本周策略观点： ChatGPT Agent 具备三大特性，自主任务执行：用户可以用自然语言下达指令，Agent 能够自主规划并 执行系列操作。协作与交互性：它会在需要时主动询问更多细节以完成目标。用户可以随时中断、重定 向任务或完全接管浏览器的控制权。安全与权限控制：在执行购买、提交表单、发送邮件或处理个人信 息等具有实际影响的关键操作前，Agent 会明确请求用户许可。同时，它被禁止执行如金融转账或提供 法律建议等高风险任务。OpenAI 还内置了针对"提示注入"等恶意攻击的防护措施。 1. 人形机器人订单加速落地，宇树科技启动IPO辅导，国内机器人产业加速拓展。近期，国内人形机器 人产业加速拓展，我们综合总结为三方向。 第一，头部机器人企业订单加速落地；第二，以宇树、智元机器人等为引领的机器人具身智能产业加速 上市进程；第三，多家机器人厂商持续推出机器人本体等产品，在技术上持续迭代。 人形机器人订单加速落地，商业化进程持续加速。2025 年7 月21 日，据中国招标投标公共服务平台》 显示，优必选科技中标觅亿（上海）汽车科技有限公司9051.15 万元机器人设备采购项目。7 月14 日， 中移（杭州）信息 ...

据匿名知情人士透露，Manus技术团队过去两个月集中攻关Wide Research开发，旨在通过差异化功能巩 固市场地位。联合创始人季峰计划通过视频演示展示新功能，延续其首次发布产品时的传播策略。 目前该公司已完成业务重心转移，运营主体迁至新加坡、东京和加州圣马特奥，产品尚未在中国境内上 线。其技术架构基于Anthropic的Claude等大型语言模型构建，早期投资方包括腾讯控股、HSG(前红杉 中国)及真格基金等知名机构。 尽管面临挑战，Manus仍展现出强劲的发展潜力。其产品已上架Microsoft Store，显示出在国际市场上 的认可度。此外，Manus在新加坡招聘AI工程师、数据科学家等岗位，提供高薪待遇，以吸引国际化高 端人才。 中国初创公司Manus即将推出其标志性人工智能平台自3月上线以来的最大规模功能升级。据知情人士 透露，这项名为"Wide Research"的新功能将通过部署数十个协作式AI代理同时处理海量数据 （603138），显著提升复杂任务的处理能力。该工具计划最早于本周上线，顶级订阅套餐定价为每月 199美元。 知情人士称，相较于OpenAI的Deep Research等同类工具 ...

国内机器人产业加速拓展，Al Agent 持续演进，持续看好相关产业 投资机会 证券研究报告 | 行业周报 2025 年 07 月 31 日 | 股票 | 股票 | 投资 | EPS （元） | | PE | | | --- | --- | --- | --- | --- | --- | --- | | 代码 | 名称 | 评级 | 2025E | 2026E | 2025E | 2026E | | 002463.SZ | 沪电股份 | 买入 | 1.85 | 2.45 | 29.45 | 22.24 | | 002881.SZ | 美格智能 | 买入 | 0.73 | 0.92 | 60.41 | 47.93 | | 300308.SZ | 中际旭创 | 买入 | 7.6 | 8.29 | 27.71 | 25.4 | | 300394.SZ | 天孚通信 | 买入 | 3.68 | 4.53 | 27.63 | 22.45 | | 301285.SZ | 鸿日达 | 买入 | 0.21 | 0.7 | 198.67 | 59.6 | | 600941.SH | 中国移动 | 买入 | 6.77 | 7.1 ...

图片来源： PlayerZero 随着硅谷加速迈向 AI Agent 主导软件编程的未来，一个新问题随之产生：如何在 AI 生成的代码缺陷 进入生产环境前将其发现。就连 OpenAI 也面临此类问题 。 PlayerZero 周三宣布完成 1500 万美元 A 轮融资，由 Foundation Capital 的 Ashu Garg 领投，他是 Databricks 的早期投资人。 这个问题还会因为大量 AI Agent 产出远超历史总量的代码而加剧。人类不可能总是实际检查所有 AI 编写的代码是否存在错误或幻觉。对于企业依赖的大型复杂代码库而言，这个问题更加严峻。 "PlayerZero 训练的模型能深入理解代码库，我们了解它们的构建方式和架构原理， "Koratana 说道。 他的技术研究企业历史中的错误、问题和解决方案。当系统出现故障时，他的产品能够 " 找出原因并 修复，然后从这些错误中学习，防止它们再次发生 " ，科拉塔纳说道。 他将自己的产品比作大型代码库的免疫系统。 此前该公司还获得过由 Green Bay Ventures 领投的 500 万美元种子轮融资，多位知名天使投资人参 与，包括 Za ...

