Ch04-L11 技术洞察 像管理者一样思考

技术洞察：像管理者一样思考

我的理解

本课提出核心洞察：GenAI 的高度通用性使其与人类助理相似，而非与功能单一的传统工具相似，因此有效管理 AI 所需的技能组合——沟通、文档管理、了解成员能力边界、评估机制、风险管理——与带人管理高度一致。这意味着提升 GenAI 效能最快的路径不是钻研 ML 论文或 prompt 工程，而是学习管理，因为高层次决策带来最大收益也浪费最多时间。以编程为例，文档管理能力正在取代算法熟练度成为 AI 时代更关键的开发者技能；以学习方式为例，GenAI 使“问题驱动的自上而下学习”成为可能，打破了传统自下而上积累的低效。核心告诫是：在委托 AI 扩大产出的同时，始终守住属于自己的核心竞争力与判断力，因为你是自己成长的真正责任人。

相关链接

• Ch04-L04 学习要点 1 有选择地把任务委托给 AI — 有选择地委托是本课管理框架的第一支柱，两课形成原则与实践的呼应
• Ch04-L06 学习 2 面向 AI 的文档管理 — 文档管理被本课确立为 AI 时代工程师核心竞争力重塑的关键，从“技巧”提升为“战略”
• Ch05-L02 引言 — 下一模块引言在“如何思考 AI 机会”层面延伸了本课管理者视角，是思维层面的自然延续
• Ch01-L06 技术洞察 3 从用户到构建者 — 本课是 AI 管理者思维的进阶升级，Ch01 奠定了从用户到构建者的基础框架，两课首尾呼应

---

原文

Lesson 36 of 68 技术洞察：像管理者一样思考 / Technical insight: Think like a manager

通过这个案例研究，我们手中多了四件新武器：

有选择地把任务委派给 AI。

文档管理

评估机制

风险管理

这些技能与带人管理者所需要的能力非常相似。仔细想想，这其实并不奇怪，因为 AI 极其通用。你不会问“我要不要把这个任务交给我的汽车？”——当你需要去某个地方时，你直接上车就行。其他工具都有专门且明确的用途，但 GenAI 没有，因为它能适配如此多样的场景。从这个角度看，它确实更像一位人类助理。

这种与人类的相似性，对我们许多人，尤其是个人贡献者（IC）而言，带来的麻烦其实多于好处，因为它要求我们去学习管理知识、沟通技巧与判断力。新晋管理者常常发现，自己带的小团队的产出反而不如他还是 IC 时高效，这种情况非常普遍。无论我们开始管理 AI 还是管理人类，这都是一种颠覆性的转变。管理 AI 相对更容易，因为我们不必担心如何激励它们，也不需要顾及工作与生活的平衡，但许多管理原则在这里同样适用。

多了一位可以分派任务的团队成员，立刻就需要沟通；而要让沟通高效，就需要文档管理；你必须了解团队成员（AI）的优势与短板，才能合理地拆解和分配任务；你需要评估机制，以便判断它们何时出错、何时做得好。当事情变得更复杂、不确定性增加时，我们就需要管理并降低风险。这一切对 AI 和对人类都同样适用。这正是为什么有效管理 AI 所需的技能组合，与管理人类如此相似。

从这个角度看，充分利用 GenAI 的最有效方式，并不是去阅读学术论文或成为提示工程的专家，而是去学习管理。高层次的决策带来最大的收益，也可能浪费最多的时间，因此我们必须把它们做对。一旦你对这些技能更加熟悉，就可以围绕 AI 来定制你的机制、流程与工作方式。设计最契合团队的机制与流程，本就是带人管理者的核心能力之一。

编程就是一个例子。借助最新的 AI，GenAI 完全有能力以良好的风格和质量编写简单或复杂的程序。传统上，评估（入门级）专业开发者的关键标准之一，是对数据结构与算法的扎实掌握。换句话说，衡量标准在于一名开发者能多快地把一个用自然语言清晰描述的问题，转化为一个可运行的程序。

然而，鉴于 AI 在这一点上可以说做得比人类更好，这一标准正在发生变化。正如我们在前面的课程中提到的，文档管理正逐渐成为一项关键技能。如果有两位开发者——一位精通传统的数据结构与算法但不懂 GenAI，另一位熟悉接口、注释与文档管理——后者借助 GenAI 很可能写出更好的程序。

这一洞察正在促使开发者围绕 GenAI 重塑自己的工作流，乃至培养计划。对于数据结构与算法，学到能够评估现有代码的程度，可能就已足够；其余工作流应聚焦于如何有效使用 GenAI 与文档管理。但软件工程领域的其他能力——例如定义“好软件”（即定义成功标准），以及让软件具备可维护性和可复用性——依然至关重要。我们可以与 GenAI 探讨这些方面以获得启发，但 AI 目前还无法替我们完成设计工作。

这种转变强调的是：要在传统知识与有效驾驭 GenAI 的能力之间取得平衡，最终带来更高产、更具创新性的开发实践。

另一个例子是，问答式机器人的新交互形式正在改变我们的学习方式。传统的学习方法是自下而上的：我们读书、看视频，逐步积累对基础知识的理解，然后把它们逐渐连接起来，形成更整体的视角。这种方法对相关知识的覆盖很好，但有两个明显缺点。第一，在学习过程中，“我为什么要学这个？”这个问题往往得不到回答，导致缺乏动力。第二，当我们带着具体问题去学习以求解决时，很难快速定位到所需的知识。

