在一次面向产品领导者的聚会上，演讲者讨论了产品经理（PM）在AI时代不断演变的角色。最初，人们普遍担心工作岗位会消失，各种角色（包括产品经理）都害怕被淘汰。然而，这种担忧已经消退，取而代之的是角色重叠和扩展的现实，开启了“AI构建者”的新时代。 “AI构建者”的概念正在受到关注，尤其是在初创公司，个人通过身兼多职来体现这一角色，类似于创始人的运作方式。虽然这在小型公司中是直接的，但演讲者探讨了这如何转化到拥有超过10,000人的组织中，特别从Atlassian公司中汲取了例子。 产品经理的基本工作——寻找产品市场契合点、构建受喜爱产品并确保业务可行性——保持不变。改变的是产品经理实现这些目标的方式，主要得益于快速发展的AI工具和增强的组织环境。例如，Atlassian公司开发了一个“团队协作图谱”（teamwork graph），以提供一个全面的“企业大脑”（enterprise brain），使产品经理能够卸载传统的行政任务，如会议记录或后续跟踪。这让他们能够专注于“加速”，利用智能和上下文来加快项目交付。 引入的一个核心比喻是，产品经理是“划桨”（亲自动手贡献）还是“掌舵”（设定方向，排除障碍）。演讲者指出，最佳方法“取决于”产品类型及其当前阶段。 以下三个Atlassian的具体例子说明了这种动态： 1. **Confluence（现有代码库中的新功能）：** 对于“Remix with Rovo”和“Confluence Slides”等功能，产品经理Aya采取了一种激进的方法。以前从未写过代码的她，在一个月内提交了26个PR（Pull Request），在工程团队提供特殊“引导”（harness）支持下。她的动机是加速用户体验（UX）修复，并让工程师能够专注于更高级别的任务。产品经理还负责Confluence Slides的评估工作，使吞吐量翻倍，并使用Atlassian的LLM平台Arise进行提示调试。设计缺陷修复通过Figma MCP和编码代理实现自动化，每小时解决14个缺陷。工程师的测试创建时间从半天缩短到10分钟。这种产品经理“划桨”且角色有意模糊的方法，使Remix在六周内上线，Confluence Slides在八周内上线，比之前六个月的时间表显著加速。 2. **RovoClaw（从零到一的产品）：** 这个从零开始的项目由产品经理Josh和设计师Kevin通过“氛围编码”（vibe coding）的方式，共同开发出了一个可用的alpha版本。最初，Josh深入参与了编码工作。然而，随着工程师的加入，他意识到他最具杠杆效应的活动发生了转变。他从编码工作中抽身，转而专注于“掌舵”——设定方向、确定优先级并为团队排除障碍。他最初的编码经验在理解挑战和有效引导项目方面被证明是无价的。他还利用AI，使用RovoClaw本身，自动化了每周更新，消除了手动报告。在这里，产品经理最初是“划桨”，但随着项目阶段的需求，他过渡到了“掌舵”。 3. **JIRA（大型现有代码库的增强）：** JIRA，一个拥有复杂代码库的20年老产品，需要一种不同的方法。产品经理直接编码被认为风险过高。目标是使JIRA“AI优先”，在JIRA内部构建AI功能以帮助客户进行软件开发。在10周内完成22个面向用户的功能的吞吐量是正常情况下的三倍。这是通过以下方式实现的： * **精简原型设计：** 使用Loom创建工作项，然后启动编码代理在托管云环境中编写代码，确保符合Atlassian的设计语言。 * **自动化反馈分类：** Slack中的JIRA代理对反馈和缺陷进行分类，并将其发送给编码代理进行修复。 * **客户洞察：** 一个“机器人代理”（robo-agent）通过JIRA Service Management审查了来自用户研究的900多条反馈，为工程团队提供了高质量的分类。在这个例子中，产品经理显然是在“掌舵”，专注于为工程师排除障碍并通过AI工具和流程提高效率。 总而言之，产品经理现在正在进行原型设计、测试、撰写评估、深入研究客户反馈，并以更快的速度获取信息，所有这些都得益于理解组织上下文的AI工具。相反，他们不再进行手动更新、手动整理研究资料或从头开始制作幻灯片，并且同步会议也可能减少了。 Atlassian培养AI流利度的策略包括“AI流利度指数”（AI Fluency Index），这是一个包含六种能力（例如，工具使用、评估、技术素养）并按1-5评分的开发工具，旨在指导产品经理进行技能发展。他们还每季度举办成功的“AI构建者周”（AI Builders Weeks）活动，这是密集的培训课程，侧重于原型设计或构建代理等实践技能，已吸引了1000多人参与，并产生了120多个新工作流程。尽管衡量这些变化的成果仍然是一个挑战，Atlassian会跟踪已部署的PR和已交付的功能等指标。 演讲者总结说，产品经理的角色正在扩大，使他们能够更有效地工作，并为客户提供更多价值，这使得现在成为产品经理是一个极其激动人心的时刻。

Addressing a gathering of product leaders, the speaker discussed the evolving role of Product Managers (PMs) in the age of AI. Initially, there was widespread nervousness about jobs disappearing, with various roles (including PMs) fearing obsolescence. However, this fear has subsided, replaced by the reality of roles overlapping and expanding, ushering in a new era of the "AI builder." The concept of the "AI builder" is gaining traction, especially in startups, where individuals embody this role by wearing multiple hats, akin to how founders operate. While this is straightforward in small companies, the speaker explored how this translates to organizations with over 10,000 people, specifically drawing examples from Atlassian. The fundamental job of a PM—finding product-market fit, building loved products, and ensuring business viability—remains unchanged. What has transformed is *how* PMs achieve these goals, primarily due to rapidly advancing AI tools and enhanced organizational context. Atlassian, for instance, has developed a "teamwork graph" to provide a comprehensive "enterprise brain," allowing PMs to offload traditional administrative tasks like note-taking or follow-ups. This frees them to focus on "acceleration," leveraging intelligence and context to expedite project delivery. A central metaphor introduced was whether a PM is "rowing" (hands-on contribution) or "steering" (setting direction, unblocking). The speaker noted that the optimal approach "depends" on the product type and its current phase. Three concrete Atlassian examples illustrated this dynamic: 1. **Confluence (New Feature in