A working notebook on agents, human agency, and the tempo of organisations. It starts at the Microsoft Work Trend Index — and then looks wider, because no single report holds a current this big. 一份关于智能体、人的能动性与组织节奏的开放笔记。它从微软《工作趋势指数》出发，再往更远处看——因为这么大的一股水流，没有哪一份报告装得下。
Point of departure → 出发点 → Microsoft WorkLab · Work Trend Index 2025
Numbers are not the story. They are the current showing on the dial. Every figure below is sourced, and every source is listed at the end so you can go past this page. 数字不是故事本身，只是刻度盘上显示出来的水流。以下每个数字都注明出处，完整来源列在文末——你可以顺着它们走得更远。
The average gap between interruptions during core working hours — roughly 275 a day. Attention is now the scarcest material in the building. 核心工作时段里，两次被打断之间的平均间隔——每天约 275 次。注意力已成为这栋楼里最稀缺的材料。
of workers and leaders worldwide say they lack enough time or energy to do their job. The demand curve crossed the human one a while ago. 全球受访员工与管理者中，有八成表示没有足够的时间或精力完成本职工作。需求曲线早已越过了人的曲线。
of leaders call this a pivotal year to rethink strategy and operations — not a year to optimise the old shape. 八成以上的管理者认为，今年是重新思考战略与运营的关键一年——而不是把旧结构再优化一遍的一年。
expect agents to be moderately or extensively part of their AI strategy within 12–18 months. That is next spring, not next decade. 八成管理者预计在 12–18 个月内，将智能体较大程度地纳入公司的 AI 战略。那是明年春天，不是下一个十年。
new jobs projected by 2030 against 92 million displaced — a net gain that will not land evenly, and 39% of today's skills change on the way. 预计到 2030 年新增 1.7 亿个岗位，同时 9200 万个岗位被替代——净增长，但落点并不均匀；沿途还有 39% 的现有技能会改变。
productivity lift for the least experienced support agents using an AI assistant, against 14% on average. The current lifts the newest boats highest. 在使用 AI 助手后，经验最少的客服人员效率提升 34%，整体平均为 14%。水流把最新的船抬得最高。
Stewart Brand drew civilisation as layers — fast on top, slow underneath — where the quick layers learn and the slow layers remember. A workplace has the same anatomy. Most of the ache of this moment is a mismatch of tempo, not a failure of will. 斯图尔特·布兰德把文明画成层——上层快，下层慢；快的负责学习，慢的负责记忆。职场有同样的解剖结构。此刻大部分的痛，来自节奏错位，而不是意愿不足。
The top layer is already sprinting. The bottom layer is where meaning is kept. Both are correct — and a company that only accelerates the top of the stack starts to shear. 最上层已经在冲刺，最下层保存着意义。两者都没错——只加速顶层的组织，会从中间被撕开。
The Work Trend Index describes the shift in three phases. Read it less as a roadmap and more as a river with three reaches — most organisations have parts of themselves in all three at once. 《工作趋势指数》把这场转变分成三个阶段。与其当成路线图，不如当成一条河的三段河道——多数组织的不同部分，此刻同时处在三段之中。
Everyone gets an assistant. The work keeps its human shape; the friction simply drops. This is the reach most people are standing in right now. 每个人都拥有一个助手。工作仍保持人的形状，只是摩擦变小。这是当下大多数人所站的那段河道。
Agents join as digital colleagues, taking whole tasks under human direction. 46% of leaders say they already use agents to fully automate a workflow. The org chart grows a new column. 智能体作为数字同事加入，在人的指挥下承担完整任务。46% 的管理者表示已用智能体全面自动化某条流程。组织结构图长出了新的一列。
People set direction, judge quality and answer for outcomes. In the report's phrase, everyone becomes an "agent boss" — the job title stays, the verb underneath it changes. 人设定方向、判断质量、为结果负责。用报告里的说法，每个人都成了"智能体的老板"——职位名不变，底下那个动词变了。
In the same survey, 33% of leaders are considering headcount reductions while 78% are considering hiring for new AI-specific roles. Both sentences are true at the same time, and that is exactly what a current looks like from inside it: water going two ways around the same stone. 在同一份调研里，33% 的管理者在考虑裁减人手，同时 78% 的管理者在考虑为全新的 AI 岗位招人。两句话同时为真——这正是身处水流之中的样子：水绕着同一块石头，往两个方向走。
The slow layers do not speed up because the tools did. Judgement, accountability, the reason a decision was made, the patience to teach someone who is new — these are ballast, not drag. They are what keeps a fast thing upright. 慢的那几层，不会因为工具变快就跟着变快。判断力、责任归属、一个决定为什么被做出、教会新人的耐心——这些是压舱物，不是阻力。它们让快的东西不至于翻。
So "change management" is the wrong instrument. Change implies an ending: a before, a switch, an after. What is in front of us has no after. It has a heading and a rate. 所以"变革管理"是错的量具。改变意味着结束：之前、切换、之后。而眼前这件事没有"之后"，它只有方向和速率。
Open research only works if you can walk back to the well. Each link below is the primary source for a claim on this page, or the next good place to keep learning. 开放研究成立的前提，是你能沿路走回水源。以下每个链接，要么是本页某个论断的原始出处，要么是继续学习的下一站。