Design as Engineering,
Engineering as Design 13 个信号
设计即工程，工程即设计

一份最新研究显示：开发者参与设计的比例在一年内增长了 50%，设计师参与开发的比例近乎翻倍。近八成产品构建者表示，至少一半的工作发生在设计 Canvas 中；五分之一的开发者更愿意从 Canvas 而非代码开始工作。有参会者提到，所在组织拥有数千名设计师与大量前端工程师，角色边界"越来越模糊"——初期经历了混乱期，但现在每个人已在组织中找到新定位。AI 工具让代码从稀缺资源变成了低成本工具，研发逻辑彻底反转。 A recent study shows: developers designing grew 50% year over year; designers coding nearly doubled. Nearly 8 in 10 product builders say at least half their work happens in the Canvas; 1 in 5 developers prefer starting in Canvas over code. One attendee noted their org has thousands of designers alongside many front-end engineers, and boundaries have been "blurring more and more" — chaos at first, but everyone has now found a new place. AI made code go from scarce resource to cheap tool; the R&D logic has inverted entirely.

传统"设计完成→交付开发"的串行模式正在退场，约 70% 的团队已转向并行工作方式。设计团队的策略也在分化：一种做法是让设计师只聚焦"高影响力项目"（大方向探索、发现性项目），将表单和标准化 UI 交给模板系统，非设计人员自行取用。设计团队转而提供"Office Hour"式咨询——"你把你的方案带过来，我们给你反馈"——而非事事亲为。 The serial "design → hand off to dev" flow is fading; about 70% of teams now work in parallel. Design teams have split strategies too: some focus designers only on high-impact work (horizon exploration, discovery), leaving standard UI and form fields to templates that anyone can use. Design then shifts to office-hour style consulting — "bring your fork here, we'll give you feedback" — instead of doing everything.

但速度也带来代价。当一切追求更快，你会失去对齐预期的时间、失去记录变化的时间。关键问题变成了——在所有人都能写代码的时代，"代理权与责任感"取代代码产出量成为核心衡量标准。过去因代码昂贵而限制非技术人员写代码的旧模式已经瓦解，但"谁写的谁负责"变得无处不在。 But speed has a cost. When everything chases faster, you lose time for clarity — time to align expectations, time to document what changed in the past year. The key question shifts: when anyone can write code, agency and accountability replace lines-of-code as the core metric. The old model of restricting non-engineers from writing code (because code was expensive) is gone — but "you build it, you own it" is now everywhere.

当 AI 让一切加速，方向感反而成为最稀缺的资源。一位参会者的比喻最为传神——"你喊'跑'，团队就跑了。可当我问'往哪跑？'，他们说：'你没说，你只让我跑。'"速度天然会吞噬清晰度：你失去对齐预期的时间、失去记录变化的时间、甚至失去判断"跑对了没有"的时间。 When AI accelerates everything, direction becomes the scarcest resource. One attendee put it vividly — "You say 'run,' and they run. But when I ask 'where to?' they say, 'You didn't say — you just told me to run.'" Speed naturally eats clarity: you lose time to align expectations, to document what changed, even to check whether you're running the right way.

IC（个人贡献者）的角色固然模糊了，但管理者的角色比以往更加模糊。团队成员期望管理者成为"岩石"——提供稳定性和方向感，即便管理者自己也在边学边做。过去你靠"比团队更懂"来建立权威，现在你需要靠"我也在学习"的谦逊来建立信任。 IC roles are blurring, yes — but the manager role is even more ambiguous. Teams look to their leaders to be the rock, the stability, the constant in times of change. Yet the reality is: we don't know what's going on either — we're learning on a dime too. Authority once came from knowing more than the team; now trust comes from admitting you're learning alongside them.

招聘标准也在改变。现在你招的不只是一个能解锁团队的人，而是一个能解锁整个组织的人。要的素质是好奇心、适应力、良好的判断力，以及对周围人的信任与谦逊。设计师不进生产代码、工程师不代做设计决策——知道边界在哪，恰恰是成熟的表现。 Hiring standards have shifted too. You're no longer hiring someone who just unlocks the team — you're hiring someone who unlocks the organization. The traits: curiosity, adaptability, good judgment, and the humility to trust the people around you. A designer who stays out of production code, an engineer who doesn't override design decisions — knowing your boundary is a sign of maturity, not weakness.

