10 月 23 日晚，“极客时间 9 周年 × QCon 上海同行之夜”在上海举行。活动以“AI 落地见真章”为主题，邀请 zoowork.ai 联合创始人徐文浩、奇点智能研究院研究员兼首席技术咨询师吴咏炜、知名游戏公司业务质量负责人何纯等嘉宾，与技术从业者、企业实践者及用户展开交流。本次活动由极客邦科技副总裁、极客时间首席专家尹会生主持。
当 AI 从“要不要用”进入“用出了什么价值”的新阶段，一系列更具体的问题浮出水面：企业里越来越多人使用 AI，为什么业务成果仍不明显？个人效率提升，如何转化为组织收益？AI 项目完成试点之后，怎样实现稳定运行？围绕这些来自一线的真实问题，活动通过圆桌对谈、现场问答、小组共创和案例发布，将讨论深入到业务需求、工程建设与组织协作等具体环节。
对于迎来 9 周年的极客时间而言，“AI 落地见真章”既是本次活动的主题，也体现了其在 AI 时代持续坚持的实践导向：帮助用户掌握技术、形成能力，并在真实工作中创造价值。
从关注技术进展到检验业务价值
活动开场，极客时间个人版总经理张浩回顾了平台九年来的发展。从后端开发、微服务、云原生，到大模型、Agent 开发与工程化，极客时间持续跟随技术演进，建设面向实践的专业内容。
在与专家和用户的交流中，张浩观察到，人们对 AI 的态度正在变得更加务实。除了关注模型能力与技术进展，越来越多人开始追问：AI 究竟给工作和业务带来了什么价值？
“能落地、能产生价值才是真本事。”张浩表示，这正是此次活动将主题定为“AI 落地见真章”的出发点。评价 AI 应用，需要回到它解决了什么问题、带来了什么实际成果。换句话说，AI 的答卷，最终要交给业务来打分。
活动现场，极客时间还发布了《AI 落地案例集》，汇集一线实践者的应用经历与探索，为用户和企业提供可借鉴的实践参考。通过整理和分享真实案例，极客时间希望让更多人看见具体场景中的落地路径，从他人走过的路中获得启发。
个人用上 AI 如何转化为组织收益
在随后的圆桌对谈中，嘉宾首先讨论了一个普遍受到关注的问题：企业已有不少员工使用 AI，为什么仍难产生实际业务成果？
徐文浩指出，如果企业内部的协作流程和职责边界没有相应调整，AI 带来的局部提效就可能被新增的沟通与协调消耗。让 AI 产生组织层面的价值，需要重新审视任务如何分工、流程如何衔接，以及由谁对最终结果负责。
如果说流程调整是组织问题，那在吴咏炜看来，答案还可以用一个更底层的定律来解释。他借助阿姆达尔定律指出：开发环节即使明显提速，如果需求澄清、评审、测试、交接等环节没有同步改善，完整交付链路的效率仍然受限。企业需要从整个流程出发寻找瓶颈，判断 AI 的应用效果。
如果说组织流程回答了改什么、系统定律解释了为什么，那落到具体业务里，还需要知道怎么做。何纯结合游戏行业的实践，给出了更具体的落地抓手。他强调结构化需求表达和企业知识库的重要性——工作中存在大量没有被明确说出的业务规则、风格要求和经验判断。将这些内容持续沉淀下来，才能让 AI 更准确地理解任务，减少反复生成与返工。
从工具使用到流程协作，嘉宾的讨论指向了更具体的评价方式：AI 应用既要关注任务执行速度，也要关注结果是否可用、交付是否可靠，以及整个业务流程是否得到改善。
AI 落地不是一道单选题，而是一场系统工程。
从试点走向持续运行 场景与工程能力缺一不可
AI 项目完成试点之后，如何持续发挥作用，是圆桌讨论的另一项重点。嘉宾分别从场景需求、工程能力和跨部门推广等角度分享了观察。
徐文浩认为，项目能否持续运行，首先要看所选场景是否具有真实、持续的需求。如果应用没有进入核心业务、没有解决实际痛点，就容易在短期尝试后失去使用动力。企业推进 AI 落地，应先判断这件事是否值得持续做，再讨论如何将它做好。
选对了场景、确认了方向，接下来的问题就是怎么让 AI 应用长期稳定地跑起来。何纯强调，面向长期运行的 AI 应用，需要具备稳定性保障、执行状态管理、过程回溯与异常处理等工程能力。将 AI 应用作为需要持续维护的软件和服务来建设，才能支撑它进入日常工作流程。
当单个场景跑通之后，如何推广到更多部门、更大范围，又是一道新的关卡。吴咏炜则提醒，不同部门的业务条件与工作任务存在差异，试点经验在推广时需要结合具体场景调整，并明确人与 AI 的责任划分。
围绕如何打通落地“最后一公里”，嘉宾还讨论了技术管理者参与一线实践的重要性。管理者通过亲身体验，能够更准确地理解 AI 的能力与限制，形成合理预期，并据此推动团队协作与流程改进。
对于个人，嘉宾从不同角度强调了判断力、业务理解、沟通能力和责任意识的价值。当 AI 承担更多执行工作，技术人员也需要更深入地理解需求、组织任务，并对最终交付结果作出判断。
