«

亚马逊发布卖家平台自主AI工具,正值监管审查趋严之际

qimuai 发布于 阅读:8 一手编译


亚马逊发布卖家平台自主AI工具,正值监管审查趋严之际

内容来源:https://www.geekwire.com/2025/amazon-unveils-new-agentic-ai-tools-for-its-crucial-seller-marketplace-amid-heightened-scrutiny/

内容总结:

亚马逊在日前举行的年度卖家大会Accelerate上发布新一代“智能体”(Agentic)AI工具,旨在为平台第三方卖家提供更主动的自动化运营支持。这些集成于"卖家助手"和"创意工作室"的功能可自动完成库存监控、广告创意生成乃至全流程广告活动策划。公司首席执行官安迪·贾西称第三方卖家生态是"零售史上最具影响力的合作模式",目前其贡献了亚马逊在线零售额超60%的份额。

然而,这份合作背后存在多重挑战。独立调查显示,部分卖家面临成本上升、政策限制与待遇不公等问题。美国联邦贸易委员会(FTC)正在进行的反垄断诉讼指控亚马逊滥用市场支配地位,另据披露该机构还在调查亚马逊与谷歌在搜索广告定价透明度方面的合规性。

新一代AI工具基于亚马逊Nova基础模型和Anthropic的Claude 4构建,强调结合平台实时数据与卖家个性化场景进行决策。为避免数据争议,亚马逊强调模型训练未使用特定卖家机密数据,所有推荐均基于聚合趋势与卖家自有数据的隔离计算。

财务数据凸显其重要性:2024年第三方服务为亚马逊带来1560亿美元收入,占其总营收近四分之一。贾西透露,25年前平台开放第三方卖家的决策曾引发内部激烈辩论,如今AI技术或将重塑零售协作生态。

中文翻译:

亚马逊公司于今日上午发布了一系列新型代理式人工智能工具,旨在为平台上的独立卖家提供主动协作支持——这些卖家贡献了亚马逊在线零售总额的60%以上。

升级版"卖家助手"和"创意工作室"工具将实现从库存监控到广告活动策划的全流程自动化。亚马逊表示,这些工具旨在进一步赋能在其运营已达25年之久的第三方市场上开展业务的企业。

首席执行官安迪·贾西周二在西雅图举办的亚马逊卖家大会"Accelerate"上对与会者表示:"亚马逊与这些卖家的合作已成为'零售史上最具吸引力且规模空前的协作关系'。"但部分卖家面临的现实更为复杂:独立分析师、供应商及咨询机构的调研显示,虽然亚马逊庞大市场带来了商机,但随之而来的是成本攀升、政策限制以及待遇不公的感受。

与此同时,美国联邦贸易委员会(FTC)正在审理的反垄断诉讼指控亚马逊通过多种手段非法滥用其对第三方卖家的支配权,该公司对此予以强烈否认。据彭博社上周报道,FTC还另案调查谷歌与亚马逊是否充分披露搜索广告定价信息——这类广告被卖家广泛用于推广商品列表。

这些宏观问题为亚马逊年度卖家大会构成了复杂却未明言的背景。本届"Accelerate"大会在西雅图会议中心举行,距亚马逊总部仅数街区之遥,吸引了全球各地卖家参与。

与当前科技行业整体趋势一致,人工智能成为本届大会的核心议题。这是亚马逊向外界展示其AI领域领导力的全面努力之一——此前该公司常被认为在追赶OpenAI、微软和谷歌等竞争对手。

亚马逊表示其现有生成式AI技术已被130万卖家使用,包括通过照片或简短描述自动生成商品列表、智能推荐产品属性与关键词、以及广告文案起草等功能。据公司介绍,新型代理式AI工具不仅具备这些能力,更能作为综合性助手——在卖家授权下可主动执行任务。

以升级版"创意工作室"为例,该工具旨在成为卖家的内置创意合作伙伴,能够研究产品与受众、构思创意方案、起草故事板,并全程协助生成视频及展示广告活动。亚马逊广告创意产品与技术副总裁杰伊·里奇曼在接受采访时表示:"用户不再只是点击按钮,而是能与创意伙伴进行自然语言对话,协同实现营销目标。"

升级版"卖家助手"的功能包括主动监控库存水平、标记潜在合规或账户健康问题、预测产品需求变化,甚至生成关于定价、促销或补货的具体建议。经卖家授权后,该工具可直接通过亚马逊系统执行发货准备、安排补货或调整商品列表等操作。

亚马逊宣称其新工具基于自研Nova AI基础模型和Anthropic公司的Claude 4大模型,并结合了平台深厚的购物数据与成功卖家经验。这些宏观智能还与每个卖家的实时运营情境相结合——包括库存水平、季节性规律和促销活动等数据,同时具备直接通过亚马逊内部系统与API执行操作的能力。

亚马逊销售合作伙伴服务副总裁达尔梅什·梅塔在本周大会采访中强调:"这不仅是基于历史数据的通用知识或能力,而是为每个卖家量身定制的解决方案。"他举例说明,针对节日季经营的商家,工具可精准建议发往亚马逊物流中心的商品数量,并在获得批准后全程安排物流运输直至生成货运标签。

