广告
加载中

李宁、维密等大牌为何集中签约AI店小蜜?AI客服为电商带来经营增长确定性

天下网商 2026-07-21 12:04
天下网商 2026/07/21 12:04

邦小白快读

EN
全文速览

本文核心介绍了淘天推出的新版AI店小蜜,已经被李宁、维密、鱼跃等多个大牌集中签约,为电商服务和经营带来了本质改变,核心干货如下

1. 相较于传统智能客服,新版AI店小蜜依托大模型能力,可完成从售前咨询导购到售后挽单、退款处理的全链路服务,能结合上下文、用户实际需求做千人千面的个性化推荐,主动计算优惠、推荐替代款,比传统客服更灵活精准。

2. 从实际落地效果看,AI店小蜜帮品牌大幅降本提效:李宁大促客服准备时间从1-2周缩短到两三天,整体工作量减少六成,售前转人工率从约40%降至20%以内,鱼跃夜间询单转化率接近翻倍,部分品牌退款率最高下降超过6%。

3. 新版AI店小蜜清晰划分了人机分工边界,AI处理标准化问题,高风险、复杂需求会自动转人工,既提升了效率,也能保障消费者的服务质量。

新版AI店小蜜为品牌降本增效、实现经营增长提供了可行路径,符合当前电商品牌拥抱AI的趋势,核心干货如下

1. 效率与增长层面:AI店小蜜可将大促客服准备工作量减少六成以上,自动消化大促进线流量压力,不需要品牌额外增派人手做专项备战,还能直接带动经营结果提升,既降低转人工率,又能提升售前转化率,最高可帮助品牌降低退款率超6个百分点,让客服从成本中心变成增长引擎。

2. 用户服务层面:AI可规模化复制品牌验证过的导购策略和经营方法,实现千人千面的顾问式导购,品牌还可根据自身品类特点自主设定AI与人工的分工边界,兼顾服务效率和风险管控,适配不同品牌的需求。

3. 产品研发层面:AI可自动聚类用户咨询、退款中的痛点问题,快速输出用户洞察,帮助商品团队及时优化产品和详情页,比如李宁优化跑鞋鞋舌问题后,相关咨询量下降60%,退货也同步减少。目前拥抱AI客服是确定性趋势,头部品牌已经开始集团级布局,提前接入可更早积累数据获得先发优势。

对于电商卖家来说,AI客服已经从效率工具升级为增长工具,当前布局有明确机会也有需要注意的风险,核心干货如下

1. 机会层面:新版AI店小蜜高阶版已经实现从接住咨询到接住成交的升级,可帮助卖家降低大促备战工作量、消化大促流量压力,还能直接提升售前转化率、降低退款率,效果已经被李宁、维密、鱼跃等多个大牌验证,拥抱AI是当前电商行业确定性的商业增长趋势。

2. 灵活适配层面:AI店小蜜支持卖家自定义导购逻辑、转人工边界,可适配不同品类的经营需求,比如医疗器械可设置“宁可多转不可漏转”的规则,内衣可设置隐私风险规避规则,灵活性很强。

3. 风险提示:对于观望的卖家来说,犹豫的成本大于试错成本,AI店小蜜的产品能力可以后续购买,但把平台能力转化为自身经营能力、积累真实会话数据需要时间,晚入场会错过先发优势;同时高阶版AI店小蜜价格高于传统客服机器人,大模型使用有Token成本,卖家需要结合自身规模和需求选择。

对于布局电商的工厂来说,AI店小蜜的普及给工厂数字化转型、产品优化带来了新的机会和启示,核心干货如下

1. 产品生产设计端:AI可自动聚类用户咨询、评价、退款中的用户反馈,快速识别用户对产品的痛点需求,把原来分散在海量对话中的用户声音,整理成可供生产设计团队参考的清晰信息,过去工厂从问题出现到发现问题通常需要2-3周,现在可缩短到2-3天,能帮助工厂更快调整产品设计、优化详情页,减少后续退货,提升用户满意度。