Existing Codebase):** For features like "Remix with Rovo" and "Confluence Slides," the PM, Aya, took a radical approach. Having never coded before, she checked in 26 PRs in a month, with engineering support providing a special "harness." Her motivation was to accelerate UX fixes and allow engineers to focus on higher-level tasks. PMs also took ownership of evaluations for Confluence Slides, doubling throughput, and used Atlassian's LLM platform, Arise, for prompt debugging. Design bug fixing was automated using Figma MCP and coding agents, addressing 14 bugs an hour. Test creation time for engineers was reduced from half a day to 10 minutes. This approach, where the PM was "rowing" and roles were intentionally blurred, enabled Remix to launch in six weeks and Confluence Slides in eight weeks, a significant acceleration from the previous six-month timeline. 2. **RovoClaw (Zero to One Product):** This ground-up project began with a PM, Josh, and a designer, Kevin, "vibe coding" a working alpha. Initially, Josh was deeply involved in coding. However, as engineers joined, he realized his highest leverage activity shifted. He stepped back from coding to focus on "steering"—setting direction, prioritizing, and unblocking the team. His initial coding experience proved invaluable in understanding challenges and effectively steering the project. He also leveraged AI, using RovoClaw itself, to automate weekly updates, eliminating manual reporting. Here, the PM started "rowing" but transitioned to "steering" as the project phase demanded. 3. **JIRA (Enhancements in Large Existing Codebase):** JIRA, a 20-year-old product with a complex codebase, required a different approach. Direct PM coding was deemed too risky. The goal was to make JIRA "AI-first," building AI features within JIRA to assist customers with software development. Throughput for 22 user-facing features in 10 weeks was three times the norm. This was achieved through: * **Streamlined Prototyping:** Using Loom to create work items, which then kicked off coding agents to write code in a managed cloud environment, ensuring compliance and adherence to Atlassian's design language. * **Automated Feedback Triage:** JIRA agents in Slack triaged feedback and bugs, sending them to coding agents for fixes. * **Customer Insights:** A "robo-agent" reviewed over 900 pieces of feedback from user studies via JIRA Service Management, providing high-quality triage for engineering. In this example, the PM was clearly "steering," focusing on unblocking engineers and enhancing efficiency through AI tools and processes. In summary, PMs are now prototyping, testing, writing evaluations, delving deeper into customer feedback, and consuming information at a higher rate, all facilitated by AI tools that understand organizational context. Conversely, they are no longer performing manual updates, compiling research by hand, or creating slides from scratch, and synchronous meetings have potentially reduced. Atlassian's playbook for fostering AI fluency includes the "AI Fluency Index," a development tool with six capabilities (e.g., tool use, evals, technical literacy) rated 1-5, guiding PMs towards skill development. They also run successful "AI Builders Weeks" quarterly, intensive training sessions focusing on practical skills like prototyping or building agents, which have engaged over a thousand people and led to 120+ new workflows. While measuring the outcomes of these changes remains a challenge, Atlassian tracks metrics like PRs deployed and features delivered. The speaker concluded that PM roles are expanding, enabling them to work more effectively and deliver more to customers, making it an incredibly exciting time to be a PM.