但 GenAI 改变了这一切。我们可以很方便地直接向 ChatGPT 提问，并在提示中附上可选的搜索结果。这样，我们总能针对自己的问题获得最直接的答案。如果愿意，我们还可以继续深入，去学习背后的底层知识。我们不再依赖自下而上的学习策略，而是可以从任何一点出发，获得最直接的学习，再根据自己的兴趣向上探索（更抽象的层面）或向下深入（更具体的细节）。这是围绕 GenAI 重塑学习工作流的又一个绝佳例子。

采用这种自上而下的学习方法，我们可以迅速解决具体的问题，然后按需扩展自己的理解。这种方法让我们保持动力，并带来更高效、更有针对性的学习体验。

作业：

=======================================================

在你的日常工作中，你觉得 GenAI 在哪些方面尤其擅长？你能否围绕 GenAI 重新组织现有的工作流，让它效率大幅提升？

=======================================================

归根结底，使用这一工具所获得的全部收益与所承担的全部责任，都由你自己承受。因此，你才是自己项目与成长的真正主人。明智地使用 GenAI，并守住你的核心竞争力。

English Original

Through the case study, we had four new weapons in our hand:

Delegate to AI selectively.

Document management

Assessment mechanism

Risk management

And these skills are very similar to those of a people manager. If we think about it twice, it’s actually no surprise, because AI is so versatile. You don’t ask, “Shall I delegate this task to my car?”—you just jump in when you need to go somewhere. Other tools have dedicated and well-defined usage, but GenAI doesn’t, because it can fit so many different scenarios. From this perspective, it’s really like a human assistant.

This similarity to humans has actually caused more trouble than benefits for many of us, especially individual contributors (ICs), because it requires us to learn management expertise, communication skills, and judgment. It’s quite common to see new managers struggle to make a small team more productive than when the manager was an IC. It is a disruptive change anyway, whether we begin to manage AIs or humans. Managing AIs is easier because we don’t need to worry about motivating them or addressing work-life balance, but many management principles still apply here.

An extra team member to delegate tasks to immediately calls for communication. You then need document management to make communication efficient. You need to understand the strengths and weaknesses of your team members (AI) so you can properly decompose and assign tasks to them. You need assessment mechanisms to know when they make mistakes or do things well. As things get more complicated and uncertainty increases, we need to manage and mitigate risk. Everything applies to both AI and humans. That’s why the skill set for effectively managing AI is so similar to that for managing humans.

From this perspective, the most effective way to fully leverage GenAI is not by reading academic papers or becoming an expert in prompt engineering. Instead, it is to learn about management. High-level decisions gain the most benefit or waste the most time, so we need to make them right. Once you become more familiar with these skills, you can customize your mechanisms, procedures, and workflows around AI. Setting up mechanisms and procedures to best fit the team is among the core competencies of a people manager.

One example is programming. With the latest AIs, GenAI is more than capable of writing simple or complicated programs in good style and quality. Traditionally, a critical evaluation criterion for (entry-level) professional developers is a solid understanding of data structures and algorithms. In other words, the measure was how quickly a developer could translate a well-defined problem in natural language into a working program.

However, with AI arguably doing this better than humans, this criterion is changing. As we mentioned in previous lessons, document management is becoming a critical skill. If there are two developers—one knows a lot about traditional data structures and algorithms but has no GenAI knowledge, and the other is familiar with interfaces, comments, and document management—the latter would likely write better programs with the help of GenAI.

This insight is inspiring developers to rebuild their workflows and even education plans around GenAI. For data structures and algorithms, it might be sufficient to learn them to a point of being able to assess existing code. The remaining workflow should focus on the effective use of GenAI and document management. However, other expertise in the software engineering field, such as defining good software (i.e., defining success criteria) and making it maintainable and reusable, remains crucial. We can discuss these aspects with GenAI to get inspiration, but AI cannot yet handle the design work for us.

This shift emphasizes the need to balance traditional knowledge with the ability to leverage GenAI effectively, ultimately leading to more productive and innovative development practices.

Another example is how the new interaction form of a Q&A bot changes the way we learn. The traditional approach to learning is a bottom-up approach. We read books, watch videos, accumulate an understanding of the basics, and then gradually connect them to form a more holistic view. This method provides good coverage of all relevant knowledge but has two significant drawbacks. First, during the learning process, the question “Why do I need to learn this?” often remains unanswered, leading to a lack of motivation. Second, when we have a specific problem in mind and try to learn to solve it, it’s difficult to quickly locate the necessary knowledge.

But GenAI has changed all this. It’s easy to simply ask questions to ChatGPT with optional search results supplied in the prompt. This way, we always get the most direct answers to our questions. If we want, we can dig deeper and learn the underlying knowledge. Instead of relying on a bottom-up learning strategy, we can start anywhere, get the most direct learning, and explore further up (more abstract) or down (more detailed) based on our interest. This is another excellent example of rebuilding our learning workflow around GenAI.

By adopting this top-down learning approach, we can quickly address specific questions and problems, then expand our understanding as needed. This method keeps us motivated and allows for a more efficient and targeted learning experience.

Homework:

=======================================================

What are the areas you feel GenAI especially excel at in your daily work? Can you wrap your existing workflow around GenAI to make it much more efficient?

=======================================================

At the end of the day, you are the one who reaps all the benefits and undertakes all the responsibilities of using the tool. So you are the actual owner of your own projects and growth. Use GenAI wisely, and maintain your core competencies.