What it takes to be a top PM today | Robby Stein (Google Search)

演讲者，一位拥有近20年经验的产品经理和构建者，其经历包括在 Instagram（Stories、Reels、Feed ranking、Direct messaging）和 Google Search（AI 时代体验）领导团队，分享了关于构建成功产品的见解。曾参与开发数十亿人使用的产品，也曾开发过“无人问津”的产品，他们识别出了一种可重复的成功模式。 产品经理（PM）的根本作用正在发生转变。尽管PM传统上专注于“把事情办成”、组织和建立势头，但演讲者认为，在一个几乎任何东西都可以被构建的时代，PM的真正价值在于“判断力”、“品味”和“把事情做到极致”。最重要的是，PM的核心技艺是**决策**，而优秀的PM是这一技能的“学生”。 演讲者概述了一个构建产品的“三章”攻略： **第一章：深入理解用户。** 这个过程始于理解人类的基本需求。援引 Clayton Christensen 的《与运气竞争》和“Jobs-to-be-done”框架，演讲者强调人们“雇佣”产品是为了满足特定的需求。一个个人例子涉及到购买一张床：对他妻子来说，关键的洞察力不是凉爽或价格，而是需要一张在他翻身时不会吵醒她的床。如果不对“为什么”进行深入的用户访谈，这种“转移”的洞察力就会被错过。 在数字领域，谷歌早期的购物AI体验揭示了用户需要灵感，这促使他们投资于多模态AI，以实现视觉产品描述（例如，祖母绿沙发）。同样，最初的 Google Search AI 模式缺乏地图和星级评分等权威上下文，而这些对于美食相关的查询至关重要；整合这些元素使产品变得“神奇”。演讲者指出，AI 可以改变这一章，可能会使用AI分析用户访谈记录来寻找jobs-to-be-done，甚至让AI代理进行大规模访谈。 **第二章：诊断根本原因。** 产品很少一开始就非常出色。成功来自于分析性地识别问题、优先处理它们、积极地解决它们，并进行迭代——非常类似于“训练模型”。对于 Instagram Stories，最初的采用受到“受众问题”的阻碍（“我的前任在上面，我的老师也在上面”）。广泛的定量调查证实了这一点，导致在两年迭代后开发了“亲密朋友”功能，该功能只有在Stories内使用时才取得了成功。另一个例子是 Instagram Reels 在巴西最初的“阅后即焚”发布，该发布失败了，因为那些为舞蹈投入精力的用户希望他们的内容是永久性的并能病毒式传播，而不是消失。这促使他们将Reels变成了持久的格式。现在AI也在协助这一过程；谷歌内部工具，例如“反重力”系统，收集选择加入的用户反馈。AI 可以处理定性评论，例如背包购物者需要关于孩子身高的信息，以揭示共同需求并推动创建“协作对话”的功能。 **第三章：工艺（契合与精修）。** 这个经常被忽视的章节关注的是有意为之的细节，这些细节传达了创造者是否关心他们的产品。它由两个因素驱动：完美的功能（没有用户痛点）和激发愉悦。在痛点方面，演讲者描述了一个AI驱动的内部工具，它充当用户，截取产品体验的屏幕截图，并根据标准评估它们，以发现故障或不符合规范的行为，例如数学渲染问题或视觉托盘缺失。这使得质量保证（QA）的规模超越了人类团队，AI也越来越能够自行修复问题。为了愉悦用户，谷歌为AI时代重新设计了其搜索框。微妙的细节，例如光标闪烁时循环显示谷歌颜色，点击时搜索框“跳出”，以及触觉反馈，都传达了“产品背后的人性”并赋予其“灵魂”。 总之，演讲者强调，尽管这些核心原则并非新鲜事物，但AI显著增强了它们的运用，使得PM的技艺比以往任何时候都更加关键。这套攻略——深入理解用户、诊断根本原因和专注于工艺——通过优先考虑“屏幕另一端的实际用户和人类”，使PM能够构建出色的产品。

The speaker, a product manager and builder with nearly 20 years of experience, including leading teams at Instagram (Stories, Reels, Feed ranking, Direct messaging) and Google Search (AI era experiences), shared insights into building successful products. Having worked on products used by billions and some that "zero people use," they've identified a repeatable pattern for success. The fundamental role of a Product Manager (PM) is shifting. While PMs traditionally focused on "getting stuff done," organizing, and building momentum, the speaker argues that in an age where almost anything can be built, the true value of a PM lies in "judging," "taste," and "doing something extremely well." Above all, the core craft of a PM is **decision-making**, and great PMs are students of this skill. The speaker outlined a three-chapter playbook for building products: **Chapter 1: Understanding People Deeply.** This process begins with understanding fundamental human needs. Referencing Clayton Christensen's "Competing Against Luck" and the "Jobs-to-be-done" framework, the speaker emphasized that people "hire" products to fulfill specific needs. A personal example involved purchasing a bed: the key insight for his wife wasn't cooling or price, but the need for a bed that wouldn't wake her when he moved. This "transference" insight would be missed without deep user interviews focused on the "why." In the digital realm, Google's early AI experiences for shopping revealed users wanted inspiration, leading to investments in multimodal AI for visual product descriptions (e.g., emerald green couches). Similarly, initial Google Search AI modes lacked authoritative context like maps and star ratings, which were crucial for food-related queries; integrating these made the product "magical." The speaker noted AI could transform this chapter, potentially using AI to analyze user interview transcripts to source jobs-to-be-done or even having AI agents conduct interviews at scale. **Chapter 2: Diagnosing Root Causes.** Products rarely start great. Success comes from analytically identifying problems, prioritizing them, fixing them aggressively, and iterating—much like "training a model." For Instagram Stories, initial adoption was hampered by "audience problems" ("My ex is on it, my teacher's on it"). Extensive quantitative surveys confirmed this, leading to the development of "Close Friends" after two years of iteration, which only succeeded when confined to Stories. Another example was Instagram Reels' initial ephemeral launch in Brazil, which failed because users who put effort into dances wanted their content to be permanent and go viral, not disappear. This led to making Reels a lasting format. AI now assists in this process; internal tools like Google's "anti-gravity" system gather opt-in user feedback. AI can process qualitative comments, like backpack shoppers needing questions about a child's height, to reveal common needs and drive features that create a "collaborative conversation." **Chapter 3: Craft (Fit and Finish).** This often-overlooked chapter concerns the intentional details that convey if creators cared about their product. It's driven by two factors: perfect functionality (no user pain) and sparking joy. On the pain side, the speaker described an AI-powered internal tool that acts as a user, screenshots product experiences, and evaluates them against a rubric to find breakages or off-spec behaviors, such as math rendering issues or missing visual trays. This scales QA beyond human teams, with AI increasingly able to fix issues itself. For delight, Google reinvented its search box for the AI era. Subtle details like the blinking cursor cycling through Google colors, the search box "jumping out" when tapped, and haptic feedback communicate "humanity behind the product" and give it "soul." In conclusion, the speaker emphasized that while these core principles aren't new, AI significantly enhances their application, making the craft of PM more critical than ever. The playbook — understanding people deeply, diagnosing root causes, and focusing on craft — empowers PMs to build great products by prioritizing the "actual people and the humans on the other side of the screen."