2026 年 10 月 24 日，1024 模力工场 · AI 社区日活动在杭州西湖区云谷中心顺利举行。
本次活动以「AI落地，此刻发生」（AI For Real Business）为主题，汇聚了企业创始人、技术与产品负责人、高校研究者等二十余位一线实践者。
AI 到底能帮我们做多少事，又有哪些事还做不好？带着这个问题，本次活动的主论坛围绕工作与生活中的常见真实需求，以 14 场“AI+？”分享，呈现 AI 在办公、科研、制造、艺术创作、商业决策等领域与场景中的实际进展、实践者踩过的坑，以及仍待解决的难题。
活动的主论坛分上午、下午两场进行。14 位嘉宾结合正在推进的项目与过往实践案例，分享了 AI 带来的实际变化、落地过程中面临的挑战，以及相关探索与思考。
在上午的主论坛中，还有一个非常重要的环节：即极客邦科技南方总部、TGO 鲲鹏会全球总部成立仪式。极客邦科技南方总部，及极客邦科技旗下的 TGO 鲲鹏会的全球总部，正式落地杭州。
除主论坛外，活动还设置了汇集 15 个摊位的 AI 造物市集，供参与者现场体验 AI 产品、与开发者面对面交流，并参与互动集章抽奖。
下午，还同步开展了「AI 社区怪点子工坊」和「30 秒预告片工厂」两场 Hands-on 工作坊，参与者在导师指导下动手制作 AI 小工具或原创预告片，并通过人气投票选出获奖作品。两场活动的一、二、三等奖分别为：价值 888 元的得到 AI 录音卡、500 元京东卡和 AI Coding 键盘等奖品。
上午主论坛：AI 让工作、研究和创作多些可能
极客邦科技南方总部、TGO 鲲鹏会全球总部成立仪式
当天上午，极客邦科技南方总部、TGO 鲲鹏会全球总部成立仪式，在主论坛现场举行。来自政府、园区、企业、社区、基金会及投资机构等多方面的 10 位代表齐聚现场。
在仪式现场，极客邦科技创始人兼 CEO 霍太稳；园区代表，云谷中心产业运营负责人杨璘璘；政府代表，西湖区委人才办朱霞颖；紫金港科技管委会代表陈婉卿；客户代表，阿里云市场品牌部总监凌丹；TGO 鲲鹏会代表芦宇峰；社区代表，魔搭社区朱琳；协会基金会代表，杭州市开源人工智能基金会苏帅；投资人代表，赛智伯乐姚东；OPC 社区代表，Y/OUR SPACE 璐瑶，共同登台见证了这一时刻。
极客邦科技南方总部，与极客邦科技旗下 TGO 鲲鹏会全球总部同步落地杭州，进一步完善了极客邦科技在企业 AI 落地服务与科技领导者交流成长方面的布局，为连接区域产业需求与技术人才、推动交流合作提供了新的支点。
极客邦科技，是一家综合性的“企业 AI 落地服务”提供商，旗下产品有 InfoQ 极客传媒、极客时间、模力工场和 TGO 鲲鹏会，累计服务超过 4000 家企业和 400 万用户。 为更好地服务华东地区、尤其是浙江的客户，极客邦科技成立南方总部，近距离支持企业 AI 落地与智能化升级。
而 TGO 鲲鹏会，是极客邦科技旗下极具影响力的科技领导者同侪成长平台，学员包括具有技术背景的 CEO、CTO、技术 VP、架构师等，采用实名付费学员制，严格审核学员资格，旨在组建全球化的科技领导者社交网络，线上线下相结合，联结杰出的科技领导者学习和成长。在 2026 年年初，因为在 AI 开源生态建设方面的突出贡献，被评为杭州“五小凤”。
当前，国家正深入推进“人工智能+”行动，让科技创新走进千行百业。杭州也在持续支持人工智能产业发展，从技术创新、人才集聚到应用场景和创业社区建设，为 AI 落地创造条件，这里活跃的技术人才与创业团队、丰富的产业需求，以及对科技创新的政策支持，为极客邦科技扎根杭州、连接更多企业与技术人才提供了良好条件。
上午主论坛，还有 6 位嘉宾围绕 AI 落地，展开了精彩的主题分享。以下内容由模力工场整理编辑，按嘉宾登场顺序呈现。
1. AI+社区｜霍太稳，极客邦创始人兼 CEO
极客邦科技创始人兼 CEO 霍太稳，为 1024 模力工场·AI 社区日的主论坛开场致辞，并以《AI 破局的生态法则：从零和博弈到共生智能》为题作主题分享。
霍太稳在开场致辞中表示，举办此次社区日、推动总部落地杭州，是希望立足本地，通过持续的社区建设与交流合作，为 AI 应用落地培育生态。
在主题分享中，他提出，AI 行业的关注点，正从大模型的参数竞争转向智能体及具体业务应用，于此同时，企业的需求正在变化：相比几天的培训，它们更希望服务方带着团队做出能用的智能体与 Skill，让学习直接连接实践。所以极客邦科技，也将“助力企业 AI 加速落地”作为业务升级的重要方向。
霍太稳表示，会用 AI 工具，不等于能用它解决业务问题。企业有需求、有预算，却不容易找到靠谱的团队；技术团队能做项目，也担心做完后收不到钱。模力工场想解决的，就是双方找人难、建立信任难的问题。
他介绍称，模力工场计划用 AI 帮助企业判断一个项目大概需要多少时间、多少预算，再根据团队过去的交付表现推荐合作方；技术团队也能据此了解企业需求和预算，选择合适的项目。双方建立合作后，平台还希望帮助跟进交付、评估结果。
他认为，这件事亟需更多生态合作伙伴一起来做，而极客邦科技过去近 20 年积累的企业与技术人才资源，正是推进这项工作的基础。
2. AI+垂类 Agent｜毕玄，贝联珠贯创始人兼 CEO
贝联珠贯创始人兼 CEO、前阿里巴巴集团研究员毕玄，以《以 SREAgent 为例：看垂类 Agent 的竞争力》为题作主题分享。
SREAgent，是一款面向企业 IT 系统的 AI 运维助手，通过分析监控数据和日志，帮助运维人员排查故障、定位原因并处理问题。
毕玄以 SREAgent 的实践为例，讨论了一个垂类 Agent 公司绕不开的问题：大模型越来越强，客户为什么还需要垂类 Agent？
他提到，SREAgent 主要通过分析监控数据和日志，帮助运维人员排查故障、定位原因并处理问题。
毕玄坦言，他们团队曾认为，让 Agent 长时间执行任务、熟练使用通用工具，是拉开竞争差距的关键；但随着大模型进步，这些能力正逐步成为模型本身的能力，难以再作为垂类公司的长期壁垒。
在他看来，机会藏在具体业务的“脏活累活”里。以运维为例，每家企业的监控系统、数据来源和采集方式都不一样，如何接入并用好这些数据，需要逐一解决。他介绍，SREAgent 3.0 已尝试让客户用自然语言描述需求，由系统生成相应的 Skill，降低数据接入门槛，但这种功能也容易被同行追上。
更难复制的是背后的专业经验：面对大量日志，哪些信息有用、该如何处理，又该怎样结合客户业务，建立反映系统运行情况的模型？这套模型还要能随着业务系统的变化持续更新，否则仍离不开人工维护。
毕玄将垂类 Agent 的竞争力归纳为四个方面：数据采集、数据处理、业务建模和模型持续迭代。而最终的检验标准，是 Agent 能否把问题解决好，运行是否可靠，行为是否受控。
3. AI+FDE｜芦宇峰，深空之序联合创始人 COO
深空之序联合创始人兼 COO 芦宇峰，以《FDE：AI 落地的核心竞争力》为题作主题分享。值得一提的是，深空之序是芦宇峰与前钉钉 CEO 叶军共同创办的 AI 公司，专注于跨境电商场景的 AI 业务智能体 Agent Nine。