会议中有一个专门的问题环节："在 AI 时代，管理者是更有价值，还是更不必要？"讨论触及了组织架构中敏感的一层——当 AI 可以产出策略分析、优先级排序、技术方案时，纯"中间层"的传话价值被严重质疑。有参会者直言："AI 让写代码变便宜了，但也让信息传递成本显得格外高昂。" A dedicated question was posed: "In an AI era, are managers more valuable or less necessary?" The discussion hit the most sensitive layer of org structure — middle management. When AI can output strategy analysis, priority rankings, and technical plans, the pure "relay" value of middle layers is questioned. As one attendee put it: "AI made code cheap, which makes the cost of middle-layer information-passing look especially expensive."

但另一面的声音同样强烈。有参会者指出，高层不断抛来兴奋的新想法——"我们得做这个，我们得做那个"——管理者需要判断什么真正值得投入，保护团队免于被信息淹没。知道"何时加速、何时减速"，是从高层向下过滤信息的守门人角色。还有人提到，在剧烈变化的时代，团队成员期望有人成为"岩石"——提供稳定性和方向感。即便管理者自己也在边学边做，这种"我也在学习"的谦逊本身就能建立信任。 But the counter-argument was equally strong. Leadership keeps throwing exciting new ideas at teams — "we gotta build this, we gotta build that" — and managers need to judge what's truly worth pursuing, protecting teams from being overwhelmed. Knowing when to speed up and when to slow down is a gatekeeper function that filters from the top. Others noted that in times of radical change, teams look for a "rock" — stability and direction — and the manager's willingness to say "I'm learning too" itself builds trust.

结论不是"要不要管理者"，而是"要什么样的管理者"。旧模型：管理者=上传下达的中间人。新模型：管理者=战略制定者+方向引导者+团队赋能者。关键转变在于——从管"人做什么"到管"人如何决策"。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 conclusion isn't "managers or not" — it's "what kind of manager". Old model: manager = middle-man relaying orders. New model: manager = strategist + direction-setter + team enabler. The key shift: from managing what people do to managing how people decide. Hiring follows: curiosity, adaptability, judgment and humility now matter more than raw technical skill.

AI 产品理念经历了三次关键跃迁。早期 AI 因缺乏实时交互与引导导致体验生硬，采用率极低。团队由此确立了"极简上手、逐步进阶"的设计理念——先用最低门槛解决"用起来"的问题，再逐步暴露高级能力。这本质上是用降低门槛来换取用户规模。 AI products went through three critical leaps. Early AI felt stiff because it lacked real-time interaction and guidance — adoption was painfully low. This drove the team to establish a "minimal to start, progressive to master" design philosophy — solve the "can I even use this" problem first, then progressively reveal advanced capabilities. It's essentially trading a lower bar for wider adoption.

核心目标很明确：让 AI 像主动提供帮助的队友一样自然。但最难的坎在于——连自家员工都对产品极为挑剔，不盲目使用。内部员工都不买账，外部推广更是难上加难。因此，交互模式必须根据用户技术背景进行差异化设计，兼顾专业开发者的自由度与普通用户的易用性。这本质上是针对技术门槛做的产品分层。 The core goal is clear: make AI feel like a proactive teammate, natural and helpful. But the hardest hurdle — their own staff are fiercely picky and won't blindly use it. If internal teams won't adopt, external push is even harder. So the interaction model must be layered by user skill level, balancing freedom for expert developers with ease for everyday users. It's product stratification by technical threshold.

信任靠透明度建立：让用户看清 AI 的处理过程和状态，从"先审批再执行"的场景入手，逐步增加自主性。在团队协作场景中，用户需要能清楚区分"这是人做的"还是"AI 做的"，便于追溯和回滚。可为 AI 分配身份（特定名称或角色），使其行为更可预测和可控。 Trust is built on transparency — show the user what the AI is doing and its current state. Start from "approve then execute" scenarios, then gradually increase autonomy. In team collaboration, users need to clearly distinguish "human" from "AI" actions for traceability and rollback. Giving the AI an identity (name, role) makes its behaviour more predictable and controllable.

一个反复被强调的关键案例：即便是 OpenAI，内部一个非技术部门的 AI 工具使用率，也从 20% 飙升到 87%。秘诀不是更好的模型——而是两名专人深入各部门，用 Office Hour、一对一沟通、手把手赋能，硬是把使用率拉了上去。连世界上最懂 AI 的公司，内部推广依然要靠人力布道和"保姆式"陪跑——服务与布道从来不是技术问题，是人的问题。 A case repeatedly highlighted: even at OpenAI, a non-technical internal department saw AI adoption jump from 20% to 87%. The secret wasn't a better model — it was two dedicated people embedding into departments, running office hours, doing one-on-one coaching, literally pulling adoption up by hand. Even the company that knows AI best still had to rely on human evangelism and hands-on hand-holding to drive internal adoption. Enablement isn't a technology problem — it's a people problem.

