第 1 步：识别你工作流程中的瓶颈
我的理解
工作流优化的起点不是“AI 能做什么”，而是“我的核心竞争力是什么、什么在阻碍它发挥”——AI 只是众多解决方案之一，而非目的本身。识别核心竞争力（让你在同事中脱颖而出的能力）与限制因素（消耗时间却产出低的任务）是分析框架的两条主线，两者共同指向委托的优先级，而非委托的边界。以 TPM 角色为例：大规模追踪与协调是核心竞争力，而铺天盖地的邮件和消息是瓶颈——将第一轮优先级排序交给 AI 既合理又低风险，最坏情况不过是 AI 帮不上忙。这种“从需求出发、AI 作为备选”的思路，是避免为用 AI 而用 AI 的根本防线。
相关链接
- Ch04-L04 学习要点 1 有选择地把任务委托给 AI — 识别瓶颈后紧接的委托原则，解释为何选择 Gmail 优先级排序这个具体任务
- Ch04-L05 第二步 探索技术可行性 — 确认委托目标后的下一步，验证方案在技术层面能否实现
- Ch05-L07 可落地的指南 哪些任务该交给 GenAI — 对哪类任务适合委托给 GenAI 的系统性指南，与瓶颈识别形成互补
- Ch02-L09 自动化的力量 以及生成式 AI 的角色 — 自动化的价值与 GenAI 在其中的定位，提供宏观背景
原文
Lesson 28 of 68 第 1 步：识别你工作流程中的瓶颈 / Step 1: Identify the bottleneck of your workflow
我们在前面的课程中已经有所涉及，这里做一个正式的介绍。任何工作流程优化的目标，都是把精力聚焦在那些需要你专业能力、批判性思维和创造力的任务上。AI 或 AI 辅助的自动化是一种手段，而不是目的。关键不在于决定“要自动化什么”，而在于决定“要优先做什么”。一旦明确了重点领域，就能找出阻碍我们专注其上的障碍，而解决方案中可能会用到 AI。
推荐的思考路径，是先识别你的核心竞争力：是什么让你在同事中脱颖而出？工作中哪一部分让你感到愉悦？接着再思考限制因素：是什么阻碍了你的成长？哪些任务让你心生疑问“今天怎么还没过完”？目标是通过委派的方式，让积极的部分更高效，让消极的部分负担更轻。
在本次案例研究中，设想我们是一名 TPM（Technical Program Manager，技术项目经理）。我们要同时追踪过去和未来的各类活动，制定计划、时间表和备选方案。
核心竞争力：大规模地进行跟踪、沟通和协商。
主要瓶颈：开不完的会议、铺天盖地的 IM 消息，以及难以管理的大量浏览器标签页。深度思考永远是一种奢侈。
明确这两方面之后，我们可以进一步发散思考：
如何让你的优势更具可扩展性、更有成效？
提示：可行方案与不可行方案：
让 AI 记住并管理所有事情大概率不是一个好主意，因为这会形成一个又长又复杂的上下文窗口，可能引发问题（参考第 2 模块中关于上下文窗口长度限制的内容）。
但让 AI 帮我们捕捉可能遗漏的事项，则可能是有益的。最坏的情况是 AI 帮不上忙；而一旦它指出我们漏掉的内容，我们就赚到了。
具体来说，我们可以把各个项目的文档喂给 AI，请它给出灵感或指出被忽视的事项。我们并不指望它给出深刻洞见，但演绎和推理本就是 AI 的强项。
好的示例对 AI 充分发挥推理能力非常重要。例如，提供项目 A 的一个有力示例，可以有效启发 AI 在项目 B 中进行类似的复核，并套用相同的逻辑。
通过实施这些策略，我们可以更好地管理工作负担，确保关键事项不被疏漏。
用 GenAI 起草邮件大概率不是一个好主意，因为 AI 目前在处理沟通方面还不够可靠（措辞往往看起来很高级，却通常缺乏情感分量）。不过，它在汇总信息、跟踪“需要和谁就什么话题沟通”方面可以是一个很有价值的工具。
我们也始终可以选择更深入地挖掘，找出具体阻碍优势规模化的瓶颈。
如何让瓶颈变得更可接受？
提示：可行方案与不可行方案：
专属优先级排序器：把第一轮优先级排序交给 AI。例如，AI 可以读取 Slack 消息、邮件，甚至会议记录（来自 Zoom 或其他工具）。作为 builder，我们可以很容易地搭建出这套系统。AI 会就重要消息向我们发出提醒，其余的则集中起来统一处理。我们将在案例研究中使用这一方案。
使用 Zoom AI Companion 生成会议纪要：借助 AI 生成会议纪要，并在不那么关键的会议中并行处理其他事务。这虽然存在一定风险，但完全有可能找到一个收益大于代价的平衡点。
不要把批判性思考交给 AI：依赖 AI 提供洞见并不可取，因为 AI 在这方面并不擅长。此外，这样做也会削弱你的核心竞争力。
采用这些策略，我们可以让瓶颈更加可控，同时确保关键优势依然具备可扩展性和有效性。
作业
==================================================
在你目前的岗位上，你自己的核心竞争力是什么？如何让它更具可扩展性、更有成效？
在你目前的岗位上，你自己的限制因素是什么？如何让它更省时、更少需要亲力亲为？
==================================================
English Original
We touched on this in previous lessons, but here’s a formal introduction. The goal of any workflow optimization is to focus on tasks that require your expertise, critical thinking, and creativity. AI or AI-assisted automation is a method, not a goal. The key is not to decide what to automate, but what to prioritize. Once we identify our focus areas, we can then pinpoint the blockers that prevent us from concentrating on them, and the solution might involve AI.
A recommended thought process starts with identifying your core competency. What makes you stand out from your colleagues? What part of your job brings you joy? Next, consider the limiting factors. What hinders your growth? What tasks make you wonder why the day hasn’t ended? Aim to make the positive aspects more efficient and the negative aspects less burdensome through delegation.
In this case study, imagine we are a TPM (Technical Program Manager). We keep track of both past and future activities, developing plans, timelines, and backup strategies.