首先，他分享了一个刺绣项目中的教训：AI 生成的图案看起来不错，老师傅一看，却发现机器需要剪线、打结 700 多次。从“生成结果”到“实际能用”，这中间的差距，正是 FDE 需要深入现场解决的问题。
在这个项目中，团队最初尝试让 AI 模仿老师傅操作软件，却发现复杂的工艺判断难以靠复制操作完成，于是转向利用标注数据和后训练，根据原图、面料和针法直接生成针迹文件。据芦宇峰介绍，部分中低难度图案已能实现分钟级产出，但具体工艺问题仍需团队驻厂，陪着老师傅实际打版、反复验证和调整。
第二个案例，是在义乌的报关项目中，团队面对的则是零散、模糊的照片和单据，通过现场收集资料、改进识别能力，将多数报关单的制作时间从至少 3 小时缩短至约 5 分钟，单份制作成本从最低约 70 元降至约 10 元。
芦宇峰强调，FDE 不能只在旁边听、拿着设备录，而要亲手做一遍业务，弄清老师傅凭经验作出的判断，以及流程中可能出现的例外。
收集到的业务资料和操作经验，一定要整理进知识库；解决问题的方法则要沉淀为可复用的 Skill，并配套覆盖真实场景和客户运行环境的测试用例，避免“开发时没问题，到了客户现场却用不了”。
在他看来，交付成效最终要看省了多少时间、降低了多少成本。而 FDE 更长远的价值，是把一次项目的经验变成产品能力，让下一个遇到同样问题的客户，不再需要团队从头驻场解决。
4. AI+办公｜张献涛，阿里云智能集团研发副总裁
阿里云智能集团研发副总裁、终端智能计算事业部负责人张献涛，以《云电脑的演进：从人的办公桌面到智能体的工位》为题作主题分享。值得一提的是，他在 2014 年加入阿里云，2017 年带领二十余人团队半年干出飞天神龙架构——中国第一套大规模商用 DPU，与 AWS 的 Nitro 仅相差一个月。这套后来成了阿里云的底座，扛住双十一流量，Gartner 测评中阿里云 IaaS 基础设施的四项核心能力全部拿下最高分。
张献涛提出，Agent 要真正替人完成工作，也需要一个“工位”：一台能运行软件、访问业务系统，并且持续在线的电脑。
过去，云电脑主要供人远程办公；如今，它也能成为 Agent 操作软件、处理跨系统任务的工作环境，让 Agent 能做的事从调用浏览器延伸到使用更复杂的应用。但给 Agent 一台电脑，并不意味着它就能在企业里顺利工作，还需要明确它以什么身份登录、能访问哪些数据，以及操作过程如何记录和审计。
他以订单处理为例：Agent 可以接收新订单、核对信息、生成报告并提交审批，但遇到登录失效、验证码或需要人工判断的情况，流程就可能中断。这时，云电脑需要保存当前工作状态，通知用户接管，待问题处理完后，再让 Agent 接着做，而不是从头重来。
据张献涛介绍，好未来的一个项目在不到两周的时间里，通过企业级云电脑部署了约 1000 个 Agent，为教师提供教案、课件和作业批改等方面的支持。
在他看来，云电脑多年积累的身份权限管理、资源调度和协作能力，正好对应了 Agent 进入企业工作流程后的实际需求：既要有稳定的运行环境，也要能管住权限，并让人在关键时刻接得上手。
5. AI+科研｜王欢，西湖大学工学院助理教授、博士生导师
西湖大学工学院助理教授、博士生导师、高效智能计算实验室（ENCODE Lab）负责人王欢，以《AI 智能体对科研的影响》为题作主题分享。
王欢有两个重点研究方向：Efficient AI（高效人工智能），关注如何提高 AI 的运行效率、降低资源消耗；AI Evaluation（AI 评测），关注如何定义和衡量智能。
这次分享中，他结合计算机和 AI 研究的经历，讨论了 Agent 带来的两方面影响。一方面正好也关于“效率”：读文献、写代码、做实验等工作正在加速；另一方面则关于“评价”：论文更容易产出，也给成果鉴别、同行评审和科研评价带来了更大压力。
他认为，Agent 确实加快了科研进程：Agent 能自主决定下一步、调用工具执行任务，Harness 则提供上下文管理、工具调用等支持，让大模型能够持续开展工作；他自己每天早上 6 点，还会收到 Agent 筛选推荐的 10 篇相关文献。
但论文产出更快，不等于科学进步更快：科研旨在增进认识，结论需要接受检验，大量成果都声称有新发现时，如何判断它们是否可靠、究竟新在哪里，就成了更紧迫的问题。
与此同时，科研评价仍看重论文等量化成果，因此当一些老师倡议“慢科研”、呼吁重质量而非数量时，也有学生反问：老师自己发了那么多论文，为什么要求我们慢下来？
在王欢看来，Agent 的影响已超出效率提升，触及科研评价制度：论文越来越容易产出，评价体系该如何识别真正的贡献？
不过，他仍持乐观态度：AI 接手部分工作，也让人有更多空间思考，哪些探索和判断仍需要自己承担，以及人的独特价值在哪里。
6. AI+艺术｜程佳伟，奇点川越创始人
复旦大学中文系硕士、奇点川越创始人程佳伟以《AIGC 何以成为艺术？浅谈生成技术演进与我的创作实践》为题作主题分享。
程佳伟指出，想让 AI 演出细腻的情绪，与其反复修改提示词，不如自己先演一遍。他分享了团队的一个实用方法：录下真人表演作为参考，让 AI 跟着处理微表情和动作，比单靠文字描述更稳定、更可控。
在他看来，如果只把 AI 当作省钱、省时间的工具，就低估了它对影视创作的影响：非科班创作者也有机会把脑中的镜头做出来，小团队能够承担更多制作任务，创意从构想到验证的距离正在缩短。
但生成变容易了，故事并不会自动变好，人物设定、角色成长和分镜脚本仍要提前想清楚，否则后面就容易跑偏、反复返工。
他介绍道，团队会先确定脚本、测试并固定提示词模板，再交给 Agent 批量生成素材，让人不必始终守着每一次生成。AI 的帮助也不止于“出画面”：音画同步生成可以减轻后期匹配音乐、音效的负担，剪辑软件中的 AI 功能也能处理部分细节工作，但专业影视剪辑目前仍不能完全交给 AI。
与此同时，人物长相保持一致，只解决了部分问题，不同镜头里的场景和空间关系能否接得上，仍是生成影像的一道难关。他还提醒，追求高质量内容不能只看画面效果，也要重视素材授权，审慎对待对他人画风、形象和声音的直接模仿，并保留创作与处理过程的记录。
程佳伟认为，工具技术改变的不只是制作速度，也会改变作品的表达方式，正如手持摄影让镜头走向街头、计算机让影像不再完全依赖实拍，AI 也在打开新的创作空间。
但新方式出现，不代表旧方式必须退场：今天仍有人用胶片拍电影，AI 也不会让此前所有创作方式失去价值。
下午场主论坛：AI 走进业务和生产现场
下午主论坛，共有 8 位嘉宾围绕 AI 落地，展开了精彩的主题分享。以下内容由模力工场整理编辑，按嘉宾登场顺序呈现。
7. AI+商业决策｜胡旭升，YouNavi.ai 创始人兼 CEO
YouNavi.ai 创始人兼 CEO 胡旭升，以《从“命令式工具”到“启发式参谋”：多源 Context 如何帮助我看清商业机会、推进关键合作》为题作主题分享。值得一提的是，胡旭升拥有约 18 年的大厂产品工作经验，曾先后在腾讯工作 8 年、滴滴 5 年、阿里 2 年、吉利 3 年。