从专家对谈到同行共创 探索具体行动路径
本次活动还设置了约 50 分钟的小组共创环节。参与者结合各自经历，围绕四大话题展开讨论，并由代表分享成果。
AI Coding 如何走向组织级研发流程？
参与者提出，个人提效走向团队实践，需要同步完善持续集成与持续交付、自动化测试、安全扫描和代码评审等基础能力。通过明确流程与质量要求，让个人使用 AI 的经验逐步转化为团队可以执行、评估和改进的工作方法。
企业 Agent 如何建立质量评估与持续迭代闭环？
讨论进一步深入意图识别、知识检索、运行轨迹和评估数据集等具体环节。参与者提出，应拆分链路、识别薄弱节点，结合业务风险引入必要的人工审核，再将运行反馈用于后续优化。
技术人如何从写代码走向解决业务问题？
参与者将关注点放在用户痛点和投入产出上：利用 AI 加快原型验证与方案试错，同时更主动地理解业务，在真实反馈中判断技术应用的价值。
当执行能力被 AI 重构，人的核心价值如何重新定义？
围绕这一话题，参与者提出，人在组织中的连接作用与责任承担值得进一步重视。一方面，需要有人将业务专家的隐性经验转化为 AI 可以理解的上下文，通过沟通、共情和信任建立跨岗位协作；另一方面，即使任务交由 AI 执行，仍需要人作出判断、承担责任。讨论也延伸到创造力、审美与个体体验，提醒从业者在提升效率的同时，持续发展自身的独特价值。
从研发流程、质量评估到业务理解与人的作用，四大话题呈现了 AI 落地涉及的不同层面。通过专家拆解、同行分享与现场共创，活动让不同岗位和行业的实践者建立连接，为各自的下一步行动提供参考。
坚持精品与专家 让学习连接真实业务
极客邦科技联合创始人、COO 司巧蕾在致辞中回顾，极客时间自创立以来，始终坚持“精品＋专家”的内容模式，以“听得懂、学得会、用得上”为品牌承诺。通过专家与内容团队共同打磨，将一线经验转化为用户能够理解、掌握和应用的专业内容，是平台长期坚持的方向。
进入 AI 时代，这一实践导向有了更具体的落点：通过大模型、Agent 开发与工程化等内容建设，帮助用户掌握方法；通过案例分享和同行交流，帮助用户理解技术进入业务场景时的条件与难点。此次 9 周年同行之夜，正是这一方向的集中呈现。
活动也与极客邦科技在 QCon 全球软件开发大会（上海站）主会场发布的新使命形成呼应。10 月 22 日，极客邦科技创始人兼 CEO 霍太稳宣布，公司使命升级为“助力企业 AI 加速落地”。进一步明确了极客邦将长期积累的内容、人才与生态资源连接到企业 AI 实践中的发展方向。
围绕这一使命，极客时间将继续发挥在专业内容、专家连接与人才培养方面的积累，帮助个人将学习转化为工作能力，支持企业将 AI 应用推进到真实业务中。
九年前，极客时间在 QCon 上海发布；九年后，平台再次与用户、专家和企业实践者相聚于此，共同探讨 AI 如何创造实际价值。技术与学习需求不断变化，“听得懂、学得会、用得上”的品牌承诺始终延续。面向新的发展阶段，极客时间将继续与同行者一起，把真实问题讲透，把实践经验沉淀下来，让 AI 在真实业务中落地见真章。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
On the evening of 23 October, the 9th anniversary of the QCon Shanghai Walking Night was held in Shanghai. The event, which focused on the theme “AI Down to the Truth”, included guests such as Seo Wen Ho, co-founder of zoowork.ai, Fellow and Chief Technical Adviser of the Institute of Intelligence in Wonderland, and the Director of Business Quality of the Distinguished Play Company, who interacted with technology practitioners, business practitioners and users. The event was moderated by Yoon Yoon, Vice-President of Polar Science and Technology and Chief Specialist of Polar Time.