2. 电商经营端:AI可以帮助工厂降低客服人力成本,提升大促接待能力,不需要大促临时扩招客服,还能提升导购转化率、降低退款率,直接带动经营增长。

3. 数字化转型启示:AI可以帮助工厂把成熟的销售服务经验规模化复制,解放出人力投入到售后回访、高价值用户服务等之前没时间开展的工作,工厂提前布局AI就能更早积累商品知识和会话数据,形成自身的数字化经营优势。

当前电商客服行业正在发生根本性变革,大模型赋能下的AI客服成为新的增长方向,核心干货如下

1. 行业发展趋势:行业对AI客服的价值拷问已经从“AI能不能回答客户问题”转变为“AI能不能带来生意结果”,AI客服已经不再只是服务部门的效率工具,变成了可以直接影响品牌经营结果的增长引擎,越来越多品牌愿意为能带来经营增量的AI客服付费,头部品牌已经开始集团级签约布局,市场需求明确。

2. 品牌客户的核心痛点:传统智能客服只能完成标准问答,无法结合上下文、用户画像、商品信息做个性化导购,也无法延伸到售后挽单、业务办理环节,同时大促备战成本高,用户反馈分散无法沉淀利用。

3. 可参考的解决方案:AI店小蜜的模式值得参考,以大模型为底座,接入平台交易数据,支持客户自定义经营策略、导购逻辑和转人工边界,明确人机分工,同时沉淀用户反馈反哺产品运营,可为客户带来降本增效和经营增长的双重价值。

从AI店小蜜的落地可以看出商家对平台AI工具的核心需求,也总结出可参考的平台运营和产品经验,核心干货如下

1. 商家核心需求:商家需要的不只是能回答问题的AI客服工具,而是能结合平台交易数据(用户画像、商品信息、订单优惠等)、商家自身经营策略,完成从售前导购到售后处理全链路服务,能直接带来经营增长的AI工具。

2. 产品设计经验:产品可分版本分层满足不同商家需求,比如分为标准版和高阶版,同时要开放商家自定义能力,允许商家根据自身品类特点设置AI能力边界、转人工规则,适配不同行业的风险要求,还要增加多道内容校验、安全监控机制,保障输出符合监管要求。

3. 招商运营经验:AI客服的增长价值已经被头部品牌验证,可推出集团级签约服务覆盖商家多品牌多店铺,同时要向商家传递先发优势:越早接入越早积累真实会话和经营数据,能力迭代更快,可吸引商家尽早布局,同时要明确人机分工的价值,帮助商家真正拿到可量化的经营结果。

本文展现了大模型落地电商客服领域的最新产业动向,呈现了很多值得研究的新变化和新启示,核心干货如下

1. 产业新动向:大模型已经从概念落地到电商客服的真实经营场景,AI客服的价值定位发生了根本性变化,从原来的降本工具变成带动经营增长的引擎,服务链路从单纯回答问题延伸到全链路经营,覆盖售前导购、售后挽单、用户洞察反哺产品等多个环节,头部品牌已经批量布局,成为确定性的产业发展趋势。

2. 新的分工与商业模式:探索出了清晰的人机协同分工模式:AI规模化复制品牌经过验证的经营方法,处理标准化问题,人工聚焦情绪价值、复杂问题和高风险场景,形成了新的服务分工;同时推出分版本定价模式,高阶版针对有增长需求的品牌收取更高费用,匹配价值定价。

3. 价值重构的新方向:客服渠道通过AI的聚类分析,已经变成可实时输出用户洞察的VOC来源,反向赋能产品、运营、供应链环节,推动品牌将“客户服务中心”改为“客户运营中心”,重构了客服部门的产业价值。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article introduces Taobao Tmall Group's new large language model-powered AI Dianxiaomi (store assistant), which has already been adopted by major brands including Li-Ning, Victoria's Secret and Yuwell, bringing transformative changes to e-commerce service and operations. Key takeaways are as follows:

1. Compared with traditional intelligent customer service, the new AI Dianxiaomi leverages large model capabilities to deliver end-to-end services spanning pre-sales consultation and product recommendation, post-sales order retention and refund processing. It can deliver personalized, context-aware recommendations tailored to individual user needs, automatically calculate discounts and suggest alternative products, outperforming traditional customer service in flexibility and accuracy.