胡旭升抛出了一个值得很多上班族思考的问题：会开了这么多，目标到底往前推进了多少？
不过他想讨论的，不只是如何用 AI 记好一场会，而是如何让它在听过多场会、了解自己的业务目标之后，成为能提供建议的“参谋”。
所谓“多源 Context”，就是把多场会议、日常对话、文档等信息联系起来，让 AI 了解他的业务目标、遇到的问题、接触过的人，以及事情推进到了哪一步。
他自述，与 AI 语音对话已超过 1000 小时，两年多录下 7000 场会，并尝试将这些信息按事实、人物关系和行动记录等整理、持续更新，让 AI 提供建议时有背景可循。比如来参加这次活动前，他就把海报发给 AI 助手 Navi，Navi 根据他的业务需求推荐了三位嘉宾，中午他就找其中一位聊了聊。
这种辅助贯穿会前、会中和会后：会前提示重点话题与问题，会中提供回答思路，会后梳理待办、责任分工和潜在风险，并推送到微信，让讨论接上后续行动。
另外，在招聘中，他也会利用 Navi 对公司日常讨论的了解，辅助准备面试问题、分析人岗匹配；每天复盘时，则要求 AI 尖锐指出自己说错了什么，再从交流中找出值得吸收、经过查证的观点。
这就是他所理解的“启发式参谋”：结合用户目标和前后信息，帮助发现遗漏、提出建议、修正判断，而人的责任仍是作出决定，并通过行动验证。
同时，他也强调，这种持续了解用户的能力，必须建立在用户授权、安全和信任的基础上。
8. AI+创业｜Mingke，茗日智能(MRS.ai) 创始人
茗日智能(MRS.ai) 创始人 Mingke，以《我在 2017 年关闭了一家 Personal Agent 公司》为题作主题分享。值得一提的是，现在大家都知道 Agent 很火，但 Mingke 其实已经做 Agent 12 年了。
他讲了一段从个人助手转向企业服务、再到探索 Agent 之间协作的经历。
早在 2016 年，Mingke 的团队就想做一个能替用户安排生活的个人智能体：不只推荐餐厅，还要真正把座位订下来，却发现听懂需求只是开始，难的是拿到餐厅、酒店等服务方的接口，把事情办成。
但一次投资人的反问让他停了下来：是你拿下所有餐厅的资源更快，还是美团、点评做出你这套技术更快？
2017 年，他停止了面向个人用户的创业方向，随后找到另一条路：团队有智能体技术框架，Visa 等企业有服务资源和履约体系，双方结合，才有机会把用户的一句话变成真正完成的订单。
这段经历让他认为，大模型虽然大幅改善了自然语言理解和推理能力，但并没有自动解决服务供给的问题：AI 知道该做什么，不等于它有条件做成。
更深的矛盾在于，品牌是否愿意把客户关系交给第三方，让自己变成后台一个被调用、被比价的接口，因此 Personal Agent 能否成立，不仅取决于技术和体验，也取决于用户与服务方是否都愿意参与。
如今，他把重点转向 Agent 之间的互操作：每个服务方保留自己的客户关系和服务能力，遇到自身无法满足的需求，再与其他 Agent 协作，并探索转介、分成等合作方式。这种协作也适用于企业内部，让不同部门、不同技术框架下的 Agent 能够配合，而不必全部推倒重建、统一成同一种系统。
在他看来，新的机会在于让分散的服务真正协作起来，而通信协议只是起点，背后还需要服务衔接、信任关系和收益分配等机制，才能把“彼此能对话”变成“共同办成事”。
9. AI+具身智能｜刘晨，云深处科技产品总监
云深处科技产品总监刘晨，以《足式机器人的技术进展与行业应用》为题作主题分享。
刘晨介绍称：机器人可以用 AI 识别路上的台阶和坑洞、规划落脚点，还能理解“把东西送到指定地点”“不要走草坪”等指令，自主规划配送路线。
他分享了一个巡检现场的意外：机器人原本每天自主执行三段任务、每段两小时，却因维修时井盖被打开而掉进了井里。
这个案例暴露出按既定路线行走的局限，也促使团队探索将环境感知与运动控制结合，让机器人提前识别坑洞、台阶等障碍，自主规划落脚点，而不只是碰到障碍后再调整姿态。
AI 的作用也体现在理解任务上：在浙江大学校园的演示中，机器人能够根据自然语言指令拆解配送任务、规划路线，并结合视觉识别，遵循“不要走草坪”“靠右行”等要求。
刘晨将机器人的核心任务概括为：运输、侦查和处置。如果要像快递员一样完成配送，机器人还得学会开门、按电梯、刷门禁，这也是团队探索用机械臂操作日常设施的原因。
不过，完成一次演示不等于能适应所有现场：在工业场景中，机器人已能执行部分配电柜按钮、旋钮操作，但换一个柜子就可能失准，如何让学到的操作适用于不同设备，仍是难点。
据刘晨介绍，云深处已有机器人在能源电力、应急消防和工业运维等场景交付运行，接下来的挑战，是让这些能力在更多不同现场稳定发挥作用。
10. AI+品牌增长｜李婷，思创客品牌公司创始人
思创客品牌公司创始人、资深品牌战略专家李婷以《AI 驱动的用户需求洞察与品牌科学增长方法论》为题作主题分享。
李婷认为，用 AI 做营销，首先要弄清用户为什么买、为什么不买。
比如，客户说“太贵了”，未必就该降价：因为这可能是预算不够，也可能是觉得竞品更划算，或是还没理解产品为什么值这个价。
原因判断不同，后续方案也会不同，AI 的作用是在此基础上帮助团队制定方案、测试效果。
她分享了一个女装案例：品牌过去什么火就卖什么，却忽略了一款销售排名第四、第五的塑形裤，团队借助 AI 分析用户反馈后发现，修饰梨形身材和大腿线条的需求被反复提及，这款产品值得重点投入。于是，团队将其定位为“超模裤”，重新梳理产品特点和直播话术，把创始人的真实经历讲清楚，让用户理解“不用勒得难受，也能穿得好看”，再通过反复测试调整表达。
李婷认为，AI 可以帮助发现用户需求、评估客户线索、复盘销售沟通和测试营销内容，但前提是提供真实的产品信息、品牌特点与用户反馈，而非套用通用模板。业务人员仍需亲自了解客户、作出判断，再借助 AI 加快执行、验证效果。
11. AI+制造业｜邱良军，正雄企业 CIO
正雄企业 CIO、TGO 鲲鹏会苏州分会负责人邱良军以《让工作长出能力：我的 AI 实践之路》为题作主题分享。
在制造业里，即使同样叫采购或销售，不同企业的具体流程和业务规则也可能差别很大，这让服务这些岗位的软件开发常常面临一个难题：业务人员说的需求，开发人员未必理解到位。
邱良军介绍，团队用 AI 辅助制作可操作的原型，再用 SPEC（规格说明）补充计算公式和业务规则，让业务人员先上手试，确认是否符合实际工作、有没有遗漏的情况，再推进开发。
他还提出一个正在探索的设想：采购人员花半天整理资料、分析需求后，能否让 AI 把这段工作过程总结成 Skill，让下次做同类任务的人不必再从头摸索？
这样能留下的，不只是一份报告，还可能是同事能复用的方法，或后续开发软件的需求依据。
不过他也提醒，个人效率提高 5 倍、10 倍，企业整体也未必提效，因为制造业的工作需要多个岗位协同，单个环节变快，不代表整条流程就顺了。要让这些 Skill 在企业内复用，还得核对方法是否准确、不同岗位的做法是否适用，并处理内部数据的访问权限与安全问题。
