[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"blog-post-ai-assist-enterprise-software-zh-CN":3,"blog-related-ai-assist-enterprise-software-zh-CN":29,"blog-adjacent-ai-assist-enterprise-software-zh-CN":64},{"id":4,"slug":5,"title":6,"excerpt":7,"content":8,"cover":9,"tone":10,"author":11,"publishedAt":12,"updatedAt":13,"readingMinutes":14,"featured":15,"tags":16,"category":19,"seoSettings":28},5,"ai-assist-enterprise-software","AI 进入企业软件：先从减少重复劳动开始","企业场景不需要“会聊天”的 AI，而需要能稳定分担重复工作、并且结果可追溯的助手。","\u003Cp>AI 是这两年被讨论最多的技术，也是最容易被误用的技术。我们的判断很简单：先找那些重复、耗时、规则清晰的工作，让 AI 去分担，而不是一上来就追求“智能决策”。\u003C\u002Fp>\n\u003Ch2>三个务实的落点\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Cstrong>内容整理：\u003C\u002Fstrong>把会议记录、客户反馈与工单归类成结构化信息，减少人工整理；\u003C\u002Fli>\n\u003Cli>\u003Cstrong>客服辅助：\u003C\u002Fstrong>在知识库基础上生成回复草稿，由人工确认后发送；\u003C\u002Fli>\n\u003Cli>\u003Cstrong>数据查询：\u003C\u002Fstrong>用自然语言完成常见报表查询，降低一线人员的使用门槛。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2>边界与人工确认\u003C\u002Fh2>\n\u003Cp>企业场景对错误的容忍度远低于消费场景。涉及价格、合同、权限与对外承诺的环节，必须保留人工确认；AI 的输出需要可追溯，能说明依据来自哪份资料。把边界写进流程，比事后补救更可靠。\u003C\u002Fp>\n\u003Ch2>如何评估效果\u003C\u002Fh2>\n\u003Cp>不要只看“生成了多少内容”，而要看节省了多少时间、减少多少重复操作、错误率是否下降。我们通常会在小范围试点，用两三周的真实数据判断是否值得扩大。\u003C\u002Fp>\n\u003Cblockquote>AI 在企业软件里的价值，不是替代人，而是把人从重复劳动里释放出来。\u003C\u002Fblockquote>\n\u003Cp>技术会继续演进，但判断标准不会变：是否让业务更稳定、更高效、更容易被一线接受。\u003C\u002Fp>","","teal","佳益方圆团队","2026-09-04T09:30:00+08:00","2026-10-18T15:58:14.704+08:00",1,false,[17,18],"AI 应用","工程实践",{"id":14,"parentId":20,"slug":21,"name":22,"description":23,"icon":24,"cover":9,"tone":25,"kind":26,"sortOrder":27,"count":20},0,"engineering","技术专栏","架构设计、性能优化、AI 集成与跨端方案等一线技术实践，由研发团队持续分享。","cpu","blue","article",60,{},[30,41,53],{"id":31,"slug":32,"title":33,"excerpt":34,"content":9,"cover":9,"tone":35,"author":11,"publishedAt":36,"updatedAt":36,"readingMinutes":14,"featured":15,"tags":37,"category":40},30,"integration-api-pitfalls","企业系统集成：接口对接最容易踩的五个坑","接口文档写得再全，真正对接时仍然会卡在字段约定、超时和重复推送这些细节上。","cyan","2026-10-17T09:00:00+08:00",[38,39,18],"系统集成","接口",{"id":14,"parentId":20,"slug":21,"name":22,"description":23,"icon":24,"cover":9,"tone":25,"kind":26,"sortOrder":27,"count":20},{"id":14,"slug":42,"title":43,"excerpt":44,"content":9,"cover":9,"tone":35,"author":11,"publishedAt":45,"updatedAt":46,"readingMinutes":47,"featured":48,"tags":49,"category":52},"why-enterprise-systems-age","为什么企业系统会在三年后开始“老化”","系统上线时功能齐全，三年后却处处别扭。问题往往不在代码，而在交付时被忽略的那些变化。","2026-10-16T09:30:00+08:00","2026-10-18T15:58:14.648+08:00",2,true,[50,51],"系统演进","架构设计",{"id":14,"parentId":20,"slug":21,"name":22,"description":23,"icon":24,"cover":9,"tone":25,"kind":26,"sortOrder":27,"count":20},{"id":54,"slug":55,"title":56,"excerpt":57,"content":9,"cover":9,"tone":25,"author":11,"publishedAt":58,"updatedAt":58,"readingMinutes":14,"featured":15,"tags":59,"category":63},31,"idempotent-design-basics","幂等设计：重复提交不只是用户手抖","网络重试、消息重投、用户连点，都会让同一个请求到达多次。幂等应该设计在接口里。","2026-10-05T10:13:00+08:00",[60,61,62],"接口设计","幂等","可靠性",{"id":14,"parentId":20,"slug":21,"name":22,"description":23,"icon":24,"cover":9,"tone":25,"kind":26,"sortOrder":27,"count":20},{"previous":65,"next":82},{"id":66,"slug":67,"title":68,"excerpt":69,"content":9,"cover":9,"tone":70,"author":11,"publishedAt":71,"updatedAt":71,"readingMinutes":14,"featured":15,"tags":72,"category":76},92,"first-month-onboarding","新人第一个月怎么过","第一个月的目标不是产出，而是让新人知道遇到问题该找谁。","indigo","2026-09-05T12:39:00+08:00",[73,74,75],"团队文化","入职","带新人",{"id":77,"parentId":20,"slug":78,"name":79,"description":80,"icon":81,"cover":9,"tone":25,"kind":26,"sortOrder":31,"count":20},4,"team-culture","团队动态","公司活动、技术分享、团建与招聘文化，了解佳益方圆团队的日常。","users",{"id":83,"slug":84,"title":85,"excerpt":86,"content":9,"cover":9,"tone":70,"author":11,"publishedAt":87,"updatedAt":87,"readingMinutes":14,"featured":15,"tags":88,"category":92},111,"ai-adoption-reality","AI 落地的现实：从演示到上线","AI 演示的效果经常很惊艳，上线之后的使用率却很低。差距在哪里？","2026-09-03T12:39:00+08:00",[89,90,91],"行业观察","人工智能","落地",{"id":4,"parentId":20,"slug":93,"name":89,"description":94,"icon":95,"cover":9,"tone":25,"kind":26,"sortOrder":96,"count":20},"industry-watch","对互联网、软件与数字化行业趋势的深度分析，输出团队自己的判断。","globe",20]