推广路径也有一套清晰的打法：先让产品成功吸引开发者——这批人是天然的意见领袖和布道者。站稳脚跟后，主动推出 IDE 插件等"自我颠覆"式产品，从极客圈层向大众团队渗透。这不是被动等待扩散，而是主动把产品送进更多人的日常工作流。 The adoption playbook is clear: win developers first — they're natural evangelists and opinion leaders. Once established, ship IDE plugins and other "self-disrupting" products, actively expanding from the geek circle into mainstream teams. This isn't passive diffusion — it's deliberately putting the product into more people's daily workflows.

结论很直接：公司必须摒弃单打独斗的"独狼"模式，重视团队协作。想靠个人硬啃 AI 工具行不通，抱团取暖才是破局关键。产品迭代能解决"能不能用"，但人工引导解决的才是"会不会用、愿不愿用"。 The conclusion is blunt: companies must abandon the lone-wolf approach and embrace team-based collaboration. Going it alone on AI tools won't work — huddling together is how you break through. Product iteration solves "can they use it," but human guidance solves "will they use it, and do they want to."

软件业正在经历根本性的转型。传统的软件开发模式是围绕特定产品和离散工作流设计的——用户需要主动学习和掌握工具，产品逻辑是"教人用工具"。新模式从"能力"出发，为用户提供可组合的通用能力，让 AI 能适应更有机、多样的使用场景。一句话总结：产品设计逻辑正从"教人用工具"向"工具主动适配人"转变。 Software is undergoing a fundamental transformation. The traditional model was built around specific products and discrete workflows — users had to actively learn and master the tool; the logic was "teach people to use tools." The new model starts from capabilities, offering composable general-purpose building blocks that let AI adapt to more organic, varied scenarios. In one line: the logic is shifting from "teach people the tool" to "the tool adapts to the person."

有参会者分享了从专注单一产品到接触企业业务的经历，坦言一线业务视角的切换直接刷新了对产品形态的固有认知——当你看到企业客户的实际工作方式时，那些预设的"标准工作流"假设会迅速瓦解。 One attendee shared their journey from focusing on a single product to engaging with enterprise customers, admitting that the frontline perspective directly refreshed their assumptions about product form — when you see how enterprise customers actually work, those preset "standard workflow" assumptions crumble fast.

但关键的限制在于：仅靠开放底层能力只能满足极客用户。若要赋能整个团队或企业，仍需将能力打包成针对特定场景的解决方案——标准化的"交钥匙工程"。ToB 落地不能只靠模型的黑盒魔法。 But the critical limitation: raw capabilities alone only satisfy geeks. To empower an entire team or enterprise, you still need to package capabilities into scenario-specific solutions — standardized "turnkey projects." B2B can't rely on black-box magic alone.

此外，Agentic 工具带来全新的设计挑战。不同于传统产品有可预测的用户路径，AI Agent 每次响应可能不同——即使输入相同的概念。这意味着产品设计不能再依赖"预设的用户路径"，而必须考虑多路径、非确定性的交互模式。这对设计思维和工程实现都提出了全新的要求。 Moreover, agentic tools introduce a fundamentally new design challenge. Unlike traditional products with predictable user paths, AI agents can respond differently each time — even with the same input concepts. This means product design can no longer rely on "preset user journeys" and must accommodate multi-path, non-deterministic interaction patterns. This places entirely new demands on both design thinking and engineering implementation.

用"组织战略投入"与"个人实践水平"两个维度，可以把组织的 AI 现状分为四个象限。这个框架的价值在于：不同象限需要截然不同的行动策略——你不能用统一象限的打法去推萌芽象限的团队。关键不是"哪个象限更好"，而是"你所在的象限下一步该做什么"。 Mapping "org strategy investment" against "individual practice level" sorts organizations into four quadrants. The value of this framework: each quadrant demands a fundamentally different playbook — you can't push a Nascent team with Unified tactics. The key isn't "which quadrant is better" but "what's the next move from where you are."

