus should be on specific use cases, such as user operations and sales enablement, to achieve tangible results [9][10]. Group 3: Practical Applications and Case Studies - Various companies shared their experiences, such as using AI for personalized user journeys and sales training, which have shown promising results [9][11]. - Solutions include AI-driven content production and distribution across multiple platforms, significantly improving efficiency and sales [15][20]. Group 4: Human Element in AI Integration - Experts emphasized that while AI can automate repetitive tasks, human judgment and emotional intelligence remain crucial in decision-making processes [26][33]. - The integration of AI should enhance human capabilities rather than replace them, ensuring that critical decisions are still made by people [26][33]. Group 5: Future Directions and Industry Collaboration - Successful AI implementation requires deep collaboration between companies and technology providers, along with continuous optimization and iteration of AI applications [36][38]. - The importance of identifying effective use cases and measuring ROI is critical for businesses to justify their AI investments and drive future growth [32][34].