他还观察到，AI 有时已经把功能做出来了，开发人员却还没真正理解业务，一旦出问题，能否解决仍是考验。而把老师傅的经验变成可复用的能力，也需要让贡献经验的人获得相应回报和成长机会，才能让他们愿意分享、持续参与。
12. AI+工程｜高阳，漫飞实验室（MindFly Lab）创始人
漫飞实验室（MindFly Lab）创始人高阳，以《Knowledge as Code：AI 原生团队的协作架构实践》为题作主题分享。
代码已经让 Agent 写了，它却不知道团队昨天刚改了需求？
高阳关注的正是这个问题：执行交给了 AI，项目知识却还散落在群聊、会议和文档里。
据他介绍，团队已全部使用 AI 进行代码研发，但项目越做越复杂，让 Agent 跟上变化，比快速生成第一版更难。
他的办法是“Knowledge as Code”：借鉴 Git 的版本管理方式，把项目知识整理成结构清晰、可查找、可追溯的文档，让人和 Agent 都能使用。其中，当前决策可以更新，历史记录则沿 timeline（时间线）保留，既告诉 AI“现在该怎么做”，也能回答“三个月前为什么那样做”。
日常需求和讨论则整理成持续更新的 live page（知识页），让 Agent 按任务查找相关资料，不必把所有内容塞进一个大文档。共享这套知识后，不同 Coding Agent 和团队成员也能接着已有进展工作，减少反复交代背景。
高阳主张，让 Agent 更多承担文档整理与维护，人重点审核需求、决策和结果，让知识库真正成为 AI 干活的依据，而不是项目做完后就没人更新的存档。
13. AI+安全｜张新波，Corint AI 创始人 & CEO
Corint AI 创始人兼 CEO 张新波，以《Agent 在风控领域的应用与实践》为题作主题分享。
张新波指出，判断一笔交易有没有风险，只需毫秒级；调整背后的策略和模型，却可能要忙上好几天。
所以他希望用 Agent 缩短这段时间，让风险发现、原因分析、策略优化衔接起来，再经人工确认上线。
要做到这一点，Agent 得先读懂实际数据库中的表、字段和关联，必要时结合源代码理解业务含义，再由人工修正歧义。一句“昨天审批通过率为什么下降了”，可能需要 10 步、20 步甚至 50 步查询，因此不能只生成一条 SQL，而要通过 ReAct 循环，边分析、边执行、边纠错，最后用图表呈现结果。
在特征工程中，Agent 提出的风险线索还需计算和验证，用 IV、缺失率、PSI 等指标检查区分能力、数据完整性和分布稳定性，并通过 Harness 保障长任务持续执行。据他介绍，在一次与 Kaggle 专家团队的对比测试中，系统用约 40 分钟生成了 700 个有效特征，原先人工处理则需要数天。
策略修改也被纳入类似代码开发的流程：用统一语言描述规则，提交到 Git，再通过 PR 审核、合并，让每次改动有迹可循。他还分享，一个客户将分析流程搬到线上后，负责人连续两周没再收到数据下载审批请求，既省去了下载后再加工的步骤，也减少了数据导出。
在张新波看来，Agent 的价值是接手繁复的分析与执行，让人把精力放在业务判断和风险把关上。
14. AI+增长｜沈涛，上海鲸阙人工智能有限公司联合创始人
上海鲸阙人工智能有限公司联合创始人，帆软软件 战略副总裁沈涛，以《AI 助力企业找增长方向、找商机客户，实现效率到利润的转化》为题作主题分享。
沈涛提出了一个足以打动大多数企业老板的问题：AI 让员工一年多做 1000 份幻灯片，能多带来多少利润？
对此，分享了一次化工企业的交流：大家讨论了两个小时的研发、生产提效，休息时却发现，选定的技术路线可能已经失去竞争力，这让他更加关注一个问题：
方向错了，AI 可能只是让企业更快地走错路。
沈涛介绍，他们团队会用 AI 综合分析企业的招聘、专利、招投标和财务数据，从中寻找潜在商机。比如，一家企业招标采购 3150 千伏安的变压器，同时大量招聘检验、测试人员，再结合相关专利信息，就可以推测它可能正在扩大产能，并有加强质量控制的需求。对检测设备商、数据分析公司和人力服务公司来说，这就是值得进一步跟进的客户线索。
在一家制造企业探索电池管理系统业务时，团队则先梳理产业链、同行和下游市场，再筛选潜在客户及其近期需求，让“寻找新增长点”逐步落到具体客户上。他还介绍，AI 可以从活动报道等资料中提取人物与企业的关联，帮助寻找可能的商务引荐路径。
在沈涛看来，专业数据、业务知识和 AI 分析结合起来，才能更有针对性地辅助企业选方向、找客户、跟进商机。而提效最终有没有价值，还要看这些线索能否转化为订单、带来利润。
两个 Hands-on 活动：有想法？动手做！
下午的两场 Hands-on 工作坊活动，将重点转向了动手实践。
「AI 社区怪点子工坊」与「30 秒预告片工厂」同步开展，从下午 13:30 开始一直持续到 17:00，吸引了在校大学生、艺术家，以及来自软件开发、金融、新媒体等不同领域的参与者。
大家带着各自的专业经验与生活观察，探索 AI 如何融入自己的工作、创作与日常生活。
在「AI 社区怪点子工坊」中，Steerable 开发者王泰担任导师，带领参与者借助 AI Coding 尝试制作可交互的小工具、网页或产品 Demo。
创作的起点，藏在参与者各自的工作与生活中：日常做决策时，能否多一个帮手？任务安排怎样更贴合个人需求？股票分析能不能借助 AI？
带着这些具体问题，大家一边交流思路，一边动手尝试，让原本停留在脑海中的想法，逐渐变成可以展示和体验的雏形。
另一边，「30 秒预告片工厂」由奇点川越创始人程佳伟担任导师，围绕故事构思、分镜设计与 AI 视频生成展开指导。
参与者借助 AniShort，将脑海中的情节与画面转化为原创短片或预告片，尝试用影像讲述自己的故事。
原本互不相识的人，因为共同感兴趣的方向组队，在交流中补充彼此的想法，也一起讨论如何将创意落到具体功能或镜头上。不同背景带来的视角差异，成为现场共创的一部分。
一个下午的时间，参与者们尝试把自己的需求和创意做成作品，也在彼此的展示与交流中，看到了 AI 在不同工作与生活场景中的用法。
##感谢共创合作伙伴
1024 模力工场·AI 社区日活动的顺利举办，离不开合作伙伴们的大力支持。
感谢紫金港科技城对本次活动的指导；感谢联合主办方云谷中心和 TGO 鲲鹏会杭州分会；感谢 Hands-on 工坊联合举办方：七牛云和 AniShort，分别为参与者提供大模型 Token 和视频创作积分支持，感谢特别合作伙伴七牛云提供 Token 卡作为现场伴手礼。感谢”得到大脑 App”提供 7 天会员卡，以及作为活动奖品的价值 888 元的得到 AI 录音卡，帮大家高效记录。
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 24 October 2026, 1024 Modular Ai Community Days were successfully held at the Cloud Valley Centre, West Lake District, Hangzhou.