AI 时代对人才的定义正在重塑。不再以代码产出量为核心指标——事实上，并非全员都需要直接交付生产代码。新的衡量维度是代理权（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。AI 落地的核心门槛不是技术，而是"清晰传达需求所需的领域专长"。 The definition of talent is being reshaped in the AI era. Lines of code are no longer the core metric — in fact, not everyone needs to ship production code directly. The new dimensions: agency, accountability, domain expertise, and judgment. This is the real meaning of "from Agent to Agency" — however strong the tool, it takes someone who knows the craft to steer it. The core barrier to AI adoption isn't technical — it's "the domain expertise required to clearly articulate what you need."

个体差异必须被尊重：不同的人有不同的优势和热情所在，角色定位应顺应个人意愿——例如让热爱与客户沟通的人负责客户相关工作，而非强迫所有人走同一条技术路径。AI 已不再是锦上添花的噱头，而是产品迭代的必选项，但将新技术融入自身产品需要差异化竞争力——不是"用了 AI"就行，而是"用 AI 做出了什么不同"。 Individual differences must be respected: people have different strengths and passions, and roles should align with personal inclination — let the person who loves talking to customers do customer-facing work, rather than forcing everyone down the same technical path. AI is no longer a nice-to-have gimmick but a must for product iteration. Yet integrating new tech into your product requires differentiated competitiveness — it's not about "using AI" but about "what different thing you built with AI."

团队结构遵循清晰的演进规律：初创期需要全能型人才，每个人覆盖多个角色；随产品壮大，必然走向专业化分工。但无论哪个阶段，核心原则不变——"小团队大责任"。通过压缩人员规模来最大化个体的权责集中度，用责任倒逼产出质量。领导者的新课题也由此而来：关注团队如何共同推理与决策，而非个人的 AI 使用量或代码生成量。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——这对建立信任至关重要。 Team structure follows a clear evolution: startups need generalists covering multiple roles; as the product grows, specialization inevitably follows. But at every stage, the core rule holds — small teams, big ownership. Compress team size to maximize individual ownership concentration; let responsibility drive output quality. The leader's new task follows: focus on how the team reasons and decides together, not individual AI usage or code generation volume. Make room for collective thinking, conflict resolution and decisions. Embrace the humility of "I'm learning too" — this is critical to building trust.

设计风格需要兼顾不同行业的差异化需求：医疗等高风险行业需要严谨、确定性的流程图设计——出错后果严重，不能容忍不确定性；而创意行业则可采用更灵活的非确定性流程，给用户探索和发现的空间。不存在"一刀切"的 AI 产品设计美学。 Design style must respect industry differences: high-stakes fields like healthcare need rigorous, deterministic flows — the cost of error is too high for ambiguity. Creative industries can afford looser, non-deterministic processes that give users room to explore and discover. There's no one-size-fits-all AI product aesthetic.

在 AI 产品中融入乐观、趣味与愉悦感被反复强调为重要课题——可通过微交互、自定义颜色等方式实现。在团队协作场景中，还需设计清晰的"存在感"和"归属感"机制：用户应能清楚看到 AI 正在修改什么内容，并能区分人类还是 AI 的操作，以便追溯和回滚。 Bringing optimism, playfulness and delight into AI products was repeatedly emphasized as crucial — achievable through micro-interactions, custom colours and the like. In team collaboration, clear "presence" and "belonging" mechanisms are needed: users must see what the AI is changing and distinguish human from AI actions for traceability and rollback.

一个关键澄清：AI 并非要取代传统软件。有参会者明确驳斥了"AI 将取代传统软件"的悲观论调，直言自己绝非"SaaS 末日论者"。过去几十年软件开发门槛过高，而 AI 的核心价值是大幅降低构建与修改软件的成本。AI 不是来砸场子的，而是来给软件行业降本增效的。设计师应保持对产品品质的追求，利用现代工具提升效率的同时保持审慎——确保设计真正解决用户的真实问题，而非为 AI 而 AI。 A critical clarification: AI is not here to replace traditional software. One attendee explicitly pushed back against the "AI will kill SaaS" narrative, stating they are absolutely "not a SaaS doomsday-er." Software has been too hard to build for decades — AI's core value is dramatically lowering the cost of building and modifying software. AI isn't here to wreck the place; it's here to make software cheaper and better. Designers should keep pursuing quality, use modern tools to boost efficiency while staying deliberate — ensuring the design truly solves real user problems, not just "AI for AI's sake."

