Samsung, PepsiCo Use AI to Cut Product Prototyping Cycles to Minutes

三星与百事利用AI将产品原型周期压缩至数分钟 生成式建模与自动化仿真消除了繁琐的手工绘图工作，在降低企业试错成本的同时，推动产品设计师从资产创作者转变为战略把关人。 设计领域的行业领袖探讨了生成式AI与计算设计如何在确保物理制造可行性的同时，将产品迭代周期从数月压缩至数小时和数天。AI并未取代人类人才，而是承担了无附加价值的机械性繁琐工作，使设计师的核心职责转向战略性叙事、情感共鸣与严格的质量把控。结合消费品、汽车工程以及生物识别可穿戴设备等案例，专家组重点阐述了各机构如何利用嵌入式虚拟助手与快速数字仿真，大幅降低“学习成本”并拓展前期的设计探索空间。 达索系统通过将专用 AI 虚拟助手原生嵌入现有软件工作流，每周能帮助专业人员收回长达一天半非增值的手动操作时间。 (1:50) 百事公司借助由设计师和技术人员组成的跨职能攻坚团队，结合生成式 AI 仅用 3 周便完成了外部机构需耗时 12 周的原型项目，将原型反馈循环压缩至数小时至数天。 (3:20) 三星电子运用整合 3D/4D 人体生物数据、数字孪生与机器人测试的计算设计流程，将可穿戴硬件的仿真模拟与人体工学验证周期从 6 周大幅缩短至数分钟或数天。 (6:48) 在产品概念初期应用生成式 AI 能够以最大杠杆优化制造性能与成本；随着资产自动生成普及，设计师的核心职责正从基础像素绘制转向战略编排、叙事沟通与品质把关。 (12:36)

Across global enterprises, artificial intelligence is transforming product design by collapsing prototyping cycles and automating low-level production tasks, shifting design priorities from manual drafting toward upfront strategic orchestration. Industry leaders from PepsiCo, Samsung, and Dassault Systèmes argue that while AI accelerates execution speed, long-term product value remains anchored in physical manufacturing constraints, human empathy, and disciplined quality control. Compressing Iteration Cycles and Learning Costs (3:20) Generative AI is significantly reducing the time and financial investment required for early-stage conceptualization. Jordan Cooke reported that a cross-functional PepsiCo "tiger team" using Claude, GitHub Copilot, and Figma Make delivered a functional raw-materials negotiation prototype in three weeks—outperforming a twelve-week outsourced agency engagement at a fraction of the cost. Prototyping feedback loops contracted from months to single days. Similarly, apparel workflows at Newell Brands transformed an exhaustive six-to-twelve-month physical sampling process into a one-hour digital pipeline by translating 2D Illustrator tech packs directly into photorealistic, animated 3D models. Panelists emphasized that the core strategic benefit of accelerated iteration is lowering the organizational "cost to learn." At Samsung, computational design frameworks cut complex simulation and iteration timelines from six weeks down to days or minutes. Cooke explained that compressing turnaround times allows teams to explore a substantially wider concept aperture and stress-test risky hypotheses early, rather than locking in sub-optimal directions before downstream production begins. Grounding Generative Output in Physical Reality (19:24) While digital imagery can be generated without friction, industrial product development faces hard physical boundaries. Jeff Erno noted that generative renderings frequently fail to reflect the physics, tooling clearances, and economic constraints of traditional subtractive and injection molding manufacturing. For AI to deliver industrial value, it must integrate with enterprise knowledge graphs that cross-reference requirements, simulations, and physical validation data. As an example, generative software on an automotive assembly detected an tooling clearance conflict and relocated panel fasteners before physical metal was cut, preventing expensive manufacturing rework. To balance algorithmic speed with physical usability, Federico Casalegno detailed Samsung's computational design framework for wearable devices like Galaxy Buds. The team combines 3D and 4D biometric human capture, synthetic ear modeling, robotic cycle testing across thousands of physical iterations, and final live-user validation. This multiphase approach prevents ungrounded aesthetic concepts from bypassing ergonomic fit and sensor reliability requirements. Shifting Roles from Asset Production to Strategy (14:17) Because generative platforms can assemble UI layouts, documentation, and foundational code in seconds, the panel agreed that designers must transition from manual asset drafters into high-level strategists and quality gatekeepers. Cooke stated that as asset generation becomes commoditized across business units, the designer's primary differentiation lies in framing, organizational storytelling, and defending craft integrity. While AI serves as "rocket fuel" for individual productivity, leaders must accept relinquishing control over individual pixels in exchange for broader strategic influence. To define this human-centric boundary, Casalegno presented Samsung’s design formula: AI = EI + HI, asserting that artificial intelligence must combine emotional intelligence with human imagination. While machine models excel at synthesizing large datasets, they lack cultural nuance, empathy, and intuitive lifestyle understanding. Drawing a parallel to photography—where ubiquitous smartphone cameras have not eliminated the demand for skilled professional photographers—the panel maintained that authenticity, human taste, and deliberate craft remain irreplaceable enterprise assets. Embedding Domain Companions and Managing Adoption (36:10) Successful AI deployment requires integrating tools directly into existing environments rather than deploying isolated external apps. Erno highlighted industry data showing workers spend 1 to 1.5 days per week on non-value-added manual tasks. To reclaim this time without disrupting core workflows, Dassault Systèmes embedded specialized virtual companions directly into its design suite: "Leo" for mechanical engineering and 3D design, "Marie" for scientific computation, and "Aura" for enterprise governance. Addressing organizational adoption, Casalegno cited the distinction between Chronos (sequential operational speed) and Kairos (the qualitative, opportune moment for human judgment), reminding leaders that faster cycle times do not inherently guarantee superior design. The panel noted that when repetitive friction is automated away, designers naturally reinvest saved hours into evaluating alternative concepts rather than shortening project schedules. However, Cooke cautioned that as early compute subsidies expire, enterprises will face stricter scrutiny to prove that token and software expenses translate into tangible operational value. As synthetic generation becomes ubiquitous across enterprise workflows, the core challenge for product organizations is balancing rapid digital execution with physical, economic, and human-centered validation. Long-term competitive moats will depend not on the raw volume of automated output, but on how effectively human designers curate, ground, and guide AI within complex real-world constraints.