2. Real-world deployment results show the AI assistant helps brands achieve significant cost reduction and efficiency gains: For Li-Ning, customer service preparation time for major promotions has been cut from 1-2 weeks to 2-3 days, with overall workload reduced by 60%, and the pre-sales human transfer rate dropped from around 40% to under 20%. Yuwell saw its nighttime order conversion rate nearly double, and for some brands, the refund rate has dropped by more than 6 percentage points at maximum.

3. The new AI Dianxiaomi establishes a clear division of labor between AI and human agents: AI handles standardized queries, while high-risk and complex requests are automatically routed to human staff, boosting efficiency while maintaining consumer service quality.

The new AI Dianxiaomi offers a viable path for brands to cut costs, boost efficiency and drive business growth, aligning with the current trend of e-commerce brands embracing AI. Key takeaways are as follows:

1. Efficiency and growth: AI Dianxiaomi can reduce brand customer service preparation workload for major promotions by more than 60%, automatically absorb peak traffic during sales events without requiring brands to deploy extra staff for preparation. It also directly improves business outcomes, cutting human transfer rates, boosting pre-sales conversion, and reducing refund rates by up to more than 6 percentage points, transforming customer service from a cost center into a growth engine.

2. User service: The AI can scale and replicate brands' validated sales guidance strategies and operational methods to deliver personalized consultant-level service at scale. Brands can also customize the division of labor between AI and human agents based on their own category characteristics, balancing service efficiency and risk control to meet the specific needs of different brands.

3. Product development: The AI can automatically cluster pain points from user consultations and refund requests, and quickly generate actionable user insights to help product teams optimize products and product detail pages in a timely manner. For example, after Li-Ning adjusted the design of its running shoe tongue based on AI-generated insights, related consultations dropped by 60% alongside a reduction in returns. Adopting AI-powered customer service is now a clear industry trend, and leading brands have already rolled out group-wide deployments. Early adoption allows brands to accumulate data sooner and gain a first-mover advantage.

For e-commerce sellers, AI customer service has evolved from an efficiency tool to a growth tool, and early deployment offers clear opportunities along with notable risks to consider. Key takeaways are as follows:

1. Opportunities: The premium tier of the new AI Dianxiaomi has upgraded from simply answering queries to facilitating full transaction journeys. It helps sellers cut promotion preparation workload, absorb peak traffic during sales events, directly boost pre-sales conversion and reduce refund rates — results already validated by leading brands including Li-Ning, Victoria's Secret and Yuwell. Embracing AI is now a clear, high-certainty trend for business growth in the e-commerce industry.

2. Flexible adaptation: AI Dianxiaomi supports sellers in customizing recommendation logic and human transfer boundaries to adapt to operational needs of different product categories. For example, medical device sellers can set a "better to transfer extra than miss any risk" rule, while lingerie brands can set rules to avoid privacy risks, making the tool highly flexible.

3. Risk note: For sellers still on the fence, the cost of hesitation outweighs the cost of testing. While the product capabilities of AI Dianxiaomi can be purchased at any time, converting platform capabilities into in-house operational capabilities and accumulating real conversation data takes time. Late entry means missing out on first-mover advantages. Additionally, the premium tier of AI Dianxiaomi is priced higher than traditional customer service robots, and large model inference incurs token costs, so sellers should choose the right option based on their own scale and needs.

For factories with direct e-commerce operations, the widespread adoption of AI Dianxiaomi brings new opportunities and insights for digital transformation and product optimization. Key takeaways are as follows:

1. Product design and manufacturing: The AI can automatically cluster user feedback from consultations, reviews and refund requests, and quickly identify user pain points. It organizes fragmented user voices from massive amounts of conversation into clear, actionable insights for production and design teams. Whereas factories previously needed 2-3 weeks to identify emerging product issues, the process can now be completed in 2-3 days, helping factories adjust product designs and optimize product detail pages faster, reduce subsequent returns and improve user satisfaction.

2. E-commerce operations: The AI helps factories cut customer service labor costs, improve peak-period service capacity during promotions, eliminating the need to hire temporary customer service staff for sales events. It also boosts sales conversion and reduces refund rates, directly driving business growth.

