[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-enterprise-agent-first-step":3,"news-related-enterprise-agent-first-step":206},{"id":4,"title":5,"author":6,"body":7,"category":180,"cover":181,"date":182,"description":183,"draft":184,"extension":185,"keywords":186,"meta":191,"navigation":192,"path":193,"products":194,"readingMinutes":196,"seo":197,"sitemap":198,"stem":199,"summary":200,"updated":204,"__hash__":205},"news\u002Fnews\u002Fenterprise-agent-first-step.md","企业智能体落地第一步：从一个高频问答场景开始","赵金添",{"type":8,"value":9,"toc":167},"minimark",[10,14,18,21,44,47,51,118,122,143,146,149,152,157,160,164],[11,12,13],"p",{},"企业智能体（Enterprise AI Agent）是在大语言模型的基础上，接入企业知识库和业务系统，理解企业内部信息、按流程调用工具并完成具体任务的 AI 助手。与通用聊天机器人不同，企业智能体回答的是“我们公司的规定是什么”“这台设备报警该怎么处理”这类企业内部问题。",[15,16,17],"h2",{"id":17},"为什么从问答场景开始",[11,19,20],{},"很多企业一开始就希望智能体“全自动处理业务”，但实践中，最容易落地的第一步往往是内部知识问答：",[22,23,24,32,38],"ul",{},[25,26,27,31],"li",{},[28,29,30],"strong",{},"高频","：新员工培训、制度咨询、产品参数查询每天都在发生；",[25,33,34,37],{},[28,35,36],{},"答案明确","：知识来自已有文档，便于校验准确率；",[25,39,40,43],{},[28,41,42],{},"效果可衡量","：可以直接统计节省的人工答疑时间和准确率。",[11,45,46],{},"我们服务的一家连锁餐饮品牌从“门店带教问答”切入，新员工独立上岗时间从约 15 天缩短到约 6 天。",[15,48,50],{"id":49},"_4-周试点节奏","4 周试点节奏",[52,53,54,70],"table",{},[55,56,57],"thead",{},[58,59,60,64,67],"tr",{},[61,62,63],"th",{},"周次",[61,65,66],{},"主要工作",[61,68,69],{},"产出",[71,72,73,85,96,107],"tbody",{},[58,74,75,79,82],{},[76,77,78],"td",{},"第 1 周",[76,80,81],{},"收集并清洗文档，按岗位和场景分类",[76,83,84],{},"结构化知识库",[58,86,87,90,93],{},[76,88,89],{},"第 2 周",[76,91,92],{},"搭建智能体，调优检索和回答格式",[76,94,95],{},"内部测试版本",[58,97,98,101,104],{},[76,99,100],{},"第 3 周",[76,102,103],{},"选择 1～2 个部门试点，收集反馈",[76,105,106],{},"问题清单与准确率报告",[58,108,109,112,115],{},[76,110,111],{},"第 4 周",[76,113,114],{},"优化后扩大范围，建立更新机制",[76,116,117],{},"正式上线与运营规范",[15,119,121],{"id":120},"知识库准备的-3-个要点","知识库准备的 3 个要点",[123,124,125,131,137],"ol",{},[25,126,127,130],{},[28,128,129],{},"先治理再导入","：过期制度、重复文档要先清理，否则智能体会给出互相矛盾的答案。",[25,132,133,136],{},[28,134,135],{},"答案标注出处","：每条回答附带文档来源，员工可以点开核对，建立信任。",[25,138,139,142],{},[28,140,141],{},"指定知识负责人","：每个知识分类明确一位更新负责人，制度变化时及时同步。",[15,144,145],{"id":145},"从问答走向任务",[11,147,148],{},"问答场景跑通后，可以逐步让智能体调用业务系统：查询订单状态、生成报表、创建工单、发送审批。智侠支持通过 MCP 等标准协议连接企业现有系统，并支持私有化部署，数据不出企业内网。",[15,150,151],{"id":151},"常见问题",[153,154,156],"h3",{"id":155},"企业数据交给大模型安全吗","企业数据交给大模型安全吗？",[11,158,159],{},"可以选择私有化部署或使用国内合规的大模型服务，知识库存储在企业自己的服务器上，并按部门和角色控制可访问的知识范围。",[153,161,163],{"id":162},"智能体回答错了怎么办","智能体回答错了怎么办？",[11,165,166],{},"通过“答案标注出处 + 用户反馈 + 定期复盘”三道机制持续优化。