When AI went from “Whether to use” to a new phase of “what value to use”, a more specific set of questions surfaced: AI was being used more and more in enterprises, why was the results of the business not visible? How can individual efficiency be translated into organizational gains? AI How can it be stabilized after the pilot project has been completed? With these real issues coming from the front line, the event, through round-table talks, on-site questions and answers, panel creation and case publish, will address specific aspects of operational needs, construction and organizational collaboration.
For the 9th anniversary period, “AI to the bottom” is both the theme of the event and the ongoing practice orientation of the AI era: helping users master technology, build capacity and create value in their real work.
From focusing on technological advances to testing business value
The event began with a review of the platform ' s nine-year development by Zhang Ho, Director-General of the personal version of the tour time. From back-end development, micro-services, clouds, to LLM, Agent development and engineering, the hours of the minibus followed the technological evolution and built a practice-oriented professional content.
In their exchanges with experts and users, Zhang Ho observed that attitudes towards AI are becoming more pragmatic. In addition to focusing on modelling capabilities and technological advances, more and more people are asking: What is the value of AI for work and business?
“It is true that you can land, you can produce value.” Zhang Ho stated that this was the point of departure for the event's designation of the theme “AI to the bottom”. Evaluation of AI applications, needing to return to what it addresses and what practical results it brings. In other words, AI responses are ultimately given to the business to score.
At the event, the ZQF94Z also featured a collection of AI landing cases, bringing together applications and explorations of front-line practitioners to provide users and businesses with useful practical references. By collating and sharing real cases, the time of the hobbies is expected to allow more people to see the landing path in the specific scene and to draw inspiration from the path that others have travelled.
How to convert AI into organizational gain
In the ensuing round-table discussions, the panellists first addressed a general concern: AI is already used by a large number of employees in enterprises, and why is it still difficult to produce real business results?
Xu Wenho stated that if the intra-enterprise collaborative processes and functional boundaries were not adjusted accordingly, the partial efficiency of AI could be consumed by additional communication and coordination. Making AI value at the organizational level requires a re-examination of how tasks are divided, how processes are linked, and who is responsible for the final outcome.
If process adjustment is an organizational issue, the answer, in the view of Wu, can also be explained by a lower law. He pointed out, with the aid of the Amdal Law, that even if the development chain was significantly accelerated, the efficiency of the complete delivery chain remained limited if the lines of clarification, evaluation, testing and handover were not simultaneously improved. Enterprises need to look for bottlenecks from the whole process to judge the effectiveness of AI applications.
If organizational processes answered what had been changed, and the system law explained why, it fell into a specific business and needed to know what to do. It is the practice of the game industry to combine it with a more specific approach. He stressed the importance of structured demand expression and the corporate knowledge base — a body of business rules, style requirements and empirical judgements that were not clearly articulated. These contents are to be deposited in order to allow AI to understand the mission more accurately and to reduce recurrence and return to work.
From the use of tools to process collaboration, the panellists ' discussions point to more specific evaluation approaches: AI applications focus both on the speed of mandate implementation and on the availability of results, the reliability of delivery and the improvement of overall business processes.
AI landing is not a single issue, but a system project.
From pilot to continuous operation, the scene and the engineering capability are indispensable.
Another focus of the round-table discussion was how the AI project could continue to function after the pilot had been completed. The guests shared their observations in terms of scenario needs, engineering capabilities and cross-sectoral outreach.
Seo Wen Ho believes that the sustainability of the project depends first and foremost on the real and continuing needs of the chosen scene. If the application does not move into core business and does not address the real pain point, it is easy to lose power after short-term attempts. Businesses push AI to land, and it should be judged whether it is worth doing on a continuing basis and then discuss how it can be done.