3. Insights for digital transformation: AI helps factories scale and replicate their proven sales and service experience, freeing up staff to focus on previously underprioritized high-value work such as post-sales follow-ups and premium user service. Early AI deployment allows factories to accumulate product knowledge and conversation data sooner, building up proprietary advantages in digital operation.

The e-commerce customer service industry is undergoing a fundamental transformation, with large model-powered AI customer service emerging as a new growth direction. Key takeaways are as follows:

1. Industry development trend: The core question around AI customer service has shifted from "can AI answer customer questions" to "can AI deliver business results". AI customer service is no longer just an efficiency tool for service departments, but has become a growth engine that directly impacts brand business outcomes. A growing number of brands are willing to pay for AI customer service that delivers incremental business value, and leading brands have already signed group-wide deployment agreements, indicating clear market demand.

2. Core pain points for brand clients: Traditional intelligent customer service only handles standardized question-and-answer tasks, and cannot deliver personalized recommendations that integrate context, user profiles and product information. It also cannot extend to post-sales order retention and core business processing. In addition, traditional solutions incur high preparation costs for major promotions, and user feedback is scattered and cannot be aggregated and leveraged.

3. A referenceable solution: The AI Dianxiaomi model offers a useful blueprint. Built on a large model foundation and integrated with platform transaction data, it supports clients in customizing operational strategies, recommendation logic and human transfer boundaries, with a clear division of labor between human and AI. It also aggregates user feedback to inform product and operations improvements, delivering dual value of cost reduction, efficiency improvement and business growth for clients.

The deployment of AI Dianxiaomi reveals the core demands of merchants for platform AI tools, and offers actionable insights for platform product development and operation. Key takeaways are as follows:

1. Core merchant demand: What merchants need is not just an AI customer service tool that can answer questions, but an AI tool that integrates platform transaction data (user profiles, product information, order discounts, etc.) and merchants' own operational strategies to deliver end-to-end services from pre-sales recommendation to post-sales processing, and directly drive business growth.

2. Product design lessons: Products should be tiered to meet the needs of different merchants, for example through separate standard and premium tiers. Platforms should also open up customization capabilities, allowing merchants to set AI capability boundaries and human transfer rules based on their own category characteristics to align with risk requirements across different industries. Multiple layers of content review and security monitoring should also be added to ensure outputs comply with regulatory requirements.

3. Business development and operation lessons: The growth value of AI customer service has been validated by leading brands. Platforms can offer group-wide signing services to cover multiple brands and stores under the same enterprise, and communicate the first-mover advantage to merchants: earlier adoption leads to faster accumulation of real conversation and operational data, and faster capability iteration, which incentivizes early deployment. Platforms should also clearly communicate the value of clear human-AI division of labor to help merchants achieve measurable business outcomes.

This article presents the latest industry developments of large model deployment in e-commerce customer service, highlighting a number of noteworthy new changes and insights. Key takeaways are as follows:

1. New industry trends: Large models have moved beyond concept to real-world deployment in e-commerce customer service operations, driving a fundamental shift in the value proposition of AI customer service: it has evolved from a cost-cutting tool to an engine for business growth. Its service scope has expanded from simply answering questions to end-to-end operational support, covering pre-sales recommendation, post-sales order retention, and user insights that inform product improvement. Leading brands have already rolled out large-scale deployments, making this a high-certainty industry development trend.

2. New division of labor and business model: The industry has developed a clear human-AI collaborative division of labor: AI scales and replicates brands' validated operational methods to handle standardized tasks, while human staff focus on emotional support, complex issues and high-risk scenarios, forming a new service分工 structure. A tiered pricing model has also been introduced, where the premium tier charges higher fees for brands with growth needs, aligning price with delivered value.

