Design as Engineering,
Engineering as Design 8 个信号
设计即工程，工程即设计

AI 工具大幅降低了编码成本，代码从稀缺资源变成了低成本工具。"你能做我的工作，我也能做你的工作"成为常态。腾讯的分享提到：四千名设计师与大量前端之间的角色边界"越来越模糊"，初期有过混沌期，但现在每个人已在组织中找到新定位。 AI slashed the cost of writing code — code went from scarce resource to cheap tool. "You can do my job, I can do yours" became the norm. As one Tencent voice noted: with thousands of designers and many front-end engineers, the boundaries kept blurring — chaos at first, but everyone eventually found a new place.

传统"设计完成 → 交付开发"的串行模式正在退场，约七成团队转向并行推进。设计团队也在重新分工：聚焦高影响力项目，把标准化 UI 交给模板，用"Office Hour"式咨询替代事事亲为。 The serial "design then hand off to dev" flow is fading — about 70% of teams now run in parallel. Design teams re-divide labour too: focus on high-impact work, leave standard UI to templates, and replace doing-everything with office-hour style consulting.

当一切都追求"更快"，方向感反而稀缺。一位参会者的比喻最为传神—— When everything chases speed, direction becomes scarce. One attendee put it vividly —

共识是：管理者模型正在改变。从"上传下达的中间人"，转向战略制定者 + 方向引导者 + 团队赋能者。招聘标准也随之变化——好奇心、适应力、判断力与谦逊，比纯技术能力更重要。 The consensus: the manager model is changing. From "middle-man relaying orders" to strategist + direction-setter + team enabler. Hiring shifts too — curiosity, adaptability, judgment and humility now matter more than raw technical skill.

AI 产品理念经历了三次跃迁。核心目标是让 AI 像队友一样自然，把用户从"监工"的位置上解放出来——而非让他时刻干预。 AI products went through three leaps. The goal: make AI feel like a natural teammate, freeing users from the role of supervisor — not forcing them to intervene at every step.

设计原则上，要极简上手、逐步进阶，并按用户技术背景做交互分层。信任靠透明度建立：让用户看清 AI 在做什么，从"先审批再执行"的场景入手，逐步增加自主性。最难的坎在于——连自家员工都对产品极为挑剔。 By principle: minimal to start, progressive to master, with interaction layered by skill level. Trust is built on transparency — show what the AI is doing, start from "approve then execute," and grow autonomy gradually. The hardest part: even their own staff are fiercely picky about the product.

一个关键案例：某团队内部一个非技术部门的 AI 工具使用率，从 20% 飙升到 87%。秘诀不是更好的工具，而是两名专人深入各部门、用 Office Hour 与一对一沟通手把手赋能。 A telling case: a non-technical department inside a tech company saw AI adoption jump from 20% to 87%. The secret wasn't a better tool — it was two dedicated people embedding into departments, coaching one-on-one through office hours.

路径上，产品先吸引开发者，再推出 IDE 插件触达更广团队——一种主动"自我颠覆"的渗透。结论很直接： The path: win developers first, then ship an IDE plugin to reach wider teams — a deliberate act of self-disruption. The conclusion is blunt:

软件设计逻辑正在反转：从"教人用工具"，走向"工具主动适配人"。 The logic of software design is inverting — from "teach people to use tools" toward "tools that adapt to people."

但仅靠开放能力只能满足极客；要赋能整个团队，仍需把能力打包成针对场景的"交钥匙工程"。ToB 落地不能只靠模型的黑盒魔法。 Yet raw capabilities only satisfy the geeks; empowering a whole team still needs capabilities packaged into turnkey solutions for real scenarios. B2B can't rely on black-box magic alone.

用"组织战略"与"个人实践"两个维度，可以把组织的 AI 现状分为四个象限——每个象限都有对应的行动策略。 Mapping "org strategy" against "individual practice" sorts organizations into four quadrants — each with its own playbook.

人才评估的维度变了：不再看代码产出量，而看代理权（Agency）、问责（Accountability）、领域专长与判断力。这正是从 Agent 到 Agency 的真正含义——工具再强，也得靠懂行的人来驾驭。 What we measure has shifted: not lines of code, but agency, accountability, domain expertise and judgment. That is the real meaning of "from Agent to Agency" — however strong the tool, it takes someone who knows the craft to steer it.

团队结构也在动态演进：初创期需要全能型人才，随产品壮大必然走向专业化分工。核心原则是"小团队大责任"——规模尽可能小，以最大化个体权责集中度，倒逼产出质量。 Team shape evolves too: startups need generalists; as products grow, specialization follows. The core rule — small teams, big ownership — keep teams small to concentrate responsibility and drive quality.

领导者的新课题：关注团队如何共同推理与决策，而非个人的 AI 使用量；并留出时间进行集体思考、冲突解决与决策。拥抱"我也在边学边做"的谦逊，对建立信任至关重要。 The leader's new task: focus on how the team reasons and decides together, not individual AI usage; make room for collective thinking, conflict resolution and decisions. Embracing "I'm learning too" builds the trust that holds it together.

设计需兼顾行业差异：医疗等高风险行业需要严谨、确定性的流程图设计；创意行业则可采用更灵活的非确定性流程。 Design must respect industry differences: high-stakes fields like healthcare need rigorous, deterministic flows; creative fields can afford looser, non-deterministic ones.

在 AI 产品中融入乐观、趣味与愉悦感非常重要——通过微交互、自定义颜色等方式实现。同时，AI 并非要取代传统软件：过去几十年软件门槛过高，AI 的核心价值是大幅降低构建与修改的成本。 Bringing optimism, playfulness and delight into AI products matters — via micro-interactions, custom colours and the like. And AI isn't here to replace traditional software: software has been too hard to build for decades; AI's core value is making it far cheaper to build and change.