The event, which was dedicated to the theme "Ail Landing, This Time" (AI For Real Business), brought together more than 20 front-line practitioners such as the founders of the business, the directors of technology and products, and the researchers of universities.
AI, how much can we do? What can we do? With this question in mind, the main forum of this event revolves around the common real needs of work and life with 14 "AI+? "Sharing, presenting the actual progress of AI in the areas of office, scientific research, manufacturing, artistic creation, business decision-making and scenes, the pits that practitioners step through, and the challenges that remain to be addressed.
The main forum for the event was held in the morning and afternoon. 14 Guests shared practical changes brought about by AI, challenges during landing, and related exploration and reflection in the context of ongoing projects and past practice.
At the main forum in the morning, there was also a very important link: the opening ceremony of the Southern Science and Technology Headquarters in the state of Polarism and the Global Headquarters of the TGO Club. The southern headquarters of Polar Science and Technology, as well as the global headquarters of the TGO Foundation under the umbrella of Polar Science and Technology, officially landed in Hangzhou.
In addition to the main forum, the event created an AI-market collection of 15 booths for participants to experience AI products in situ, to interact face-to-face with developers, and to participate in an interactive stamp draw.
In the afternoon, two workshops were held simultaneously, "AI " and "30-second trailer studio ", at which participants worked under the guidance of their mentors to produce an AI small tool or original premonition and to vote for the winning work. The first, second and third prizes for the two events were: an AI recording card for $888, a 500 Won tonca and an AI Coding keyboard.
Main morning forum: AI to make work, research and creation more possible
The opening ceremony of the Global Headquarters of Teletech, TGO.
That morning, a ceremony was held at the main forum for the inauguration of the South Indian Science and Technology Headquarters in the state of Polarism and the TGO Global Headquarters. Ten representatives from a wide range of government, parks, businesses, communities, foundations and investment agencies gathered.
On the occasion of the ceremony, the founders of the technology of the state of Tomonaga and CEO Houta; the representative of the park, Yang Zheng, head of industrial operations at the Yunya Centre; the representative of the government, Zhu Xia Wing, the representative of the West Lake District Commission; the representative of the Thongqing Port Authority for Science and Technology; the representative of the client, Ling Dang, director of the Alibaba Cloud market brand; the representative of the TGO Society, Joo Feng; the representative of the community, Joo-Lin; the representative of the Foundation of the Association, General Commander of the Open-Agen Foundation of the City of Hangzhou; the representative of the investor, Sai Ji-Blu-Yao; the representative of the OPC community, Y/OUR SOPACE, who came to the podium to witness the moment.
The southern headquarters of Polar Science and Technology, which synchronizes with the global headquarters of TGO under the umbrella of Polar Science and Technology, in Hangzhou, has further refined the layout of Polar Science and Technology for enterprise AI to exchange growth with science and technology leaders, providing new niches for linking regional industrial needs and technological talent and promoting communication and cooperation.
Polar State Technology, a comprehensive "business AI landing service" provider, has a variety of products such as InfoQ ' s extremist media, time, modulation workshops and TGO conferences, with a cumulative service of over 4 million businesses and 4 million users. To better serve the clients of the East China region, especially Zhejiang, Polar Science and Technology has set up a South Headquarters to support enterprise AI landing and intellectual upgrading.
The TGO Club, a peer-to-peer platform for highly influential science leaders under the umbrella of technology in the state of extremes, includes technical-based CEOs, CTOs, technical VPs, architects, etc., using the real-name paid student system, with a rigorous review of their qualifications, aimed at forming a global network of science leaders, combining them online and offline, and linking them to the learning and growth of outstanding science leaders. At the beginning of 2026, given its outstanding contribution to the construction of open-source ecology in AI, it was rated Hangzhou as the "Fifeng".
At present, the State is moving forward with the "Advanced Intelligence Plus" initiative to bring science, technology and innovation to the forefront. Hangzhou is also continuing to support the development of artificially intelligent industries, from technological innovation, the pooling of talent to applications and community-building for entrepreneurship, to create conditions for AI to land, where dynamic technocrats and entrepreneurial teams, rich industrial demand, and policy support for science and technology innovation provide good conditions for polar science and technology to take root and connect more firms to technological talent.
In the morning, the main forum and six other guests made an excellent thematic sharing around AI. The following are edited by the model workshop and presented in the order of guests.
1. The AI+ community is too stable, founder and CEO of the extremist state
The founder of Teletech and CEO Ho was able to open the main forum for the 1024 Modified Fields-AI Community Day and to share it under the theme " The Code of Ecology of the AI Buster: From zero to symbiotic intelligence ".
In his opening remarks, Houtha said that the organization of the Community Day and the promotion of the Landscape of Hangzhou at its headquarters were meant to foster ecology for AI applications through local and sustained community building and communication cooperation.
In the thematic sharing, he suggested that the focus of AI was shifting from LLM ' s parameter competition to AI agent ' s and specific business applications, and that, at the same time, the needs of enterprises were changing: compared with the training of a few days, they would have preferred service providers to bring a team to work with Skill, so that they could learn to connect directly to practice. As a result, Ultra-Hot State technology has also made “Assisting Enterprise AI Accelerating Landfall” an important direction for business upgrading.
Haugh said it would use an AI tool, not the same way it would solve operational problems. Businesses have needs and budgets, but they do not easily find reliable teams; technical teams are able to do projects and fear that they will not receive money after completion. The problem of finding and building trust between the two sides is the problem that the model workshops are trying to solve.
He described the model workshop plan as using AI to help enterprises determine how much time and budget a project might take and to recommend partners based on the team ' s past delivery performance; the technical team would also be able to learn about the enterprise ' s needs and budget and select the appropriate project. With the establishment of cooperation between the two parties, the Platform also hopes to help follow up on the delivery and evaluation results.
In his view, the matter urgently needed to be done by more ecological partners, and the business and technological resources that had accumulated over the past 20 years in Polar Science and Technology were the basis for taking the work forward.
2. Agent Bixon, founder and CEO of the United Nations University
The founder and CEO of the United Nations Development Programme (UNDP) and former researcher of the Z9Z Group, Baekgent, shared the theme “In the case of SREAgent: Looking at the Competitiveness of Agent”.
SREAgent, an AI transport assistant for the enterprise IT system, helps transport personnel to sort out failures, locate reasons and deal with problems by analysing surveillance data and logs.
By using the practice of SREAgent as an example, it was discussed that there was no going around Agent: LLM was getting stronger, and why did the client still need Agent?
He mentioned that SREAgent helped transport staff to sort out failures, locate causes and deal with problems, mainly by analysing surveillance data and logs.
As the LLM advances, these capabilities are gradually becoming the capacity of the model itself and are no longer a permanent barrier to proximate companies.
In his view, opportunities were hidden in the “dirty work” of specific operations. For example, each enterprise has different monitoring systems, data sources and collection methods, and how to access and use these data needs to be addressed on a case-by-case basis. He explained that SREAgent 3.0 had tried to allow clients to describe demand in natural languages, generate the corresponding Skill by the system and lower the data access threshold, but that this function was also easy to follow by peers.
What is more difficult to replicate is the professional experience behind it: in the face of a large number of logs, what information is useful, how should it be handled, and how should it be combined with client operations to create models that reflect the system ' s performance? The model will also need to be continuously updated as business systems change, otherwise manual maintenance will remain indispensable.