Key Competency: Tracking, communicating, and negotiating at scale.
Key Blockers: Endless meetings, an overwhelming number of IM messages, and difficulty managing numerous open browser tabs. In-depth thinking is always a luxury.
After identifying these two sides, we can further brainstorm:
How to make your strengths more scalable and effective?
Hint: potential solutions and non-solutions:
It’s probably not a good idea for AI to remember and manage everything, as this could create a long and complicated context window, potentially causing issues (referencing the context window limit from Module 2).
However, it may be beneficial to ask AI to capture potentially missed items. In the worst-case scenario, the AI offers no help, but if it highlights something we missed, we gain an advantage.
Specifically, we could feed our AI documents on various projects and ask for inspirations or overlooked items. We don’t expect deep insights here, but deduction and reasoning are among AI’s strengths.
Good examples are important for AI to leverage its deduction capabilities fully. For instance, having a strong example from project A can effectively inspire the AI to double-check and apply similar logic to project B.
By implementing these strategies, we can better manage our workload and ensure nothing crucial slips through the cracks.
It’s probably not a good idea to use GenAI to draft emails, as AI is not yet reliable for handling communications (its wording often appears sophisticated but usually lacks emotional impact). However, it can be a valuable tool for summarizing or keeping track of who to communicate with and on which topics.
We always have the option to dive deeper to identify specific blockers that prevent our strengths from becoming more scalable.
How to make the blocker more acceptable?
Hint: potential solutions and non-solutions:
Your Own Prioritizer: Delegate the first round of prioritization to AI. For example, AI could read Slack messages, emails, and even meeting transcripts (from Zoom or other tools). As builders, we can set this up easily. The AI then notifies us of important messages and keeps others for consolidated processing. We’ll use this for the case study.
Use Zoom AI Companion for Meeting Summaries: Utilize AI to generate meeting summaries and multitask during less critical meetings. While this carries some risk, it’s possible to find a sweet spot where the benefits outweigh the downsides.
Avoid Delegating Critical Thinking to AI: Relying on AI for insights is not advisable, as AI is not particularly strong in this area. Additionally, this would undermine your core competency.
By adopting these strategies, we can make blockers more manageable and ensure that our key strengths remain scalable and effective.
Homework
==================================================
In your role, what is your own core competency, and how can you make them more scalable and effective?
In your role, what is your own limiting factor, and how can you make them less time consuming and more hands off?
==================================================