来自百事可乐（PepsiCo）、三星电子（Samsung Electronics）与达索系统（Dassault Systèmes）的设计领导者指出，人工智能正在深刻重塑产品研发流程：它不仅将原型迭代周期从数月压缩至数天乃至数分钟，还通过消除机械性非增值任务，促使设计师从底层的素材绘制者转变为战略架构师、跨学科协调者与品质把关人。 压缩迭代周期与降低前期试错成本 (3:20) 百事可乐全球UX、服务设计与研究负责人Jordan Cooke指出，AI最核心的价值在于“扭曲效率曲线”并大幅降低“试错成本”（cost to learn）。在百事可乐的一项采购谈判工具原型研发中，由跨职能人员组成的敏捷突击队（tiger team）借助Claude、GitHub Copilot与Figma Make，仅耗时3周便交付了可用原型，而此前外部代理机构完成类似任务需要12周，且成本高昂。这种模式将反馈循环压缩至小时级，团队在周一上午立项，周一下午或周二即可拿出可测试的实体或数字化原型，UX团队还利用合成角色（synthetic personas）加速定性探索。 类似的大幅提速也出现在消费品工业流程中。论坛主持人结合此前在纽威尔集团（Newell Brands）旗下户外品牌Marmot的实践补充道，传统服装样衣制作需要将二维工艺包交由工厂缝合打样并跨国邮寄，历经6至12个月的漫长周期；如今将二维图纸直接导入三维CAD系统后，团队在约1小时内即可生成高保真渲染，并完成20多种配色在虚拟走秀模特身上的即时动态模拟。Cooke认为，低廉的试错成本使团队能在早期快速验证或推翻高风险假设，避免了在错误路线上浪费资源。 计算模拟与物理制造约束的前置融合 (6:48) 三星电子执行副总裁Federico Casalegno分享了其在Galaxy Buds等可穿戴设备中应用的“计算设计”（computational design）框架。三星通过采集人体三维和四维生理数据构建高精度数字孪生模型，在软件中模拟数以百计的人体工学变体与传感器对齐，再将运动轨迹输入机械臂进行数万次自动化物理测试。该体系将以往高度依赖人工打样的复杂模拟流程从6周缩短至数天甚至数分钟，大幅加快了数据驱动的设计论证。 达索系统技术院士Jeff Erno则强调，数字像素的无限生成必须受制于物理制造规律与成本现实。在汽车“白车身”结构装配等工业场景中，生成式系统不仅能自动计算紧固件布局，还能结合装配工具的操作间隙，在上游排查出下游工厂数周后才会暴露的干涉错误。Erno指出，行业数据显示专业人员每周要耗费1至1.5天在无增值价值的手工琐事上；现代AI通过串联企业知识图谱（打通设计需求、仿真与测试记录），将传统耗时数小时的有限元分析（FEA）推向前置概念阶段，确保方案在早期就契合减材制造和注塑成型等重资产工艺要求。 从像素绘制转向战略叙事与品质把关 (14:17) 面对生成式工具能够瞬时输出逼真产品渲染图的现状，Jordan Cooke认为，未来5到10年内产品与体验设计师花在“抠像素”上的时间将显著减少，核心精力将转向商业价值、系统架构、跨部门协调以及围绕产品展开的叙事传达。虽然AI能为个人产出提供如“火箭燃料”般的杠杆效应，但人人皆可一键生成素材并不等同于具备美学价值与真实情感。正如智能手机普及后专业摄影师的构图与审美依然不可替代一样，严谨的人类技艺（craftsmanship）与真实意图仍是设计的根本壁垒。 Jeff Erno进一步表示，自动化工具的普及正促使设计师从基础制作者转变为战略编辑者与把关人。以往需要按住快捷键花费25分钟精细勾勒套索的繁重工作被单键消除，团队因而能将省下的精力投入到全局体验的审查上。Casalegno同样指出，当生成图像或视频只需“弹指一挥”时，设计方案背后的逻辑深度与跨部门沟通能力成为决定项目成败的关键，生成式叙事工具正是帮助设计师向工程与管理层精准传达设计意图的核心桥梁。 人本内核：情感智慧、专用伴侣与时间哲学 (16:41) Federico Casalegno强调，技术工具无论如何演进，设计团队必须始终坚持“围绕人类做设计，而非让人类适应设计”。他提出了三星内部遵循的核心公式：“AI = EI + HI”，即真正的人工智能产品设计必须融合情感智慧（Emotional Intelligence）与人类想象力（Human Imagination），以此跨越社会学、心理学与工程技术的边界。Casalegno援引古希腊哲学区分了线性的执行速度（Chronos）与做决定时的定性时机（Kairos），警示领导者不能将交付加速误认为设计品质的跃升。为掌握技术主动权，其团队过去两年已自主开发了15至20款针对特定设计工作流的专有AI程序。 在协同落地上，达索系统推行了渐进式的“虚拟伴侣”架构，而非推倒既有流程。Jeff Erno介绍，达索针对不同学科语境构建了具备专业身份特征的角色化助手——面向三维工程的“Leo”、面向科学研究的“Marie”以及处理企业运营治理的“Aura”，避免跨专业协作时的语境断层。专家们一致建议，初涉AI的企业应遵循从信息检索、项目级工作区到多文件智能协同的渐进路径，且不可因追逐技术热潮而降低伦理标准与人本设计原则。 该研讨会明确指出，评估AI设计成效不能单纯以工期压缩为单一指标，因为优秀的创作者往往会将省下的时间重新投资于更激进的形态探索与边缘性能打磨。然而，与会专家也发出预警：随着早期云厂商的算力补贴逐步退潮，未来模型调用与软件授权的实际开销将显著上升，企业必须保持理性克制，建立严格的ROI评估机制，确保AI投入能够转化为真实的商业与社会价值。

## Background 本内容由频道 Ai4 于 2026 年 8 月 21 日发布在 YouTube 平台上。视频记录了一场汇聚百事公司、三星电子与达索系统等行业领军设计主管的圆桌研讨会，探讨人工智能如何超越单纯的效率工具，在概念生成、快速原型、硬件人体工学模拟与跨职能协同等全流程中重塑产品设计范式。 ## TL;DR 论坛围绕 AI 对现代软硬件产品设计全流程的深刻改造展开探讨。百事公司分享了如何借助 AI 驱动的紧凑小团队在 3 周内完成原本需要外包 12 周的原型交付；三星电子展示了结合 3D/4D 人体数据与多层模拟算法的“计算设计”如何将复杂形态验证从数周压缩至数分钟；达索系统则回顾了从物理样机到虚拟 AI 伴侣的工程演进历程。三位专家一致认为，AI 能够剔除非增值琐事并大幅降低学习试错成本，但设计的核心始终在于“情绪智力与人类想象力”，设计师也将升级为把控战略方向与叙事体验的“主编”角色。 ## Atmosphere & style 整场圆桌讨论严谨务实且极富前瞻性，充满了顶尖工业设计与工程实践者之间的智力碰撞。嘉宾们没有沉溺于空洞的概念炒作，而是带着清晰的行业痛点与量化案例娓娓道来，既有对自动化提效的兴奋展望，也有对物理规律与人文价值的冷静坚守。现场氛围专业而开放，在轻松幽默的代际对比与深刻的哲学思辨交织中，呈现出一场兼具技术深度与美学情怀的高水平对话。 ## Key takeaways - [01:50](https://www.youtube.com/watch?v=OyCaNu3I034&t=110s) 达索系统的核心策略是将 AI 直接内嵌到日常设计与工程软件中，以无缝剥离非增值的重复性劳动，避免设计师在孤立的外部工具间频繁切换。 - [03:21](https://www.youtube.com/watch?v=OyCaNu3I034&t=201s) 百事公司通过组建跨职能敏捷突击团队并结合 Claude、Copilot 及 Figma 等生成式 AI 工具，仅用 3 周就交付了以往传统外包需耗时 12 周的智能谈判工具原型。 - [06:48](https://www.youtube.com/watch?v=OyCaNu3I034&t=408s) 三星电子开创“计算设计”学科，通过高精度人体动态建模与“数字虚拟仿真-机械臂动作复现-真人终验”三层闭环体系，精准优化可穿戴设备的贴合度。 - [09:22](https://www.youtube.com/watch?v=OyCaNu3I034&t=562s) AI 与强大算力将硬件人体工学形态计算从数周压缩至数分钟，推动硬件设计从传统经验试错迈向“科学驱动型创新”范式。 - [12:36](https://www.youtube.com/watch?v=OyCaNu3I034&t=756s) 机器学习模型的秒级推演替代了耗时数小时的传统有限元分析（FEA），让团队在早期概念阶段就能实时评估可行性并迅速收敛至全局最优解。 - [14:17](https://www.youtube.com/watch?v=OyCaNu3I034&t=857s) 设计师的职能重心正在从基础的“像素制作与搬运”脱离，结构性升级为战略连接者、体验叙事构建者以及工艺质量把关人。 - [16:41](https://www.youtube.com/watch?v=OyCaNu3I034&t=1001s) 三星提出“AI = EI（情商）+ HI（人类想象力）”的哲学公式，强调技术工具必须在以人为本的情感与文化洞察下运转，才能产生真正的设计价值。 - [19:24](https://www.youtube.com/watch?v=OyCaNu3I034&t=1164s) 物理世界的制造法则、模具加工与经济效益构成了生成式狂欢的“节流阀”，数字端生成的几何方案必须接受严密工程仿真与物理落地的校验。 - [25:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=1554s) 当 AI 使数字资产的生成门槛大幅降低后，设计的核心竞争力随之转移至作品背后的真实性（Authenticity）、精湛工艺与深层意涵构建。 - [28:47](https://www.youtube.com/watch?v=OyCaNu3I034&t=1727s) AI 真正消灭的是低价值的机械劳作，设计师正向统筹方向的“主编”角色演进，从而释放精力攻克绿色环保、无障碍包容性等长周期高维命题。 - [36:10](https://www.youtube.com/watch?v=OyCaNu3I034&t=2170s) 达索系统按工程设计（Leo）、科学研发（Marie）和商业治理（Aura）等垂直专业语境构建内嵌式虚拟 AI 伴侣，实现低干扰的人机协同。 - [37:08](https://www.youtube.com/watch?v=OyCaNu3I034&t=2228s) AI 在服装领域将打样试穿周期从数月压缩至 1 小时，在汽车白车身装配中实现紧固件自动排布与工装防碰撞，达成“首次即正确”的工程突破。 - [42:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=2574s) AI 赋能的本质并不是单向压缩项目交付天数，而是解放探索边界，让设计师在给定时间内开展更深度的创意发散与方案迭代打磨。 - [45:10](https://www.youtube.com/watch?v=OyCaNu3I034&t=2710s) 评估 AI 投入产出比（ROI）应聚焦于降低组织试错与学习成本（Cost to Learn），同时需警惕未来算力补贴退坡后工具使用成本上升的风险。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=OyCaNu3I034&t=4s) 嘉宾介绍与近期 AI 实践焦点 圆桌论坛在主持人的开场中拉开帷幕，探讨大型企业如何将人工智能实际落地于产品设计流程中。主持人首先邀请三位来自不同行业的领军人物介绍各自背景，并分享过去几周在 AI 领域的核心工作重点。 百事公司（PepsiCo）全球用户体验、服务设计与研究负责人 Jordan Cooke 率先发言。他指出，自己今年以来的核心思考始终围绕着如何将层出不穷的新型 AI 工具深度融入现有工作流中，以实现管理层所期望的“弯曲效率曲线（bending the efficiency curve）”，在提升产出的同时重塑工作模式。 随后，三星电子（Samsung Electronics）设计执行副总裁、三星旧金山设计创新中心（Samsung Design Innovation Center）及全球设计工作室负责人 Federico Casalegno 分享了自己的关注点。他强调，随着 AI 工具的爆发式增长，设计团队面临的核心课题在于如何在运用新技术的同时坚守“以人为本的设计（human-centered design）”，让设计始终扎根于人类体验与人性化诉求。 达索系统（Dassault Systèmes）技术 Fellow Jeff Erno 随后致辞。他笑称很高兴能与三星和百事这两家在产品设计领域极具盛名的企业同台。达索系统目前最聚焦的方向，是将虚拟助手/智能伴侣（Virtual Companions）直接引入全企业级产品链中，覆盖从前瞻概念设计到制造工程的完整生命周期。 - **Jordan Cooke**：百事公司全球 UX、服务设计与研究负责人，专注推动 AI 融入日常工作流与效率提升。 - **Federico Casalegno**：三星电子设计执行副总裁，主导旧金山创新中心，探索 AI 时代下人本设计的深化。 - **Jeff Erno**：达索系统技术 Fellow，致力于在全流程工程与设计软件中内嵌 AI 虚拟助手。 --- ### [01:50](https://www.youtube.com/watch?v=OyCaNu3I034&t=110s) 将 AI 深度嵌入日常工具：剥离非增值工作 主持人就实际落地问题向 Jeff Erno 提问，探讨达索系统在推动哪些具体的 AI 项目以及它们如何重塑产品研发管线。Jeff 指出，团队目前最核心的策略是“将人工智能直接内嵌到用户日常使用的软件中”，而不是要求设计师或工程师在日常任务之外再去打开一套互不相连的独立 AI 工具。这种无缝嵌入的方式能最大程度降低对原有工作流程的干扰。 Jeff 结合自身早年做设计师、工程师以及后来开发与销售软件解决方案的复合经历谈到，设计与工程人员在日常工作中往往耗费了海量时间在低价值的重复劳动上。行业研究表明，知识工作者每周有多达 1 到 1.5 天的时间被这些“非增值工作（non-value-added work）”吞噬；而且这还低估了许多原本被误认为是“必要增值”、但理论上完全可以通过自动化消除的手动流程。 当前的 AI 工具正是为了将这些被琐碎任务侵占的时间“重新夺回来”。Jeff 感慨道，现在团队所打造并推向市场的 AI 辅助能力，正是他当年在一线做设计和写代码时梦寐以求的利器。 --- ### [03:21](https://www.youtube.com/watch?v=OyCaNu3I034&t=201s) 百事公司的敏捷攻坚：从 12 周外包到 3 周交付 当主持人询问现场观众中有多少设计师和工程师并邀请嘉宾分享 AI 如何加速早期产品创新时，Jordan Cooke 分享了百事公司内部刚刚完成的一项概念验证（POC）项目。 百事组建了一支被称为“老虎队（Tiger Team）”的小型专项突击团队。这支团队打破了传统僵化的职能边界，不拘泥于特定角色分工，而是由创意技术专家（Creative Technologist）、设计师、设计工程师和用户研究员等复合型实干家组成。在短短 3 周时间内，该团队便快速完成了多轮原型迭代，最终交付了一套功能完备的内部智能谈判工具（Internal Negotiation Tool）工作原型，用于辅助百事在日常大宗原材料采购中进行更具智能化的谈判决策。 Jordan 对比指出，这支紧凑的内部小团队在 3 周内交付的高质量产出，远超此前某外包团队花费 12 周且耗资巨大的成果。他直言，设计的精髓在于迭代，“设计就是看我们能在有限的时间内将一个方案演进多少轮，从而向世界呈现出真正美妙的产品”。利用生成式 AI，团队如今可以在周一启动探索，当天或周二就拿出一套可交互、可评估的原型来获取反馈并即时迭代，将整个迭代周期压缩到了极致。 在工具链方面，百事团队重点依赖 Claude 与 GitHub Copilot 进行代码与逻辑搭建，在设计端使用 Figma（含 Figma Make），同时在用户研究环节积极尝试合成用户画像（Synthetic Personas）等新型 AI 工具，以大幅提速前期洞察。 ```mermaid flowchart LR accTitle: 百事公司敏捷 POC 与传统外包流程对比 subgraph 传统外包模式["传统外包研发流程（耗时 12 周）"] A1["繁杂角色分工"] --> A2["多层级漫长沟通"] --> A3["成本高昂 / 产出有限"] end subgraph 敏捷攻坚模式["百事 Tiger Team + 生成式 AI（耗时 3 周）"] B1["复合型极简团队<br/>(设计/工程/研究/技术专家)"] --> B2["周一启动探索 / 当天或次日出原型"] B2 --> B3["借助 Claude / Copilot / Figma 快速迭代"] B3 --> B4["高效交付智能谈判工具原型"] end ``` --- ### [06:48](https://www.youtube.com/watch?v=OyCaNu3I034&t=408s) 三星计算设计：3D/4D 人体建模与三层仿真闭环 Federico Casalegno 随后介绍了三星在实体硬件产品设计中的 AI 应用。数年前，三星设计团队便开辟了一门名为“计算设计（Computational Design）”的新学科，将人工智能、机器学习与先进计算技术深度融合于硬件研发之中。 在可穿戴设备（如 Galaxy Buds 耳机、Galaxy Watch 手表、Galaxy Ring 智能戒指）的研发中，三星的核心目标是打造极致的人体工学贴合度与传感器精准度。团队首先采集高精度的真实人类 3D 及 4D 动态形态数据（例如耳朵与手部的微观几何特征与运动轨迹），设定关键绩效指标（KPI），并在此基础上训练专属算法模型，随后进入高强度的多维模拟与迭代流程。 Federico 详细拆解了三星独特的三层验证体系： 1. **数字合成环境**：在虚拟空间中高保真复制人体耳道与手部形态，全天候模拟成千上万种产品变体的贴合状态； 2. **机器人动作仿真**：捕捉真实人类使用产品时的运动轨迹，并将其转化为机械臂的物理动作，由机器人进行机械化、高重复性的物理佩戴与受力模拟； 3. **真实用户测试**：在前期虚拟与机械验证收敛后，最后交由真实用户进行最终体验与佩戴测试。 ```mermaid flowchart TD accTitle: 三星计算设计与产品研发流程 subgraph Data["数据采集与模型构建"] A["高精度采集真实人体 3D / 4D 动态数据"] --> B["设定研发 KPI 并训练专属模型"] end subgraph Sim["三层闭环验证体系"] B --> C["1. 数字虚拟仿真<br/>(高频计算数千种产品形态变体)"] C --> D["2. 机械臂动作模拟<br/>(捕捉真实动作并在机器人上复现)"] D --> E["3. 真实用户终验<br/>(实体样品真人佩戴测试)"] end E --> F["输出更舒适贴合的 Buds / Ring / Watch 硬件"] ``` --- ### [09:22](https://www.youtube.com/watch?v=OyCaNu3I034&t=562s) 从数周缩至数分钟：科学驱动的硬件设计范式转变 当主持人问及目前的设计流程与 5 年前相比最大的本质变化时，Federico 强调，现代设计已经迈入了“科学驱动型创新（science-driven innovation）”时代。三星所使用的数据均来自严格筛选、高度精选的真实人体样本，保证了底层数据的可信度；而 AI 强大的算力则使得在极短时间内计算海量数据点成为可能。 在过去，某些复杂的形态模拟与人体工学计算需要耗费整整 6 周时间；在引入 AI 计算流程后，这一周期先是缩短到两三天，如今更是被压缩至“数分钟”之内完成。除了时间维度的量级跃迁，这种快速生成与验证能力还彻底改变了团队协作模式——工程师与设计师能够在掌握充分科学数据的基础上展开对话，极速围绕产品变体进行迭代决策，从而大幅降低研发风险。 --- ### [10:49](https://www.youtube.com/watch?v=OyCaNu3I034&t=649s) 从物理原子到虚拟比特：工程软件与迭代能力的演进史 主持人结合自身此前在 Newell Brands 的工业设计管理经验补充道，生成式 CAD 工具的普及让设计师不再需要因为时间限制而过早扼杀大胆的想法。以往团队面对大量手绘草图时，往往只能极为谨慎地挑选少数几款进行 CAD 建模和 3D 打印原型验证；而现在，设计师可以在极短时间内探索更多激进或具有潜在风险的设计构想。 Jeff Erno 从工业软件提供商的视角对这一演变做出了呼应。他回顾了工业工程探索产品迭代的历史轨迹： - **算盘与计算尺时代**：在计算机诞生前，工程师只能在现实物理世界中使用“物理原子”手动造样机并进行破坏性测试，整个产品生命周期中能进行的迭代次数极其有限； - **穿孔卡片与早期计算时代**：依然依赖耗时的大量人工数据准备； - **商业 CAD 与有限元分析（FEA）黎明期**：Jeff 入行时正值商用计算机辅助设计与仿真软件兴起。行业迎来了里程碑式的飞跃，研发人员不再需要频繁消耗物理原子，而是能够在计算机系统中调度“0 和 1 的虚拟比特”进行仿真测试，单产品的迭代次数从一两次大幅跃升到五六次； - **AI 与生成式优化时代**：如今通过内嵌 AI 与全局优化算法，迭代数量和探索维度被推向了全新的量级，设计与工程从单纯的“验证假设”全面演进为“由算法辅助生成最优解”。 ### [12:36](https://www.youtube.com/watch?v=OyCaNu3I034&t=756s) 早期设计优化与机器学习的提速力量 在传统的计算机辅助设计（CAD）与参数化建模流程中，专业门槛极高。设计师和工程师必须掌握复杂的参数化建模逻辑，而让一个模型在经历反复修改后依然保持参数关系的稳健、不发生崩溃，是极其艰难的专业技能。以往在评估设计时，即使经过了几十轮迭代，团队依然会存疑：这究竟是不是最优解？受限于计算成本与算力，过去执行一次详细的有限元分析（FEA）往往需要数小时。 如今，生成式设计技术与机器学习模型的结合彻底改变了这一现状。通过训练有素的机器学习模型来快速推演性能结果，工程师可以在极短时间内获得设计反馈，不必再苦等漫长的传统物理仿真。 ```mermaid flowchart TD accTitle: 早期设计流程从传统到AI驱动的转变 subgraph Traditional["传统早期设计模式"] T1["参数化 CAD 建模"] --> T2["耗时数小时的有限元分析 (FEA)"] T2 --> T3["局部、局限的设计微调 (迭代缓慢)"] end subgraph AIDriven["AI 驱动的新型设计模式"] A1["生成式设计拓展探索空间"] --> A2["机器学习模型秒级性能评估"] A2 --> A3["快速收敛至全局最优解 (降本提效)"] end ``` 这种提速在产品开发的最早期阶段尤为致命。产品性能与最终成本的最大决定因素正是在前端概念构思期确立的，越是在组织全面铺开细化工作之前做足前置验证，越能找到全局最优解，而不是仅仅在初始设计周围做局部微调。AI 技术不仅能激发全新的概念方案，更能实时评估可行性，让团队尽早拿到优质答案。 --- ### [14:17](https://www.youtube.com/watch?v=OyCaNu3I034&t=857s) 设计师角色的进化：从“像素搬运工”到战略连接者 主持人抛出一个现实场景：当营销人员或高管仅凭 Prompt 就能瞬间生成极其逼真、吸睛的产品渲染图，却完全不了解背后的制造工艺或客户体验（CX）全链路时，产品设计师在未来 5 到 10 年的定位到底是什么？百事公司的 Jordan Cooke 指出，设计师未来的核心精力将显著剥离基础的“像素制作”，全面转向价值定义、战略把控以及围绕产品体验的叙事构建。 在 Jordan 看来，无论是像素、代码还是文档，AI 工具都可以在数秒内批量产出，而且工具的能力只会越来越强。但工具终究只是工具，当一位富有才华的专业人才配上强大的 AI 工具时，就像注入了“火箭燃料”，个人的产出上限和视野将被无限放大。 ```mermaid flowchart LR accTitle: 设计师职能重心转移模型 subgraph Past["过去的核心职责"] P1["像素绘制与排版"] P2["基础代码编写"] P3["繁琐文档沉淀"] end subgraph Future["未来的核心职责"] F1["战略设计与价值挖掘"] F2["跨领域协作与连接"] F3["工艺质量把关与高维叙事"] end Past -.->|"AI 基础资产生成接管"| Future ``` 随着设计制作门槛的大众化与平民化，设计团队应当坦然接受“不再百分之百掌控每个像素”的现状，转而扮演更高维度的角色： - **质量捍卫者与把关人**：为设计工艺和品牌叙事设立极高的标准，确保所有最终输出的产品与体验符合团队定义的品质门槛。 - **促进者与战略家**：在组织内部充当跨学科连接的桥梁，用宏观战略统筹由工具高速生成的零散资产。 --- ### [16:41](https://www.youtube.com/watch?v=OyCaNu3I034&t=1001s) 三星的设计哲学：AI = EI（情商）+ HI（人类想象力） 来自三星电子的 Federico Casalegno 阐述了三星的设计演进路径。他强调，无论技术如何翻天覆地，设计的核心使命从未改变：理解人、理解文化、理解生活方式，并在深刻洞察的基础上，构想如何利用前沿技术为最终用户创造出色的产品与体验。“以人为本的设计（Human-centered Design）”原则依然牢固——“我们是围绕人类来进行设计，而不是让人类去适应我们的设计”。 在三星，设计团队将前沿科技与人文思考凝练为一条核心公式： $$\text{AI} = \text{EI (Emotional Intelligence)} + \text{HI (Human Imagination)}$$ 即**人工智能 = 情绪智力 + 人类想象力**。 Federico 解释道，AI 无法凭空替代人类去产生真实的同理心并洞悉复杂的社会文化。设计师的角色在今天不仅没有被削弱，反而变得更加核心。他们需要从社会学、心理学等多维度理解人类社会，并在快速变化的经济、工程与创新生态之间穿针引线，将抽象的情感和需求“实体化”为具体的硬件产品、数字体验或交互界面。 --- ### [19:24](https://www.youtube.com/watch?v=OyCaNu3I034&t=1164s) 现实的节流阀：为什么制造与物理法则无法被算法架空 主持人分享了一个关于摄影发展的量化对比：从 19 世纪摄影术发明到约 2015 年智能手机普及，人类在 150 年间拍摄的照片总数，如今全球智能手机用户每分钟就能拍出相当的数量。然而，尽管人人都是“摄影师”，专业摄影师与摄像师的价值依然不可替代。 达索系统的 Jeff Erno 对此表示赞同，并指出物理世界本身就是 AI 生成狂欢的“节流阀（Throttle）”： - **生成图像易，落地实体难**：数字图像只是简单的二进制数据（0 与 1），但在物理世界中制造产品，必须面对严苛的现实条件。 - **制造方式与经济效益**：尽管某些场景下可以采用 3D 打印快速成型，但大规模量产通常必须依赖传统的减材制造或精密模具加工，这意味着高昂的时间与制造成本。 - **打通企业知识图谱**：AI 可以辅助遍历企业庞大的数字资产和知识图谱，但最终设计必须与现实碰撞——从虚拟仿真到物理测试，严密校验各项工程参数与市场接受度。 --- ### [22:20](https://www.youtube.com/watch?v=OyCaNu3I034&t=1340s) 拥抱变革与坚守真实性价值 面对这场技术巨变，设计团队管理者正在从团队工作流与设计哲学上积极调整。Jordan 提到，设计行业本身的魅力就在于其永无止境的演进，优秀的设计师绝不会“躺在过去的功劳簿上不思进取”。 当任何人都能在几秒钟内用生成工具创作一首歌、开发一个 App 或生成一幅艺术画作时，海量内容的涌现并不代表高品质。人类用户天生渴望**真实性（Authenticity）**、**精湛的工艺（Craftsmanship）**以及**动人的故事（Storytelling）**。正如每个人口袋里都揣着高清摄像头，但人们依然愿意花重金聘请专业摄影师为全家拍摄珍贵的家庭合影一样，蕴含人类深度思考与情感连接的设计永远无可取代。 ### [24:21](https://www.youtube.com/watch?v=OyCaNu3I034&t=1461s) 工具与人的协同效应：打破创作门槛与效率跃迁 Jordan Cooke 谈到，尽管技术日新月异，人们每年依然在创作电影、艺术和音乐，人类的创造力从未消亡。关键在于如何把这些前沿工具交到对的人手中。他以音乐创作为例：过去一个在地下室怀揣音乐梦想的年轻人，可能根本没有预算雇佣一整支管弦乐团；而如今，借助 AI 工具，创作者能够将丰富的交响乐感融入作品中，依托现有条件成倍放大自身的能力。Cooke 对团队的核心要求与期望正是如此——探索如何利用工具武库与个人专业背景相结合，大幅提升产出上限。他坚信“人加上工具，其价值必定大于各部分的简单相加（The sum is greater than the parts）”，设计管理者的职责就是精准搭建人与工具之间的桥梁，实现更优质的业务成果。 主持人随即风趣地感叹时代变迁带来的“代沟”：“当我告诉年轻团队成员‘当年我想抠出一个人像时，得按住 Option 键、拿着套索工具沿着边缘小心翼翼地描上 25 分钟’时，我真的觉得自己老了；而现在的你们只需要轻点一个按钮。” ```mermaid flowchart LR accTitle: 人与AI工具的协同增效模型 subgraph S1["个体能力"] A["专业背景与审美"] end subgraph S2["AI 工具赋能"] B["自动化重复任务"] C["低成本能力扩展<br/>(如管弦编曲/即时抠图)"] end subgraph S3["综合成果"] D["1 + 1 > 2 协同增效<br/>聚焦高价值问题与意义构建"] end A --> D B --> D C --> D ``` --- ### [25:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=1554s) 自研 AI 体系、跨学科融合与设计“意义”的重构 Federico Casalegno 分享了三星设计团队在转型过程中的实践路径。他坦言管理的关键在于“招募极其优秀的人才，随后全力保护并赋能他们”。三星拥有一支小而精悍的高效计算设计（Computational Design）团队，在过去几年里，团队成员根据自身实践需求，亲手编写了 15 到 20 个定制化 AI 专属程序。他们并非被动接受现有技术，而是主动“弯折并驯服技术发展”，使其精准服务于具体的设计诉求。 Casalegno 进一步强调了数字化时代“真实性”与“意义（Meaning）”的演变。当今的技术使得打个响指就能瞬间生成一段视频，生产速度与基础质量不再是瓶颈，核心竞争力随之转移至作品背后的深层意涵。设计团队当下最迫切需要的是沉淀、讨论、思辨与推敲论点的“安全空间（Safe place）”。面对层出不穷的新产品与全新体验形态，设计界并没有现成标准答案，这属于一个全然开放的探索领域。为此，Casalegno 延续了他在麻省理工学院（MIT）媒体实验室（Media Lab）任教二十年的跨学科理念——打破单一学科壁垒，将不同专业背景的顶尖人才汇聚一堂共同碰撞，从而开辟创新的未知路径。 --- ### [28:47](https://www.youtube.com/watch?v=OyCaNu3I034&t=1727s) 破除失业焦虑：消除机械劳作与晋升“主编”角色 当主持人问及设计团队在落地 AI 时面临的最大阻碍是工具匮乏、技能短缺还是主观抵触时，现场指出设计师通常天生具备探索新工具的好奇心与敏感度，因此设计部门内部的阻力远低于其他业务板块。然而，业界普遍蔓延着一种“AI 是否会取代自身工作”的深层焦虑。 Jeff Erno 对此进行了针对性回应。他指出，AI 真正消灭的是那些“低价值、高度重复的枯燥苦活（Drudgery）”。在职业生涯中，他从未见过有人会抱怨“原本要耗费我整整两天的重复机械操作，你居然用 5 分钟就帮我搞定了”。