试点阶段建议对高风险问题设置人工复核。",{"title":168,"searchDepth":169,"depth":169,"links":170},"",3,[171,173,174,175,176],{"id":17,"depth":172,"text":17},2,{"id":49,"depth":172,"text":50},{"id":120,"depth":172,"text":121},{"id":145,"depth":172,"text":145},{"id":151,"depth":172,"text":151,"children":177},[178,179],{"id":155,"depth":169,"text":156},{"id":162,"depth":169,"text":163},"企业智能体","\u002Fimages\u002Fnews\u002Fenterprise-agent-first-step.webp","2026-07-05","企业智能体是能理解企业知识、调用业务系统并完成任务的 AI 助手。本文介绍中小企业如何选择首个场景、准备知识库并在 4 周内完成试点上线。",false,"md",[180,187,188,189,190],"AI Agent","知识库","私有化部署","智侠",{},true,"\u002Fnews\u002Fenterprise-agent-first-step",[195],"zhixia",7,{"title":5,"description":183},{"loc":193},"news\u002Fenterprise-agent-first-step",[201,202,203],"企业智能体 = 大模型 + 企业知识库 + 业务系统工具调用。","第一个场景应选择高频、标准答案明确、容易衡量效果的问答类工作。","4 周试点节奏：整理知识、搭建调优、小范围试点、推广复盘。",null,"L1YgkAuX07Lm5ulrIxGHZC0qxmkA3qMk71oJaFP4lm0",[207,215,222,228,235],{"path":208,"title":209,"description":210,"category":211,"date":212,"cover":213,"readingMinutes":214},"\u002Fnews\u002Fanti-association-guide","跨境多账号防关联指南：浏览器指纹、IP 与操作习惯三件事","跨境店铺关联是指平台通过设备指纹、网络 IP 和操作行为判断多个账号属于同一主体。本文讲清关联原理、常见误区和用指纹浏览器做环境隔离的落地步骤。","跨境运营","2026-09-12","\u002Fimages\u002Fnews\u002Fanti-association-guide.webp",8,{"path":216,"title":217,"description":218,"category":219,"date":220,"cover":221,"readingMinutes":196},"\u002Fnews\u002Fcross-border-rpa-scenarios","跨境电商 RPA 自动化：最值得先做的 8 个场景","跨境电商 RPA 是用软件机器人模拟人工在店铺后台的重复操作。本文按投入产出比整理 8 个最值得优先自动化的场景，并给出评估方法与落地建议。","RPA 自动化","2026-08-26","\u002Fimages\u002Fnews\u002Fcross-border-rpa-scenarios.webp",{"path":223,"title":224,"description":225,"category":211,"date":226,"cover":227,"readingMinutes":196},"\u002Fnews\u002Ftiktok-matrix-operation","TikTok 矩阵运营实战：从 10 个账号到 100 个账号要解决的 4 个问题","TikTok 矩阵运营是用多个定位互补的账号协同获取流量。本文总结账号规模化过程中的环境、养号、内容和数据四个关键问题与解决方法。","2026-08-08","\u002Fimages\u002Fnews\u002Ftiktok-matrix-operation.webp",{"path":229,"title":230,"description":231,"category":211,"date":232,"cover":233,"readingMinutes":234},"\u002Fnews\u002Fsku-profit-analysis","多店铺 SKU 利润分析：跨境卖家算清每一分钱的 5 个步骤","SKU 利润分析是把收入和各项成本精确分摊到单个商品，计算真实利润。本文给出跨境多平台卖家统一口径、分摊费用和建立预警的 5 个步骤。","2026-07-22","\u002Fimages\u002Fnews\u002Fsku-profit-analysis.webp",6,{"path":236,"title":237,"description":238,"category":211,"date":239,"cover":240,"readingMinutes":234},"\u002Fnews\u002Faigc-cross-border-content","AIGC 在跨境电商内容生产中的 5 个实用场景","AIGC 指利用生成式人工智能自动生成图片、文案和视频。本文介绍跨境卖家在商品图、多语言详情页、广告素材等 5 个场景的落地方法与质量把控要点。","2026-06-16","\u002Fimages\u002Fnews\u002Faigc-cross-border-content.webp"]