AI Agent（指"AI代理"，或称"AI智能体"）如何落地企业的业务发展？ 在2025年世界人工智能大会（WAIC）期间，容联云在AI Agent落地实践产业创新论坛上正式推出"容 犀"Agent&Copilot平台，它能够通过质检代理、座席助理、座席代理、洞察代理四个智能引擎，为企业 提供覆盖营销、客服、质检、数据洞察等全场景的AI能力。 论坛期间，容联云大模型负责人唐兴才接受了《中国经营报》记者的采访。记者了解到，2024年年初， 容联云正式立项AI Agent，将原本只能"辅助人"的模型升级为"能自己干活"的智能体，目前已在多家银 行、券商机构跑通了质检、回访等关键场景。在采访中，就AI Agent落地中的机会与挑战、企业客户对 于AI Agent的需求变化，以及企业级 Agent的未来发展趋势等问题，唐兴才一一给出了回答。 《中国经营报》：你提到AI Agent"跑通"了关键场景，能不能结合具体案例解释下究竟是怎么"跑 通"的？ 唐兴才：最典型的一个标杆案例是长江证券，涉及的主要业务流程是开户质检回访，所谓质检是指质检 管理人员对服务质量实行实时监控、录音抽检、评分、问题反馈与改进建议，以确保服务 ...

从某种层面来看，如果说C端Agent平台封装是更多通用场景和个人生活场景的AI沉淀，那么在BetterYeah Agent开发平台底层，融入的是对真实的员工岗 位理解、业务节点、企业知识库沉淀、数据安全等的企业级全链条能力。 这些被封装的企业能力体现到前端，则是能在真实业务节点中落地的企业级"AI数字员工"形态。 作者|皮爷 出品|产业家 "企业级AI的本质是角色，AI即角色。"一位国内某新能源汽车厂商技术负责人表示，"就是要用AI来完成业务中的某些环节，或者节点。" 这是一个正在被达成的新共识。 如果说2025年什么是科技领域的新主题，那么AI Agent则恰会是多数人给出的答案，而在其中，对大部分企业侧的负责人而言，他们的回答则是会更为细 颗粒度——企业级Agent。 但这并不是一件容易的事。 "我们之前更多是用开源平台的方式自己搭建对应的Agent，但最终效果一言难尽。"一位金融机构的相关负责人告诉我们，"一方面是没办法基于智能体很 好地调度企业知识库，另一方面无法和现有系统进行集成，只能作为'独立项'外挂在整个系统之外，更多作用是copilot助手。" 在今年的WAIC大会上，一张特殊的答卷被给出，交 ...

Core Insights - The 2025 World Artificial Intelligence Conference highlighted the theme "Practical AI, Real Results," emphasizing the importance of AI applications in business and the need for effective implementation strategies [4][8]. Group 1: AI Development Trends - 2025 is identified as a critical year for breakthroughs in AI foundational theories, with some technologies reaching world-leading levels [11]. - The core of AI development is based on a "triple helix" model involving collaboration between foundational resources (computing power, algorithms, data), with high-quality data being a significant bottleneck [11]. - The rise of AI Agents is seen as a key trend, which will reshape human-computer interaction and create new service models, particularly in SaaS companies [11][12]. Group 2: Industry Applications and Challenges - The focus on practical applications emphasizes the need for effective business models, such as the "contractor model" for value sharing, while recognizing uneven industry penetration [11]. - AI's role in manufacturing supply chains is highlighted, particularly in addressing challenges like non-standard drawing interpretation, which requires converting data into structured formats for better factory matching [14][19]. - The integration of AI in healthcare, particularly in stroke prevention and treatment, showcases its potential to enhance precision and reduce complications [62][63]. Group 3: AI Infrastructure and Governance - The necessity for AI governance is underscored, with a call for a systematic framework to manage the risks associated with advanced AI technologies [47][48]. - The development of AI-specific chips (ASICs) is seen as a crucial trend, with significant market potential projected for the coming years, particularly in the Chinese market [41][42]. - The importance of data quality and governance in enterprise AI readiness is emphasized, with a focus on ensuring high-quality data for effective decision-making [33][35].