Pixon has summarized the competitiveness of Precipient Agent into four areas: data acquisition, data processing, business modelling and continuous iterative modelling. And the ultimate test is whether Agent can solve the problem, whether it works properly and whether it acts under control.
3. The co-founder of the Deep Space Order, Al+FE, Peak
Co-founder of the Deep Space Order and CoO Top, on the theme " FDE: A core competitiveness ". It's worth mentioning that the deep order of space is the AI company, co-founded by the CEO Ye Army, which focuses on the cross-border electricity business of the AI AI agent Agent Nine.
First of all, he shared the lessons of an embroidery project: AI created a pattern that looked good, and the teacher looked at it and found that the machine needed wire cutting and knotping 700 times. The gap between "generation of results" and "practical use", is exactly what FDE needs to do on the ground.
In this project, the team initially tried to get AI to imitate the teacher to operate the software, but found that complex process judgments were difficult to complete by replicating, and then shifted to using tagging data and post-training to directly generate needle files based on original drawings, fabrics and needles. According to Joo-Fung, some of the low- and medium-cost patterns have resulted in minute-level output, but specific process issues still require a team presence at the factory, accompanied by the teacher, who does the actual editing, validation and adjustment.
In the second case, in the customs clearance project in Yiu, the team faced scattered and vague photographs and documents that reduced the production time for most customs declarations from at least 3 hours to about 5 minutes, and the cost of single production from a minimum of about $70 to about $10, by collecting information on the ground and improving identification capabilities.
Joo-Fung stressed that the FDE could not just listen to and record the equipment, but would do it hand-in-hand to ascertain what the teacher had learned from experience and what exceptions might have been made in the process.
The collection of operational information and experience must be consolidated into the knowledge base; the solution to the problem is to settle into a reusable Skill, accompanied by test examples covering the real scene and the client's operating environment, so as not to “be developed in such a way that it is not used at the customer's site”.
In his view, delivery would ultimately depend on how much time and cost was saved. The longer-term value of the FDE is the ability to turn the experience of a project into a product, so that the next client with the same problem will no longer need a team presence to solve it.
4. Vice-President, Development and Research, Alibaba Cloud Smart Group, AI+ Office
Alibaba Cloud, Vice President of the Research and Development of Smart Group, and Head of the Terminal Intelligence Computing Enterprise, Zhang Yantao, shared the theme on the theme " The evolution of cloud computers: from human desktops to jobs in AI agent ". It is worth mentioning that he joined the Alibaba Cloud in 2014 and led a team of more than 20 people in 2017 to work six months out of the Dragon Flying Dragon structure - China's first large-scale commercial DPU, just a month from the AWS Nitro. This was later the base of Alibaba Cloud, holding up the double-xi flow, and Gartner measured Alibaba Cloud Iaas ' four core capabilities as the highest score.
Zhang Yantao suggested that if Agent was to really do his job, he would also need a “position”: a computer that could run software, access business systems and maintain a continuous online presence.
In the past, the cloud computer used to be used mainly for teleworking; it can now become the working environment for Agent's operating software, handling cross-system assignments and allowing Agent to move from a browser to more sophisticated applications. But giving Agent a computer does not mean that it can work well in the enterprise, but it needs to be clear as to what identity it logs in, what data it can access, and how the operation is recorded and audited.
He cited as an example the handling of orders: Agent could receive new orders, check information, generate reports and submit them for approval, but the process could be interrupted if the login lapses, the authentication code or the need for manual judgement. At this point, the cloud computer needs to maintain its current working status, inform the user to take over, and then get Agent to do it again, instead of starting over.
According to Zhang Jingtao, in less than two weeks, a project for the future has deployed about 1,000 Agents through enterprise cloud computers to support teachers in terms of teaching, course materials and job modification.
In his view, the identity permission management, resource mobilization and collaborative capacity that cloud computers have accumulated over the years corresponded to the real needs of Agent after entering the enterprise workflow: to have a stable operating environment, to be able to manage permission and to be able to take over at a critical time.
5. Assistant Professor and doctoral director, Faculty of Technology, University of West Lake Division
Assistant Professor at the University of West Lake College of Technology, Ph.D. Teacher, Head of ENCODE Lab, shared on the theme " Impact of AI AI agent on scientific research ".
Wang Jing has two main areas of study: Efficent AI (Efficient Artificial Intelligence), which focuses on how to improve the efficiency of AI operations and reduce resource consumption; and AI Evaluation, which focuses on how to define and measure intelligence.
In this sharing, he discussed the two implications of Agent in the context of his computer and AI research experience. On the one hand, the work on “efficiency”: reading literature, writing codes, doing experiments, etc. is accelerating; on the other hand, “evaluation”: papers are easier to produce and put more pressure on results identification, peer review and scientific evaluation.
In his view, Agent did speed up the scientific process: Agent was able to decide next step on its own, call tools to carry out his mission, and Harnesse provided context management, tool call, etc., to enable LLM to continue its work; he himself received 10 relevant literature recommended by Agent for screening every morning at 6 a.m.
However, the output of the papers is faster than that of scientific progress: scientific research is aimed at increasing awareness and conclusions need to be tested, and it becomes even more urgent to determine whether or not they are reliable and where they are new when a great deal of results claim to be new.
At the same time, scientific evaluation still values the results of quantization, such as dissertations, so when some teachers advocate “slow science” and call for quality rather than quantity, others ask: why do teachers ask us to slow down when they have published so many papers?
In Wang Jing ' s view, the impact of Agent has exceeded efficiency by touching on the system of scientific evaluation: papers are becoming easier to produce, and how should evaluation systems recognize real contributions?
However, he remained optimistic that AI would take over part of the work and give more room for reflection, what exploration and judgement still had to take on itself, and where the unique value of human beings was.
6. The founder of the Al+ art arts programme, Cheng Jia Wei, Kitagawa
The Chinese master's degree in Chinese at the University of Jordan and founder of Kitagawa, Cheng Jia Wei, how can it be art? It's about generating technological evolution and my creative practices.
Cheng Jia Wei said that instead of changing prompt over and over again, I would like to do it myself. He shared a practical approach of the team: recording reality performances as a reference to allow AI to follow micro-images and movements and be more stable and manageable than textual descriptions alone.
In his view, if AI is used only as a tool to save money and time, its impact on video creation is underestimated: Non-scientific creators also have the opportunity to create shot in their minds, with small teams able to take on more production tasks, and the distance between conception and validation of creativity is shrinking.
But it becomes easier to generate stories that do not automatically change, and character creation, role growth and storyboard scripts still have to be thought out in advance, otherwise they are prone to running back and back.
He explained that the team would first identify the script, test and fix the prompt template and then give it to Agent, which would produce asset in bulk, so that it would not have to keep an eye on each generation. AI's help goes beyond "shows": Synchronization of acoustic production can reduce the burden of post-production voiceover music, sound effects, and some of the AI features in the editing software can handle some of the details, but professional video clips are still not fully available to AI.
At the same time, the image of the person is consistent and solves only part of the problem, and whether the different scenes in shot and the spatial relationship can be connected remains a challenge in generating images. He also cautioned that the pursuit of high-quality content should not only be visual, but also focus on the asset mandate, be careful about the direct imitation of the style, image and voice of others and keep records of the process of creation and processing.
Cheng Jia Wei believes that tools change not only the speed of production, but also the way the work is expressed, just as hand-held photography takes shot to the streets, computers make images less fully dependent on reality, and AI is opening up new creative spaces.
But new ways do not mean that old ones have to leave: Today there are still people making films on film, and AI won't make any of the creative ways before them.