3. New direction for value restructuring: Powered by AI clustering analysis, the customer service channel has now become a real-time source of Voice of Customer (VoC) insights that反向 empowers product, operations and supply chain teams. This has pushed brands to restructure their customer service departments from "cost centers" to "customer operation centers", redefining the industrial value of the customer service function.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

从接住咨询,到提升成交。

|沈嵩男

曾经每逢大促前夕,李宁的智能客服运营团队有一项固定工作:拆解活动规则。

定金什么时候退,满减如何叠加,会员券能否与积分同时使用……每项规则都要被改写成问答,再预判消费者可能使用的几十种表达,配置关键词,反复测试。如果规则有调整,相关话术也要跟着改。

这套准备通常要提前一至两周。进入大促前,团队几乎全员扑在知识库上。

现在,他们只需要用完整的活动规则、玩法说明和权益文档训练AI店小蜜,再抽检少数核心场景。准备时间缩短到两三天,整体工作量减少六成以上。

变化不只发生在准备阶段。

过去,在大促开门红期间,客服后台的排队提示不断跳红;如今,同一时段的排队时长已接近日常高峰时的水平。一线客服不必为了追赶进线速度而匆忙回复,可以把时间留给更“非标”的售前疑难杂症。

李宁内部将这种变化概括为:不是靠堆人把流量硬扛过去,而是让AI把压力真正消化掉。大促应对也从一场需要额外增派人力、扩充知识库和安排排班的“专项备战”,逐渐变成日常服务能力的一次常规运转。

2026年5月11日,AI店小蜜发布新版本,产品分为标准版和高阶版。其中,高阶版重点增强售前转化能力,将用户画像、商品理解、行业导购策略和商家自定义能力带入会话,能够主动推荐、比较商品、计算优惠并推动成交;售后侧则从回答规则进一步延伸到办理业务,开始处理退差价、退款协商、挽单和多轮排障等事宜。

大模型每一次理解、推理和生成,都伴随实际的Token消耗,规模扩大也无法像传统软件一样将边际成本充分摊薄。所以商家们愿意采用它,并不是因为“便宜”,而是传统知识库和简单的RAG难以同时调用用户特征、商品信息、上下文、优惠权益和导购策略,更难据此自主判断下一步应该作答、推荐还是执行操作。淘天集团客户运营部商家服务体验及AI店小蜜运营负责人袁雷(花名宗布)认为,“以前的客服系统,是把一个有个性、有温度的人训练成严格执行标准话术的机器;现在,是把人的经营思路教给AI,由AI完成精准回复,让真人客服提供情绪价值和高品质的个性化服务。”

行业对AI客服价值的拷问,也从“AI能不能回答客户问题”变为“AI能不能带来生意结果”。

“但犹豫的成本,可能大于试错的成本”,袁雷坦言,AI店小蜜高阶版的价格会高于传统知识库回复的机器人的报价,但他也观察到,拥抱AI已经是一个确定性的商业趋势。第一批签约的品牌中,不少是集团级签约,覆盖旗下多个品牌、多家店铺,他们看重的是经营增量的确定性。

从部分接入高阶版能力的品牌近两个月的数据来看——李宁售前转人工率由约40%降至20%以内,鱼跃夜间询单转化率接近翻倍……

这些数字共同指向一个变化:AI正在从接住咨询,走向接住成交。所以对大部分商家来说,真正的问题是:当越来越多同行正在把客服变成增长引擎时,犹豫的成本究竟有多高?

一次客服会话被拉长成一条经营链路

传统智能客服的任务通常结束于“给出答案”:

用户问尺码,系统返回尺码表;询问活动,系统调取活动规则;说商品没货,系统回复暂时无库存。它回答了问题,却很少继续追问:这个人真正想买什么,顾虑是什么,还有没有其他成交机会。

AI店小蜜高阶版将一次会话的任务链向后拉长了。

在李宁,消费者可能会说自己平时穿42码皮鞋、脚型偏宽,再询问某款跑鞋应该选什么尺码。AI会记住前文信息,结合鞋楦、商品尺码和使用场景给出建议。面对满减、红包、会员权益叠加等问题,它也不再原样转发规则,而要把到手价和操作路径说清楚。

但消费者并不总会准确表达需求——在维密,一名消费者可能正在浏览有钢圈内衣,却只说自己想要“舒服一点”“不要勒”。店小蜜需要先识别当前商品与真实需求的错位,再推荐更合适的无钢圈产品;商品无货时,它也不再停在“暂无库存”,而是继续询问版型、杯型和穿着偏好,寻找替代款。这种结合上下文、当前商品和消费者实际偏好的动态回复,也让AI客服从标准问答进一步走向千人千面的个性化服务。