值得关注的是,亚马逊获取电商平台商户机密数据的行为曾在其他场景引发争议,有指控称公司内部人员利用个体卖家信息开发竞争产品。对此梅塔表示,新AI工具刻意避免使用卖家特定数据训练模型,确保不会向竞争卖家泄露商业机密。亚马逊通过聚合宏观趋势与洞察,再结合卖家自身情境数据进行分析,同时严格隔离每个卖家的信息。

"这对隐私保护至关重要,"梅塔补充道,"没有人希望自己的数据被用于增强竞争对手优势。"他表示目标是建立信任机制,让卖家安心授权AI代理执行操作。

对亚马逊和卖家而言,其中的经济利害关系极为重大:

这一切使得该业务的诞生更具传奇色彩。贾西周二下午与梅塔同台演讲时透露,25年前公司内部就是否允许第三方卖家入驻曾引发"激烈辩论"。当时高管们担心第三方商户无法达到其品牌与服务标准,但经过数周研讨,最终认定更多卖家将带来更丰富的选品和更低廉的价格。

"我们共同成就的事业非同寻常,"贾西表示,并预测新一代AI工具将使合作更加成功,"但我认为这仅仅是个开始。"

英文来源:

Amazon this morning unveiled a new wave of agentic AI tools designed to act as proactive partners for the independent sellers who generate more than 60% of its online retail sales.
The upgrades to Amazon’s Seller Assistant and Creative Studio promise to automate everything from monitoring inventory to producing polished ad campaigns. Amazon says they’re meant to further empower the businesses that operate on its 25-year-old third-party marketplace.
Amazon’s partnership with those sellers has turned into “probably the most compelling and substantial collaboration in the history of retail,” CEO Andy Jassy told the crowd during a Tuesday appearance at the company’s Accelerate seller conference in Seattle.
Some sellers see a more complicated reality, where the benefits of access to Amazon’s giant marketplace come with rising costs, restrictive policies, and a sense of unequal treatment, as reflected in independent surveys from analysts, vendors, and consulting firms.
Meanwhile, an ongoing FTC antitrust case alleges that Amazon has illegally leveraged its power over third-party sellers through a variety of practices that the company strongly denies.
Bloomberg News reported last week that the FTC is separately looking into new questions about whether Google and Amazon gave sufficient disclosures about pricing for search ads that, in the case of Amazon’s marketplace, are heavily used by sellers to promote listings.
Those larger issues form a complex yet largely unstated backdrop for Amazon’s annual seller conference. Accelerate is taking place this week at the Seattle Convention Center, a few blocks from Amazon’s Seattle headquarters, drawing sellers from across the country and the world.
Expanding on existing AI tools
As with much of the tech industry right now, AI is front-and-center at this year’s conference. It’s part of Amazon’s companywide effort to demonstrate leadership in a field where it has sometimes been seen as playing catch-up to OpenAI, Microsoft, Google, and others.
Amazon says its existing generative AI technologies have been used by 1.3 million sellers. They include tools that automatically generate product listings from a photo or short description, suggest attributes and keywords to include in listings, and draft ad copy, for example.
The new agentic AI tools go beyond those capabilities to serve as broader assistants — in some cases acting proactively, with the seller’s permission, according to the company.
The additions to the Creative Studio, for example, aim to give sellers a built-in creative partner that can research products and audiences, brainstorm concepts, draft storyboards, and generate full video and display ad campaigns from start to finish.
“So rather than clicking a button, you’re actually able to have a natural-language conversation with a creative partner that is working with you to help you achieve your goals,” said Jay Richman, vice president of creative products and technology for Amazon Ads, in an interview.
Features of the upgraded Seller Assistant include the ability to actively monitor inventory levels, flag potential compliance or account-health issues, and anticipate shifts in product demand — even generating specific recommendations on pricing, promotions, or restocking.
With a seller’s approval, the tool can then take action by preparing shipments, scheduling replenishment, or adjusting listings directly through Amazon’s systems.
Going beyond generalized knowledge
Amazon says it’s powering the new tools with its Nova AI foundation model and Anthropic’s Claude 4 in conjunction with its deep knowledge of shopping, successful seller approaches.
That broad intelligence is then combined with real-time context about each seller’s specific situation and history on the platform — such as inventory levels, seasonal patterns, and promotions, plus the ability to act directly through Amazon’s internal systems and APIs.
“So it’s not just some generalized knowledge or generalized capability from all the history. It is tailored to their specific situation,” said Dharmesh Mehta, Amazon’s vice president of selling partner services, in an interview about the new tools at the conference this week.
For example, the tools could create tailored recommendations for a holiday-driven business on exactly how many units to send into Fulfillment by Amazon — and then, if approved, schedule and facilitate the shipments, all the way down to creating the shipping labels.
Amazon’s access to confidential merchant data on its e-commerce platform has been a sticking point in other situations, giving rise to accusations that people inside the company have mined individual seller information to launch competing products. Amazon has repeatedly denied these allegations.
With the new AI tools, Mehta said, the company has been careful not to train its models with seller-specific data, ensuring that no confidential information informs its recommendations to a competing seller. Instead, Amazon aggregates broad trends and insights, and combines them with the seller’s own context, while keeping that seller’s information siloed.
“That’s really important for privacy,” Mehta added, noting that no one would want their personal data used to give a competitor an advantage. The goal, he said, is to build trust so that sellers are comfortable letting the AI act on their behalf with their permission.
The broader financial stakes are huge for Amazon and sellers.

Geekwire

文章目录


    扫描二维码,在手机上阅读