随着底层低效劳动的自动化，设计师的角色正在发生结构性跃升——就像从撰写具体稿件的撰稿人升级为统筹全局的“报社主编”。当设计师从基础事务中抽身，便能跨越低垂的果实，将精力投向需求金字塔更高阶的复杂命题，例如绿色环保科技、无障碍设计（Accessibility）等以往因工期有限而无暇顾及的领域。 - **消除低价值消耗**：自动化处理机械式劳动，团队对效率工具的接受度会自然提升。 - **角色向“主编”转型**：设计师从事无巨细的基础产出者转变为把控方向、筛选质感与全局判断的审阅者。 - **攻克更高维度挑战**：释放出的脑力与工时可用于攻克无障碍包容性、可持续绿色技术等长周期高价值课题。 --- ### [31:15](https://www.youtube.com/watch?v=OyCaNu3I034&t=1875s) 设计负责人的落地指南：心态重构、循序渐进与坚守设计准则 面对尚未深度引入 AI 的设计管理者，讨论嘉宾给出了一套由浅入深的实操战略建议： 1. **破除恐惧与开放心态**：Jordan Cooke 强调，管理者首先要鼓励团队积极上手实验，寻找 AI 增强个人专长与热情的切入点。切忌因恐惧被取代而盲目闭目塞听（Put blinders on），唯有保持开放与持续学习，才能找到契合团队架构的演进道路。 2. **超越速度的心智转型**：Federico Casalegno 指出，除了企业层面必须解决的安全合规、技能培训与高昂工具成本等基础问题外，最根本的变革在于“思维方式（Mindset）”。他援引古希腊语中关于时间的两个词汇： - **Chronos（时序时间）**：代表客观时间的流逝与任务节奏的物理缩短。 - **Kairos（时机/决断时刻）**：代表做出关键决策的恰当时机。 速度快并不等同于高质量，高效产出也不等于具备深层价值。Casalegno 提出了团队的座右铭——"没有人类内核的技术，只是没有意义的完美（Technology without humanity is perfection without purpose）"，强调技术工具必须在以人为本的心智框架下运转。 3. **从小步快跑演进到多任务协同**：Jeff Erno 建议采取“由小做大、逐步演进”的路线： - *第一阶段（替代搜索）*：将大语言模型作为更智能的搜索引擎，跳过繁琐的逐条网页点击，即时获取结构化答案。 - *第二阶段（项目级归纳）*：按项目对相似工作进行归类并沉淀可复用资产。 - *第三阶段（本地协同与多文件生成）*：借助 Copilot/Co-work 等高级智能体，在本地环境协同管理并一次性生成复杂的代码库或知识工作流文件。 4. **坚守以人为本的设计底线**：Cooke 补充提醒，切勿因为外界的狂热炒作（Fervor）而随波逐流降低标准。无论引入何种先进工具，团队长久建立的道德准则、以人为本的设计原则（Human-centered design）与设计思维方法论必须始终作为不可动摇的核心支柱。 ### [36:10](https://www.youtube.com/watch?v=OyCaNu3I034&t=2170s) 专业垂直领域的虚拟 AI 伴侣：Leo、Marie 与 Aura 在现有设计和工程软件的基础上，达索系统（Dassault Systèmes）正在探索更契合专业工作流的交互形态，即在软件内集成具备专属语境的虚拟 AI 伴侣（Virtual Companions）。Jeff Erno 指出，当专业人员与 AI 交互时，AI 必须使用该领域的专业语言与术语：科学家需要聚焦于科学严谨性的对话，工程师和设计师需要围绕工业设计与工程规格展开讨论，而涉及业务和项目治理时则需要偏向商业决策的语言体系。 为避免打破既有工作流并降低交互门槛，团队针对不同角色设计并命名了专门的虚拟助手。例如，负责工程技术与设计的虚拟伴侣名为 Leo，专门服务科学研发的伴侣名为 Marie，而处理商业与管理治理事务的伴侣则命名为 Aura。通过为不同岗位提供高度对齐专业语境的虚拟伴侣，工具能够更加无缝、低干扰地融入专家的日常工作中。 ```mermaid flowchart TD accTitle: 专业领域虚拟 AI 伴侣分工架构 subgraph Users ["专业用户角色"] ENG["工程师与工业设计师"] SCI["研发科学家"] BIZ["商业与项目管理者"] end subgraph Companions ["垂直领域虚拟 AI 伴侣"] LEO["Leo: 工程与 3D 设计专属语境"] MARIE["Marie: 科学实验与研发对话"] AURA["Aura: 商业决策与项目治理"] end subgraph CoreSoftware ["核心工业软件与数据环境"] ENV["一体化设计协同平台"] end ENG --> LEO SCI --> MARIE BIZ --> AURA LEO --> ENV MARIE --> ENV AURA --> ENV ``` --- ### [37:08](https://www.youtube.com/watch?v=OyCaNu3I034&t=2228s) 魔法般的创新时刻：从服装 3D 渲染到汽车制造防错 主持人向嘉宾提出核心问题：在 AI 加速创新的周期中，团队做出的哪件事称得上是“如同魔法一般”的突破？主持人首先分享了服装品牌 Marmot 的案例：传统制作一件夹克需要从 Adobe Illustrator 中绘制的技术制版图（Tech Pack）开始，经历跨国寄送缝制实物样品、真人试穿、反复修改的漫长过程，耗时往往长达 6 到 12 个月；如今借助 AI 与 3D 数字化工作流，设计师只需将二维线框图直接导入，即可一键生成高精度的 3D CAD 渲染模型，并置入虚拟摄影棚软件中套用在虚拟模特身上，不到 1 小时便可实时尝试 20 种不同的配色方案并让模特动态走秀。这彻底打破了繁琐的物理工序束缚，让设计师能够全速释放创造力。 Jeff Erno 随后给出了汽车白车身（Body in White）制造领域的震撼案例。在传统工序中，工程师在两个部件拼接处设置大量紧固件（Fasteners），往往只能先凭经验估算，直到下游开模投产才发现现场制造工装设备根本无法伸入特定位置进行紧固作业，从而导致极为痛苦且成本高昂的返工循环（Rework Loop）。而最新的 AI 算法不仅能自动生成并排列所有紧固件，还能同步推演制造工艺与现场工装的可行性，一旦识别出装配冲突，便会自动重算并调整紧固件位置，实现“首次即正确”（Right the first time）。这种直接绕过下游致命返工的设计智能，让工程团队感到不可思议。 - **Marmot 服装打样周期缩短**：从传统的二维 Tech Pack 到多配色、动态试穿的 3D 虚拟摄影棚，耗时从 6 至 12 个月剧降至 1 小时左右。 - **白车身制造自动校验**：AI 在设计生成紧固件的同时检测装配工具干涉，消除工程设计与制造车间之间的脱节与返工。 --- ### [39:38](https://www.youtube.com/watch?v=OyCaNu3I034&t=2378s) 个体赋能、海量数据驱动与叙事表达 在消费品与硬件电子领域，AI 同样重构了设计效率与表达形式。百事公司（PepsiCo）的 Jordan Cooke 虽因保密原因未能透露正在研发的具体项目，但他强调 AI 最神奇之处在于“对个体的超级赋能”（Individual Empowerment）。过去一个设计师要将脑海中的复杂构想落地成原型，需要跨部门协同与漫长排期；而现在，单一设计师凭借 AI 工具在两三天内就能独立完成概念构思并构建出可交互的原型，极大地降低了创意表达和产出设计实物（Artifact）的门槛。 三星电子（Samsung Electronics）的 Federico Casalegno 则从“海量数据处理”与“设计叙事”两个维度阐述了设计范式的演进： 1. **瞬时吞吐并操纵数百万数据点**：在设计诸如 Galaxy Buds 等无线耳机产品时，为了追求极致的人体工学贴合度与佩戴舒适度，设计团队需要处理数以百万计的耳道形态数据点。以往设计师不可能在短时间内快速调动这些海量数据，而如今 AI 能够实时辅助形态迭代，直接在海量数据指引下优化产品的几何曲面。 