这种能力并非意味着让AI自由发挥。维密的搭配关系仍由运营团队配置,鱼跃的医疗器械推荐逻辑则由客服团队与产品团队共同梳理。品牌先明确什么人适合什么商品、应当追问哪些信息、哪些表达必须避免,再由AI将策略带入每一通对话。换言之,AI规模化的不是某一句话术,而是品牌已经验证过的经营方法。据悉,目前AI店小蜜高阶版已经支持品牌根据不同品类和客群的需求,封装不同的客服Skill,用于与消费者沟通,提升AI客服整体的精细化接待能力。

在鱼跃,当消费者说“家里老人需要一台制氧机”,AI会依次询问使用目的、身体状况和使用环境,再根据医疗需求给出两三款产品及推荐理由。过去的客服话术往往先推热销款,现在则是先问需求,再给方案——从推销式导购转向顾问式服务。

这条链路甚至还延伸到了退款之后。

过去,用户发起退款,自动客服通常只负责解释规则或提供入口。现在,AI开始根据退款原因和品牌策略判断:这笔订单是否存在挽回的可能,应当直接退款,还是可以通过补偿、换货或其他方案解决问题。

比如,一家运动户外品牌在业务A/B实验中,退款率下降4个百分点;在某服饰品牌的退款场景中,AI介入后退款率下降超过6%……当一个客服系统开始影响转化率和退款率,它便不再只是服务部门的效率工具,而开始直接影响品牌的经营结果。

降低转人工率,也要守住服务边界

降低转人工率,是AI店小蜜提升接待效率的重要结果;与此同时,商家也可以根据自身行业、商品和服务要求,自主定义AI与人工介入的边界。

关于何时转人工,鱼跃为店小蜜设置的原则是:“宁可多转,不可漏转。”

用户说“制氧机报警”,AI会先确认指示灯颜色和蜂鸣方式,再询问机器是否出氧、环境温度以及近期是否移动过设备,最后初步判断故障原因,并确定是过滤网堵塞、散热异常,还是需要进一步检修。

同类问题过去由人工处理,平均需要12至15分钟。AI通常经过三至四轮交互,基本可在2分钟内定位问题并给出初步方案,完整解决问题的平均时长约4分钟;即使最终仍需转人工,AI也已提前收集好设备和使用环境信息,人工不必从头再问。

但医疗器械不能猜。

若出现异响等可能涉及安全风险的情况,必须立即转人工;用户连续表示问题没有解决,必须转人工;模型判断的置信度不足,不允许勉强作答;涉及投诉、退换货权限和医疗专业判断,也要交给人工。对于用氧安全等高风险问题,AI回答的同时还会向人工坐席发出预警。

除了商家自行配置的转人工规则,AI店小蜜还通过多道内容校验、安全模式和辅助监控机制,保障输出符合监管要求和商家的服务要求。

真正成熟的AI,不只是能够解决更多问题,还要知道什么时候停止处理。“要给AI划一条清楚的能力边界。如果问题已经超出它的能力,或者我们判断由人工接待效果更好,那就直接转给人工。”袁雷补充道,这正是他理解的AI接待与人工介入之间的边界。

维密的边界,则在于隐私与冒犯风险。

当消费者完整提供上、下胸围、杯型和穿着偏好时,店小蜜可以给出准确的尺码建议。但内衣尺码还受到版型、舒适度和个人习惯影响,信息不足时仍可能误判。品牌因此把“不越界、不冒犯、不做不当引导”,放在导购效率之前。

李宁的边界藏在商品差异中。

鞋类的尺码、鞋楦、材质和科技参数相对客观,AI通常能快速达到较高准确度;服装中的“会不会显胖”“面料是否闷热”“穿起来软不软”,没有唯一答案,需要持续用优质的人工对话训练AI的表达能力。

AI能够越来越像人一样说话,却仍很难实现真正的情绪反转。复杂投诉、突发风险、深度专业判断和高价值用户关系维护,依然需要经验丰富的人工客服。

这不是AI能力不足后的被动兜底,而是一种新的分工。

客服不再只对一通会话负责

在维密,客服过去约70%的时间都用于处理查订单、问物流、解释活动规则等标准问题。AI接手后,人工转向处理复杂售后、高价值用户和敏感场景,同时把自己的优质对话、商品经验和服务判断用于训练模型。