Afternoon home forum: AI Access to business and production sites
In the afternoon, the main forum, with eight guests, featured an excellent thematic sharing around AI. The following are edited by the model workshop and presented in the order of guests.
7. AI+ Business Decision-Making Hu Wing, Younavi.ai Founder and CEO
Younavi.ai's founder and CEO Hu Wok-kwan, shared the theme "From command tools to inspirational staff officers: how multi-source Context helps me see business opportunities and advance key partnerships." It is worth mentioning that Hu Winglón has about 18 years of work experience in large plant products, eight years in Tencent, five years in drops, two years in Ali and three years in Gilly.
Hu Wing has thrown out a question worth thinking by many workers: how much has gone and how much has gone forward?
But what he wanted to discuss was not just how to remember a meeting with AI, but how to make it a “staff officer” who could provide advice after hearing many meetings and knowing his business objectives.
The term "multi-source subject" is to link information about multiple meetings, daily conversations, documents, etc., so that AI can understand his business objectives, problems encountered, people who have been contacted, and where things go.
He stated that the dialogue with AI had been over 1,000 hours, recording over 7000 meetings over two years, and tried to organize and keep this information up-to-date by facts, personal relationships and records of actions, so that AI could provide recommendations in a context. For example, before coming to the event, he sent the poster to AI's assistant, Navi, who recommended three guests based on his business needs, and he talked to one of them at noon.
This support runs through the pre-session, in-session and post-session sessions: pre-session alerts to key topics and questions, providing answers, post-session mapping of in-session, segregation of duties and potential risks, and forwarding to WeChat for follow-up to discussions.
In addition, in recruitment, he would use Navi ' s knowledge of the company ' s day-to-day discussions to assist with the preparation of interview issues and the matching of analysts; and on a daily round-up, he would ask AI to point out what it had said wrong and to identify from the exchange points of view worthy of absorption and verification.
This is what he understands as an “inspiring officer”: helping to detect omissions, make recommendations, amend judgements in the context of user objectives and information, while the human responsibility remains to make decisions and to verify them through action.
At the same time, he stressed that such a continuous understanding of user capabilities must be based on user authorization, security and trust.
8. Founder of AI+ Enterprise Mingke, MRS.ai
Mingke, founder of MRS.ai, shared the topic on the theme " I closed down a personal Agent in 2017 ". It's worth mentioning that now everyone knows Agent is hot, but Mingke has actually done Agent for 12 years.
He spoke about the experience of moving from personal help to business service to exploring collaboration between Agent.
As early as 2016, the Mingke team wanted to be a personal AI agent who could arrange a life for a user: not just recommended restaurants, but actually booked seats, but found that understanding needs was only the beginning, and that it was difficult to get access to service providers, such as restaurants, hotels, etc., and get things done.
But one of the investors asked him to stop: did you take the resources of all the restaurants faster, or did you take the company and rate your technology faster?
In 2017, he stopped his entrepreneurial orientation towards individual users and then found another way: the team had a technical framework of AI agent, and companies such as Visa had service resources and performance systems, which combined to make it possible to turn a user ' s sentence into a real completed order.
This experience led him to believe that LLM, while significantly improving the ability to understand and reason natural languages, did not automatically resolve the problem of service provision: AI knew what to do, it did not mean it was conditional.
The deeper paradox is whether the brand is willing to hand over customer relationships to a third party and turn itself into a second-stage, second-hand, competitive interface, so the establishment of a personal delegate depends not only on technology and experience, but also on the willingness of users and service providers to participate.
Today, he turns his focus to inter-operability between Agent: Each service provider retains its customer relationship and service capacity, meets its own unmet needs, then collaborates with other Agents and explores ways of collaboration, such as referral, division, etc. This collaboration also applies within the enterprise, so that Agent, in different sectors and in different technical frameworks, can work together, without having to be completely re-established and integrated into a single system.
In his view, the new opportunity lay in genuine collaboration between decentralized services, and communication agreements were only the starting point, with mechanisms such as service interfaces, trust relationships and distribution of benefits behind them, in order to turn “dialogues” into “business together”.
9. AI+ is smart, Liu Sun, Director General of Science and Technology Products in Deep Clouds
Liu Sun, Director General of Scientific and Technological Products, deep in the clouds, shared the theme “Technology advances and industrial applications of foot robots”.
Liu ' s introduction states that robots can plan their distribution routes autonomously using AI to identify steps and pits on the road, plan their landing points, and understand instructions such as "getting things to a designated location" "do not walk the lawn".
He shared an accident at the inspection site: the robot, which was supposed to perform three missions per day and two hours per day, fell into the well because the head was opened during maintenance.
This case has exposed the limitations of walking along a given route and has prompted teams to explore the integration of environmental perception with movement control, allowing robots to identify obstacles such as pit holes, steps, etc. in advance, and to plan autonomously for landing points, not just to adjust postures after encountering them.
The role of AI is also reflected in the understanding mission: in a demonstration on the campus of Zhejiang University, robots were able to dismantle distribution missions, plan routes in accordance with natural language directives and, in conjunction with visual recognition, follow the requirement of “do not walk the lawn” to the right.
Liu Morning summarized the core mission of robots as transport, detection and disposal. To complete the distribution like a courier, robots have to learn to open doors, press elevators, brush doors, which is why teams explore the use of mechanical arms to operate everyday facilities.
However, completion of a demonstration does not mean adaptation to all the sites: In the industrial scene, robots have been able to execute the distribution of electric buttons, buttons and buttons in the department, but the change of cabinets may fail, and it remains difficult to adapt the lessons learned to different devices.
According to Liu ' s presentation, robotics are already operating deep in the clouds in the areas of energy power, emergency firefighting and industrial transport, and the challenge is to stabilize these capabilities in more different locations.
10. Al+ Brand Growth Li Ting, founder of Creative Commons
Li Ting, founder of Creative Brands and senior brand strategy specialist, shared the theme " AI-driven user demand insight and brand science growth methodology ".
Li Ting believes that marketing with AI first needs to find out why users buy, why not.
For example, customers say “too expensive” and it is not necessary to reduce the price: it may be a lack of budget, it may be a more profitable competition, or it is not clear why the product is worth it.
For different reasons and for different follow-up programmes, AI ' s role is to help teams develop programmes and test results.
She shared a case of women: brands sold what they had in the past, ignored a number of plastic pants that ranked fourth and fifth in sales, and the team analysed user feedback with AI, and found that the need to fix pear body and thigh lines had been repeatedly mentioned, a product that deserved focused input. As a result, the team positioned it as “super-styres”, recouping the product characteristics and live talk, clarifying the true story of the founders, making it clear to the user that “it is not too difficult to get strangulation or dress well” and expressing it through repeated testing.
Li Ting believed that AI could help identify user needs, assess customer leads, double-sales communications and test marketing content, provided that it provided real product information, brand characteristics and user feedback, rather than applying generic templates. Operational personnel still need to know their clients first-hand and make judgements before using AI to expedite implementation and validate results.
11. AI+ Manufacturing Zhu Liang Army, Zhengyu Enterprise CIO
Qiu Liang, the head of the Zhuizhou branch of the Zhengyu Company, shares the theme of "Making Work Work Work Work: My Ai Way to Practice."
In manufacturing, even if it is also called procurement or sales, the specific processes and business rules of different enterprises may vary considerably, which often presents a challenge for software development to serve these positions: the needs described by the practitioners are not necessarily understood by the developers.
The team uses AI support to produce an actionable prototype and SPEC to complement the formulae and operational rules, allowing the operator to test first to ascertain whether the actual work is being performed and if there are omissions before further development.