And the next question is how to get AI to run in a stable and long-term way. It is simply emphasized that long-term AI applications require engineering capabilities such as stability assurance, performance management, process backtracking and unusual treatment. The AI application is built as a software and service that requires continuous maintenance to support its entry into the daily workflow process.
Once a single scene runs out, how to extend it to more sectors, to a wider range and to a new level. U Wu, for its part, cautioned that the operating conditions of the different departments differ from the tasks performed, and that the pilot experience needs to be adapted to specific scenarios and to clarify the division of responsibilities between people and AI.
The panellists also discussed the importance of involving technology managers in front-line practice around how to reach the “last kilometre”. Managers are able, through first-hand experience, to understand more accurately the capabilities and limitations of AI, to develop reasonable expectations and, by doing so, to drive teamwork and process improvement.
For individuals, the guest emphasized the value of judgement, operational understanding, communication capacity and accountability from different perspectives. When AI undertakes more implementation, technicians also need to understand needs, organize tasks and judge the final delivery results in greater depth.
From experts talking about peer-to-peer co-creation, exploring specific action paths.
The event also included about 50 minutes of team-building. Participants took into account their experiences and discussed four main topics and shared the results among delegates.
AI Coding, how to move to an organizational R & D process?
Participants suggested that personal effectiveness moves towards team practice, synchronizing basic capabilities such as continuous integration and continuous delivery, automated testing, security scanning and code evaluation. By clarifying process and quality requirements, the individual experience of using AI is gradually translated into working methods that can be implemented, evaluated and improved by teams.
How does the Enterprise Agent establish a quality assessment and continuous iterative loop?
Discuss specific links such as further in-depth intent identification, knowledge retrieval, operational trajectory and assessment of data sets. Participants suggested that split links, identifying weak points, introduce the necessary manual clearances in conjunction with operational risks, and use the operational feedback for subsequent optimization.
How can technicians move from writing codes to solving business problems?
Participants focused on user pain points and input outputs: using AI to speed up prototype validation and programme error while understanding business more proactively and judging the value of technology applications in real feedback.
How can the core values of human beings be redefined when the capacity for implementation is reformulated by AI?
In this context, participants suggested that people ' s connectivity and responsibility in the organization deserved further attention. On the one hand, there is a need to translate the hidden experience of operational specialists into AI ' s understandable context, which builds cross-post collaboration through communication, mutuality and trust; on the other hand, even if the task is to be carried out by AI, it still requires judgement and accountability. Discussions also extend to creativity, aesthetics and individual experiences, reminding practitioners to continue to develop their unique values while improving efficiency.
From research and development processes, quality assessments to business understanding and the role of people, four main topics present different dimensions of AI landing. Through expert disassembly, peer-sharing and co-creation in the field, the campaign connects practitioners in different jobs and industries to inform their respective next step operations.
I'm going to stick to the quality and the expertise to connect learning to real business.
In his message, COO Schuire, Co-founder of Polar Science, recalled that, since its creation, he had adhered to the content model of “precision plus experts” with the brand commitment of “understanding, learning, using”. The translation of first-line experiences into professional content that users can understand, master and apply is a long-standing orientation of the Platform, through which experts and content teams work together.
This practice has led to more concrete points of entry into the AI era: through LLM, Agent development and engineering, to help users master the methodology; and through case-sharing and peer-to-peer exchange to help users understand the conditions and difficult points in which technology enters business. This nine-year-old peer-to-peer night is the focus of this direction.
The event also responded to the new mission of Polar Science at the QCon Global Software Development Conference (Shanghai Station) main venue publish. On October 22nd, the founder and CEO Hossein of Teletech in the state of Polarism announced that the company's mission had been upgraded to “Strategic Enterprise AI Accelerated Landfall”. Further clarity was provided as to the direction in which the extremist states are linking the long-term accumulation of content, talent and ecological resources to enterprise AI practice.
With this mission, the kiosk time will continue to accumulate in professional content, specialist connections and talent development, helping individuals to translate learning into work skills and supporting enterprises in moving AI applications into real business.
Nine years ago, at ZQpublish in QCon Shanghai, the platform came together again with users, experts and business practitioners to explore how AI could create real value. Technology and learning needs are constantly changing and brand commitments to “understand, learn and use” continue. In a new phase of development, the time of the rave will continue to work with peers to tell the real questions, to sink practical experience and to give AI the real story in real business.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.