客服的角色从“服务好自己接待的用户”,转向“帮助机器服务好更多人”。

这种人机协同还在向服务质检延伸。过去,品牌主要依靠满意度和小比例抽检判断服务质量;现在,AI开始对更大范围的会话进行诊断,不只识别错答和违规,也识别回避问题、没有提供实质解决方案、专业度不足等“没有出错但服务不好”的情况。优秀客服的经验可以被沉淀为AI训练素材,AI识别出的低效服务又能反向用于客服培训,推动服务评价从满意度进一步走向“好服务率”。

在鱼跃,释放出来的人力被投入到制氧机、呼吸机的售后回访,以及长期用户的耗材补货提醒和VIP服务中。过去,这些事一直被认为有价值,却因日常咨询过多而无法真正展开。

李宁的团队则从用户咨询中发现,消费者频繁询问某款跑鞋的鞋舌是否容易跑偏。商品团队随后在详情页增加结构说明和图解,这一问题的咨询量下降60%,相关退货也随之减少。

类似的经营反馈也开始从依赖客服主观感受,转向由AI主动识别。鱼跃一款血压计有关“包装体验”的咨询出现异常增长时,系统当天就能标记异常;过去从问题出现到团队真正关注,通常需要2至3周,现在缩短到了2至3天。AI对咨询、评价和退款原因进行自动聚类后,客服渠道也开始成为商品、运营和品控团队可及时利用的VOC洞察来源。

如今,越来越多品牌正在将“客户服务中心”改称“客户运营中心”。消费者为什么犹豫、最关心什么功能、哪些问题正在推高退款率——每一次咨询,都是一份未经修饰的经营反馈。过去,这些信息分散在海量对话中,难以被系统利用;如今,AI开始将其反馈给商品、运营、供应链和品控团队。

据AI店小蜜团队介绍,AI店小蜜以千问全系大模型为底座,针对客服场景进行专属训练和蒸馏;同时接入淘系平台内的用户画像、商品、订单、优惠和履约能力,再叠加商家的专属知识库与导购Skill。AI店小蜜能够把模型能力、淘天的交易体系和品牌自己的经营策略,放进同一次会话中,这使它不只是回答问题,还能判断下一步应该推荐什么、提供什么权益,或直接办理什么业务。因此,率先接入的品牌也能更早积累商品知识、导购策略和真实会话数据。

下一场大促到来时,一些品牌使用的已经是一套经历过真实流量和售后周期的AI客服;另一些品牌可能才刚开始第一轮训练。对仍在观望的商家而言,产品能力可以后续购买,但将淘天的平台能力转化为品牌自身经营能力所需要的时间成本,却无法一次性弥补。

注:文/天下网商,文章来源:天下网商(公众号ID:txws_txws),本文为作者独立观点,不代表亿邦动力立场。

文章来源:天下网商

广告
微信
朋友圈

FAQ回顾

AI店小蜜高阶版有哪些核心功能?

2026年5月11日发布的AI店小蜜高阶版,售前侧可结合用户画像、商品理解、行业导购策略主动推荐商品、计算优惠推动成交;售后侧可处理退差价、退款协商、挽单、多轮排障等业务,还支持商家自定义客服Skill。

AI客服能为电商商家带来哪些实际价值?

AI客服可将电商大促前的规则配置准备时间从1-2周缩短至2-3天,整体工作量减少六成以上;还可降低售前转人工率、提升询单转化率、降低退款率,同时聚类消费者咨询反馈为商品、运营等团队提供经营洞察,助力商家获得增长确定性。

电商场景下AI客服和人工客服如何分工?

AI客服主要承接标准咨询、优惠计算、基础故障排查、常规退改办理等标准化服务,人工客服则负责复杂售后、高价值用户服务、风险场景处理、专业判断类需求,同时人工的优质对话经验可用于训练AI,形成人机协同的正向循环。

这么好看,分享一下?

朋友圈 分享

APP内打开

+1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0