2. **跨组织叙事与创意沟通（Storytelling）**：设计不仅是制造实体或 3D 打印模型，更在于如何向管理层、跨部门团队和合作伙伴生动地传达设计理念。AI 极大地加速了数字化故事板与动态渲染的生成，让好点子不再因为缺乏展现手段而被埋没在组织沟通中。 --- ### [42:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=2574s) 现场问答：创造力是否成为大型企业的新护城河？ 在观众提问环节，一位现场观众针对 AI 自动化提出核心探讨：研究数据显示，AI 带来的积极岗位收益与消极收益比例约为 4 比 1；随着繁琐重复的低级步骤被工具消除，创造力（Creativity）是否正在成为企业新的“护城河”？设计团队究竟是将节省的时间用于缩短周期，还是转向了更大量的创意发散与方案探索？ Jeff Erno 对此给出了极具行业洞察的解答：企业过去常常误以为自动化技术的核心价值在于“缩短设计交付周期”，但实际情况是，顶尖的设计师和工程师“永远不会主动放下手中的铅笔”（They're not going to put the pencil down）。如果工序提速了，设计团队绝不会提前收工，而是会利用多出来的时间不断打磨、推翻重来、尝试更多激进方案，直到项目管理者下达硬性截稿指令（Finish Line）为止。因此，AI 的真正价值从来不是压缩工期，而是解放设计师的探索边界，让他们在给定时间内探索更多潜在解法并验证其可行性。 --- ### [45:10](https://www.youtube.com/watch?v=OyCaNu3I034&t=2710s) 衡量 AI 投入的 ROI：“学习成本”与未来算力商业模式 针对如何衡量 AI 在设计流程中的成功与投资回报率（ROI），Jordan Cooke 提出了“弯曲效率曲线”（Bending the Efficiency Curve）的双向评估模型： | 评估维度 | 传统度量重点 | AI 带来的深层价值转变 | |:--- |:--- |:--- | | **效率与时间** | 单纯压缩从概念到原型或上市的日历天数 | 避免过度压缩周期，转而在单位时间内增加迭代密度以提升产出质量 | | **设计视野 (Aperture)** | 受限于人力，仅能评估少数精选方案 | 极大地拓宽探索视野，并行探索与概念化海量前沿创意 | | **学习成本 (Cost to Learn)** | 实物打样与线下用户调研成本高昂、试错代价极大 | 大幅降低验证“哪个点子最具市场影响力”的试错成本 | Jordan Cooke 特别提醒道，当前衡量 AI 工具回报率时必须保持清醒认知。现阶段各大 AI 平台的算力信用额度（Credits）和接口费用处于巨额补贴期，但这种低成本红利在未来不可能永远持续。随着补贴退坡与真实使用成本上升，企业必须建立更加理性和审慎的评估机制，精准权衡“工具采购与算力成本”与“实际业务收益”之间的投入产出比，确保在设计流程中有针对性、有克制地部署 AI 技术。 ## Notable quotes > [01:50](https://www.youtube.com/watch?v=OyCaNu3I034&t=110s) 将人工智能直接内嵌到用户日常使用的软件中 — Jeff Erno > [03:21](https://www.youtube.com/watch?v=OyCaNu3I034&t=201s) 设计就是看我们能在有限的时间内将一个方案演进多少轮，从而向世界呈现出真正美妙的产品 — Jordan Cooke > [16:41](https://www.youtube.com/watch?v=OyCaNu3I034&t=1001s) 我们是围绕人类来进行设计，而不是让人类去适应我们的设计 — Federico Casalegno > [16:41](https://www.youtube.com/watch?v=OyCaNu3I034&t=1001s) AI = EI (Emotional Intelligence) + HI (Human Imagination) — Federico Casalegno > [24:21](https://www.youtube.com/watch?v=OyCaNu3I034&t=1461s) 人加上工具，其价值必定大于各部分的简单相加（The sum is greater than the parts） — Jordan Cooke > [25:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=1554s) 招募极其优秀的人才，随后全力保护并赋能他们 — Federico Casalegno > [31:15](https://www.youtube.com/watch?v=OyCaNu3I034&t=1875s) 没有人类内核的技术，只是没有意义的完美（Technology without humanity is perfection without purpose） — Federico Casalegno > [42:54](https://www.youtube.com/watch?v=OyCaNu3I034&t=2574s) 顶尖的设计师和工程师永远不会主动放下手中的铅笔（They're not going to put the pencil down） — Jeff Erno ## Mentioned Jordan Cooke, Federico Casalegno, Jeff Erno, Grace Engels, PepsiCo, Samsung Electronics, Samsung Design Innovation Center, Dassault Systèmes, Ameriprise Financial, Newell Brands, Marmot, MIT Media Lab, Claude, GitHub Copilot, Figma, Figma Make, Galaxy Buds, Galaxy Watch, Galaxy Ring, Leo, Marie, Aura, Adobe Illustrator ## Video description Jordan Cooke - Global Head of UX, Service Design & Research, PepsiCo Federico Casalegno - Executive Vice President of Design, Samsung Electronics Jeff Erno - Technical Fellow, Dassault Systèmes Grace Engels - Chief Strategy Officer, PWA Practice, Ameriprise Financial AI is no longer just a tool for efficiency—it’s reshaping the creative process itself. This panel brings together design leaders who are using AI to generate concepts, accelerate prototyping, and deliver highly personalized user experiences. Panelists will share how they balance human creativity with machine intelligence, what tools are driving the biggest shifts, and how AI is changing collaboration between designers, engineers, and product managers. Walk away with a practical understanding of how AI is enhancing product design today—and where it’s headed