设计不仅是产出界面，更是改变组织文化的有力杠杆。有分享者讲述了所在大型企业（$630 亿收入、251 次收购）的转型故事：公司起初是典型的工程驱动文化——产品经理更像是项目协调人，设计师地位极低。变革从启动"设计评审"（design reviews）开始，最初遭遇巨大阻力。 Design isn't just about producing interfaces — it's a powerful lever for changing organizational culture. One speaker shared the transformation story of a large enterprise ($63B revenue, 251 acquisitions): the company started as a typical engineering-driven culture — PMs were more like program coordinators, designers had very low status. The change began with launching design reviews, which met massive resistance at first.

关键转折点：让一线个人贡献者（IC）参与设计评审，并对全员进行全栈教育——让每个人理解技术栈的不同层次。这增加了跨职能的同理心与理解力。设计师逐渐成为与工程师、产品经理平起平坐的平等伙伴，内部自发出现了"重新设计"的运动。公司还创办了名为 D-zone 的内部设计大会，面向全部 90,000 名员工开放，目标是"让每个人都像设计师一样思考"，在整个组织中培养设计素养与灵敏度。 The turning point: getting individual contributors involved in design reviews, and rolling out full-stack education — helping everyone understand different layers of the stack. This increased cross-functional empathy and understanding. Designers gradually became equal partners alongside engineers and product managers; an internal "redesign" movement emerged organically. The company also launched an internal conference called D-zone, open to all 90,000 employees, with the goal of "making everyone think like a designer" — cultivating design fluency and dexterity across the entire organization.

成果显著：所有产品（包括 251 次收购带来的产品线）现在共用同一套统一设计语言和统一管理页面。设计成为与工程、产品管理平起平坐的"第三条腿"。硬件设计中也有可持续性创新——例如 Webex 扬声器的面料包裹设计，一项小改动压缩了供应链周期并减少了多种颜色外壳和面料的需求。核心信念是：每一个细节都值得被设计。 The results are tangible: all products (including those from 251 acquisitions) now share a single, unified design language and a unified management page. Design became the "third leg of the stool," equal to engineering and product management. Hardware sustainability innovations followed — for example, the fabric wrapping on Webex speakers: one small design change compressed supply-chain cycle time and reduced the need for multiple colour cases and fabrics. The core belief: every detail deserves to be designed.

设计领导者面临的核心挑战：如何向企业证明设计不只是"好看"，而是直接驱动商业结果。产品构建的三个目标被清晰总结：① 打造人们热爱的产品（让客户愿意向朋友家人推荐）；② 实现规模采用（Adoption 是价值的代理指标——没有规模就没有价值）；③ 创建开放生态系统（产品应在开放生态中运行，而非封闭的公司专属体系）。 The core challenge for design leaders: proving that design isn't just "pretty" — it directly drives business outcomes. Three goals for product building were crisply summarized: ① Build products people love (so customers tell friends and family); ② Achieve scale adoption (adoption is a proxy for value — no scale, no value); ③ Create an open ecosystem (products should work in an open ecosystem, not a closed company-specific one).

核心洞察：能打动人心的产品卖得更多。情感吸引力直接转化为商业回报。设计领导者需要展示"好设计→商业模型"的连接，才能改变企业文化。但一个大前提是——如果老板不重视设计，可能很难推动变革。设计师应确保与老板在"通过设计打造伟大产品"的核心价值观上保持一致，否则不如换一个更设计友好的环境。 The core insight: products that appeal to people emotionally sell more. Emotional appeal translates directly to commercial returns. Design leaders need to show the "good design → business model" connection to shift corporate culture. But a big caveat — if your boss doesn't value design, it may be hard to drive change. Designers should ensure alignment with their boss on the core value of building great products through design — otherwise, find a more design-friendly environment.

更深层的问题：大公司常沉迷于商业数学，失去了产品的灵魂——脱离了"为什么存在"的初心和一线实际。领导者需要痴迷于创造令人惊叹的产品、用案例展示何为"在乎"、消除官僚主义、改变权力结构、给员工代理权去探索和辩论。而判断一家公司是否真正重视设计，一个细微指标就足够：看细节——那些客户不一定看得见但公司依然精心打磨的细节。 A deeper problem: large companies often get lost in business math, losing the soul of the product — disconnected from the "why we exist" purpose and the front-line reality. Leaders need to be obsessed with creating amazing products, show what "caring" looks like through examples, undo bureaucracy, change the power structure, and give employees agency to explore and debate in a safe space. And one subtle metric reveals whether a company truly values design: look at the details — the ones customers might not even see but the company still sweats over.