Investment Rating - The report maintains a "Buy" rating for several companies including Lenovo Group, SMIC, Xiaomi Group, Hua Hong Semiconductor, and Industrial Fulian [5]. Core Insights - The WAIC 2025 indicates that AI has entered a new phase of application, with a focus on commercial viability and integration into traditional sectors such as finance, law, programming, and healthcare [1][2]. - AI agents are becoming a "killer application" across various industries, enhancing efficiency and altering business logic and organizational structures [2]. - The demand for computing power is shifting towards post-training and inference needs, moving away from pre-training server architectures [3]. - The report highlights a competitive yet cooperative state emerging in the AI industry between China and the US, driven by advancements in AI applications and infrastructure [3]. Summary by Sections AI Application and Market Trends - The WAIC 2025 showcased a record attendance of over 305,000 people and online views exceeding 2.36 billion, marking a 21.6% increase from the previous year [1]. - Major companies are focusing on AI agents for practical applications in daily office tasks, finance, education, entertainment, and healthcare [2]. Computing Power and Infrastructure - Server manufacturers are emphasizing the need for AI-integrated systems that support post-training and inference, with companies like Huawei and Lenovo leading the charge [3]. - The report suggests that most companies are shifting their focus to the requirements of post-training and inference, utilizing foundational models combined with enterprise data [3]. Robotics and AI Devices - Robotics and AI glasses are gaining attention, but their market performance and technological maturity remain to be validated [3]. - Companies like Rokid and Xreal are actively promoting AI glasses, although improvements in weight, battery life, and functionality are still needed [3]. Investment Opportunities - The report identifies key investment opportunities in companies with strong competitive advantages in the AI application phase, including Xiaomi's IoT ecosystem, Lenovo's AI solutions, and SMIC's semiconductor manufacturing [5][49].

Core Viewpoint - The article highlights the unprecedented enthusiasm and advancements in the AI industry showcased at the Shanghai World Artificial Intelligence Conference (WAIC), emphasizing the transformative impact of DeepSeek and the emergence of various trends in AI technology and applications [3][4]. Group 1: Key Trends in AI - Trend 1: DeepSeek has fundamentally changed the perception of AI in China, with a growing belief in the potential for achieving AGI (Artificial General Intelligence) [6][7]. - Trend 2: New foundational large models are not only focused on state-of-the-art (SOTA) performance but also on reasoning, multimodality, and cost-effectiveness [8][11]. - Trend 3: Open-source large models have entered a new phase in China, with significant players like Tongyi Qianwen leading the way [17][18][28]. Group 2: Integration of Hardware and Software - Trend 4: The integration of chips and models is creating a fully domestic AI ecosystem, with a focus on collaboration between hardware and software [32][34]. - Trend 5: AI infrastructure is rapidly developing, with vertical industry models providing direct productivity benefits, as seen in sectors like energy and finance [50][60]. Group 3: Consumer-Focused Innovations - Trend 6: AI innovation is shifting towards consumer-facing products, with AI agents becoming a new focal point in various applications [66][81]. - Trend 7: The first wave of commercial AI terminals includes automobiles, headphones, and glasses, indicating a growing market for AI-integrated hardware [88][99]. Group 4: Robotics and Non-Transformer Architectures - Trend 8: The field of embodied intelligent robots is experiencing rapid growth, with advancements in capabilities and applications [112][134]. - Trend 9: Non-Transformer architectures are emerging from research into practical applications, showcasing innovative approaches in AI development [144][146]. Group 5: Competitive Landscape - Trend 10: The gap between China's AI capabilities and those of Silicon Valley has narrowed to approximately six months, highlighting China's unique advantages in resources and talent [150][155].

OpenAI发布具备自主思考能力的ChatGPT智能体；零一万物发布企业级Agent智能体"万仔"……一时之 间，AI Agent（AI智能体）正从概念加速落地，国内外科技巨头纷纷布局。 7月28日，在2025世界人工智能大会暨人工智能全球治理高级别会议（WAIC 2025）上，安永大中华区 发布AI Agent产品——安永智能问答3.0。安永大中华区人工智能与数据咨询服务联席主管合伙人陈剑光 在接受《每日经济新闻》记者（以下简称NBD）采访时表示，衡量AI Agent"好用"的关键指标，需兼顾 技术效能与业务价值。 陈剑光强调，AI Agent是否"好用"，技术上需关注准确性、响应速度等指标，业务上则需关注效率提 升、成本优化、风控增强等实际价值，需体现对业务目标的支撑度。 除此之外，不同行业的Agent 需求差异显著。 不过，各行业在人事、行政等职能部门存在共性需求，核心是提升运营效率。 谈Agent应用：存在两大核心痛点 NBD：目前企业在落地Agent应用时，最常遇到的痛点是什么？从您的实践经验来看，这些痛点背后的 核心原因是什么？ 这些痛点的背后，核心原因在于双重壁垒。一是技术整合壁垒。打通异构系统 ...