He also proposed an idea that was being explored: could procurement staff, after spending half a day collating information and analysing needs, allow AI to summarize the process into Skill, so that those who would be doing similar tasks would no longer have to look at it?
This could leave more than just a report, but could also be a way for colleagues to reuse or the basis for subsequent software development.
He cautioned, however, that the increase in individual efficiency was five or 10 times, and that the enterprise as a whole was not likely to be effective, as the work in manufacturing required multiple job synergies, and the individual chain was fast enough not to represent the whole process. Re-use of these Skills within the enterprise also involves checking the accuracy of the methodology, the applicability of different job practices and processing internal data access to permission and security issues.
He also observed that AI had in some cases put the functionality in place, that developers had not really understood the business, and that resolution was still the test if problems arose. Turning the teacher ' s experience into a reusable capacity also requires that those who contribute experience receive commensurate returns and opportunities for growth in order for them to be willing to share and remain engaged.
12. Founder of the AI+ Engineering Zhongyang, MindFly Lab
The founder of the MindFly Lab, Gaoyang, shared the theme of " Knowledge as Code: the practice of collaboration structures of the original team ".
The code has already been written by Agent, but he doesn't know the team just changed its needs yesterday?
It is this issue that is of concern to Gaoyang: implementation has been given to AI, while project knowledge is scattered in community chats, meetings and documents.
According to him, the team has all used AI for code development, but the more complex the project is, the more difficult it is for Agent to keep up with changes, than to quickly generate the first edition.
His approach is “Knowledge as Code”: using Git's version management to organize project knowledge into clearly structured, searchable and traceable documents that can be used by both people and Agent. Among them, the current decision-making can be updated and the historical record retained along the timeline (time line) to tell AI “what to do now” and to answer “why did it three months ago”.
The day-to-day needs and discussions are organized into a live page (knowledge page) that is continuously updated, allowing Agent to search for relevant information by task, without having to stuff everything into a big document. By sharing this knowledge, different Coding Age and team members can follow up on the work that has been done to reduce the number of cross-references.
Gaoyang advocates that Agent take more responsibility for documentation and maintenance, that people focus on reviewing needs, decision-making and results, and that the knowledge base be truly the basis for AI to work, rather than an archive that is not updated after the project is completed.
13. AI+ Security Zhang Xinbo, Founder of Corint AI & CEO
Corint AI founder and CEO Zhang Xinbo, shared on the theme “Agent applications and practices in the field of wind control”.
Zhang Shinbo points out that it is only milliseconds to determine whether a transaction is risky and that adjustments to the strategies and models behind it may take days.
So he wants to shorten the period with Agent, linking risk detection, cause analysis, strategy optimization, and then go online through human confirmation.
In order to do this, Agent has to read tables, field and connections in the actual database and, if necessary, understand business meanings in combination with the source code before manually correcting ambiguity. One sentence, "Why did the approval pass rate drop yesterday," may require 10, 20 or even 50 steps, so it is not possible to generate only one SQL, but rather to circulate through Rect, analysis, side-by-side execution, correction of errors, and finally to present the results in graphs.
In the feature project, the risk trail presented by Agent also needs to be calculated and validated, using indicators such as IV, the missing rate, PSI to check the ability to distinguish, the integrity of the data and the stability of their distribution, and to ensure the continuity of long missions through Harness. According to him, in a comparison test with the Kaggle expert team, the system produced 700 effective features in approximately 40 minutes, compared to days of manual processing.
Policy changes are also incorporated into processes developed in similar codes: describe rules in uniform language, submit them to Git, then review them through PR, merge them, and follow each change. He also shared the fact that, after a client had moved the analysis process online, the person in charge had not received a request for data download approval for two weeks in a row, both saving the steps of downloading and processing and reducing data export.
In Zhang ' s view, the value of Agent is to take over complex analysis and implementation and to focus attention on operational judgement and risk management.
14. Co-founder of AI+ Growth, Shanghai Whale Slaughter Artificial Intelligence Ltd.
The co-founder of Shanghai Whale Scrambling Artificial Intelligence Ltd., Sail Soft Software, Vice-President of Strategy Shen Tao, shared the theme on the theme " AI Helping Enterprises Looking for Growth, Business Clients, Efficiency to Profit ".
Shen-tao asked a question enough to impress most business owners: "Ai, how much profit can it bring when employees do 1,000 more slides a year?"
In response, a chemical industry exchange was shared: two hours of R & D, production efficiency was discussed, while rest periods revealed that the chosen technology route may have lost competitiveness, which has increased his focus on one issue:
The wrong direction, AI might just let the business go the wrong way faster.
Shen-tao explained that their teams would use AI to provide a comprehensive analysis of recruitment, patents, bids and financial data from enterprises to identify potential business opportunities. For example, when an enterprise tendered for the purchase of 3150 kVA transformers, while a large number of test and test personnel were recruited, combined with relevant patent information, it could be assumed that it was expanding its capacity and that there was a need for enhanced quality control. This is a customer trail that deserves further follow-up for the equipment detector, data analyser and Manpower Services.
When a manufacturing enterprise explores BMS operations, the team first combs the industrial chain, peers and downstream markets, then screens potential customers and their immediate needs, and gradually puts the search for new growth points on specific clients. He also described AI ' s ability to extract connections between individuals and businesses from information such as activity reports and to help identify possible business referral routes.
In Shentao ' s view, a combination of professional data, business knowledge and AI analysis can provide more targeted support to enterprises in their choice of direction, client identification and business follow-up opportunities. It is also important to see if these leads can be translated into orders and profit.
Two Hands-on events: an idea? Do it!
Two afternoon Hans-on workshops shifted the focus to hands-on practice.
The "AI " community-based weirdo workshop, run in tandem with the "30-second trailer factory", has continued from 13:30 p.m. to 17:00, attracting university students, artists and participants from various fields, including software development, finance and new media.
With your professional experience and your life, you will explore how AI can be integrated into your work, creativity and everyday life.
In the "AI community weirdo workshop", Steerable Developer Wang Tai, a mentor, led participants to try to create interactive little tools, web pages or products through AI Coding.
The starting point for creation is hidden in the respective work and life of the participants: can one more help in everyday decision-making? How can the mission be more tailored to individual needs? Is stock analysis possible with AI?
With these specific issues, you have shared ideas and tried to make ideas that were in your mind gradually a prototype that could be shown and experienced.
On the other hand, the 30-second trailer film factory, which was mentored by the founder of Chiagawa, Cheng Jiawei, was designed to guide the story, storyboard design and AI video generation.
Through AniShort, participants tried to tell their stories with images by turning the stories and images of their minds into original short films or trailers.
People who are not familiar with each other, because teams in the direction of mutual interest complement each other ' s ideas in exchanges, and together discuss how to put creativity on a specific function or shot. Differences in perspectives brought about by different contexts are part of a co-generation on the ground.
In one afternoon, participants tried to make their needs and ideas work, and in their presentations and exchanges, they saw the use of AI in different work and life scenarios.
# Thanks to the partners
1024 The successful organization of the Moot AI Community Days will require strong support from partners.
Thanks to the City of Technology of Purple Harbor for its guidance of the event; to the co-hosting of the Fang Cloud Valley Centre and the TGO Zhangzhou chapter; to the co-organizers of the Hands-on workshop: Zine Cloud and AniShort, for providing LLM Token and video creation credits, respectively, to the participants; and to the special partner, Token Card, for providing live companion gifts. Thanks for getting brain App's 7-day membership card and getting an AI recording card for the value of $888 as an event prize to help you record it efficiently.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.