设计师在软硬件开发生命周期中的角色将发生重大变化——设计将在全流程中扮演更重要的角色，而不再局限于某个特定阶段。统一的设计语言变得至关重要，它是让整个组织围绕产品核心身份对齐的基础设施。 The designer's role in the software and hardware development lifecycle will change significantly — design will play a more material role throughout the entire process, no longer confined to a specific stage. A unified design language becomes crucial — it's the infrastructure that aligns the entire organization around the product's core identity.

工具（如 Figma）和设计 Agent 正在民主化设计：PM 和工程师都可以参与设计过程，更多人参与塑造产品的核心身份。这让设计师变得更有价值而非更不重要——因为设计师的角色从"唯一的设计产出者"转变为"设计系统的架构者和引导者"，能够调动更多人参与设计，放大设计的影响力半径。 Tools (like Figma) and design agents are democratizing design: PMs and engineers can all participate in the design process, with more people shaping the product's core identity. This makes designers more valuable, not less — because their role shifts from "sole producer of design output" to "architect and facilitator of the design system," engaging more people in design and amplifying its radius of influence.

最重要的洞察：随着 AI 和自动化增加，人的一面——判断力、品味、直觉——将成为更重要的瓶颈。与"AI 会减少设计岗位"的担忧相反，设计岗位反而会增加，因为对以人为中心的设计技能需求在持续上升。AI 不是替代设计师，而是让设计师从重复性工作中解放出来，把精力投入到更需要人的判断力、品味和直觉的地方。 The most important insight: as AI and automation increase, the human side — judgment, taste, intuition — will become an even more important bottleneck. Contrary to the fear that "AI will reduce design jobs," design jobs will actually increase because demand for human-centred design skills keeps rising. AI doesn't replace designers — it frees them from repetitive work so they can invest their energy where human judgment, taste and intuition matter most.

这是一条贯穿全场、却很少被直接说出口的暗线：在 AI 时代，真正的创新与洞察往往诞生于亲手实践的现场，而非会议室里的汇报与旁观。当技术每周都在变，二手信息的"保质期"急剧缩短——你听到的汇报，可能在传达的那一刻就已经过时。唯一可靠的体感，来自亲自把工具用起来、把东西做出来。 This is a quiet thread running through the whole roundtable, rarely stated outright: in the AI era, genuine innovation and insight are born from hands-on practice — not from reports in a meeting room or watching from the sidelines. When the technology shifts weekly, the shelf-life of second-hand information collapses — a report may be stale the moment it's delivered. The only reliable intuition comes from using the tools yourself and building the thing with your own hands.

多处论据指向同一结论：OpenAI 内部推广案例——采用率从 20% 跃升至 87%——靠的不是更好的模型，而是两个专人嵌入一线、手把手陪跑。连最懂 AI 的公司都要靠人力布道，更何况其他组织？洞察在陪练的过程中被亲手"摸"了出来。四象限模型也给出同样的处方：让团队从评估工具的"沙盒试点"转向在真实项目里动手的"流程试点"，因为只有真刀真枪地做，才会暴露真问题、积累真经验。 Multiple data points converge on the same conclusion: the OpenAI internal adoption case — 20% to 87% — wasn't about a better model, it was two people embedded on the front line, coaching hands-on. If even the company that knows AI best needs human evangelism, what does that say for everyone else? The insight was felt out through doing alongside others. The four-quadrant model prescribes the same medicine: move teams from sandbox pilots that merely evaluate tools to process pilots that put AI to work in real projects — because only real, in-the-trenches work surfaces the real problems and accrues real experience.

对管理者而言，这意味着权威的来源被改写。过去你靠"比团队更懂"来发号施令；如今，当连你自己都在边学边做时，"我也在亲手试"的躬身入局，比"我听了汇报"的居高临下更能建立信任。领导者的新课题不是关注个人的 AI 使用量，而是和团队一起推理、一起决策——这本身就是一种把手弄脏的实践。结论很朴素：想要新洞察，先动手；离实践越近，离真相越近。 For managers, this rewrites where authority comes from. You once gave orders by "knowing more than the team"; now, when even you are learning on the fly, the willingness to get your own hands dirty — "I'm trying it too" — builds more trust than the top-down "I read the report." The leader's new task isn't tracking individual AI usage but reasoning and deciding together with the team — itself a form of hands-on practice. The takeaway is plain: if you want fresh insight, start doing; the closer you are to practice, the closer you are to the truth.