next. [Source](https://www.youtube.com/watch?v=OyCaNu3I034)

[1:50] Jeff Erno: 将人工智能直接内嵌到用户日常使用的软件中 [3:21] Jordan Cooke: 设计就是看我们能在有限的时间内将一个方案演进多少轮，从而向世界呈现出真正美妙的产品 [16:41] Federico Casalegno: 我们是围绕人类来进行设计，而不是让人类去适应我们的设计 [16:41] Federico Casalegno: AI = EI (Emotional Intelligence) + HI (Human Imagination) [24:21] Jordan Cooke: 人加上工具，其价值必定大于各部分的简单相加（The sum is greater than the parts） [25:54] Federico Casalegno: 招募极其优秀的人才，随后全力保护并赋能他们 [31:15] Federico Casalegno: 没有人类内核的技术，只是没有意义的完美（Technology without humanity is perfection without purpose） [42:54] Jeff Erno: 顶尖的设计师和工程师永远不会主动放下手中的铅笔（They're not going to put the pencil down）

[14:17] Jordan Cooke: The designer is probably going to spend a lot less time focused on the pixels and a lot more time focused on the value and strategy and storytelling around those pixels. Because anything, pixels, code, documentation, it can all be generated within seconds by any of these tools. [16:41] Federico Casalegno: We design around humans, not humans around our designs. [22:39] Jordan Cooke: It's possible now for anybody to put something out into the world... to go on a tool right now and make a song or make an app or create an image or a piece of artwork. But that doesn't mean that it's going to be good. And it doesn't mean that it's going to come from an authentic place. And I think people as human beings value authenticity. They value craftsmanship. [31:15] Federico Casalegno: Technology without humanity is perfection without purpose. [3:20] Jordan Cooke: design is all about iteration [0:04] Jordan Cooke: bending the efficiency curve [1:50] Jeff Erno: claw that back [36:10] Jeff Erno: don't throw the baby out with the bath water

[0:04] Good morning. Thank you all for joining us. So,  we're excited to talk to you all about the ways [0:09] we've been using AI uh in large enterprises for  product design. So, we'll get started. Uh first, [0:15] I'm going to have each of you guys introduce  yourselves and tell us uh one thing that has [0:20] been your priority in the last couple weeks with  AI. Yeah, I I guess I'll start. I'm Jordan Cook. [0:26] I'm global head of UX service design and  research at PepsiCo. Uh, and what has been, [0:32] I guess, top of mind for me really this whole year  is how to integrate all of these new and emerging [0:38] AI tools into our workflows. Um, and uh do what my  my leadership calls bending the efficiency curve. [0:49] Good morning everyone. My name is Fedrico  Kazaleno. I'm uh the executive vice president [0:55] at Samsung Design. I'm heading the Samsung  Design Innovation Center in San Francisco in [1:00] the global design studios. And something that  um I'm thinking uh those days is actually how [1:07] we can pursue the quest of human centered design  with the rising of AI tools and how we can push [1:15] forward the design anchoring uh the human aspect  of of design. Hi, I'm Jeff Erno. I'm a technical [1:24] fellow from Deso System. Good to be on stage  here with companies Samsung and Pepsi who make [1:30] great products and do great design. Uh the thing  that we're mostly focused on is bringing virtual [1:38] companions to the various tools that we use  that are used across the enterprise from upfront [1:44] design all the way through the manufacturing  engineering of products. Awesome. And Jeeoff, [1:50] I'm going to ask you this first question about  practical application. Tell us about a specific [1:56] AI initiative your team has been implementing uh  and how that is affecting your pipeline. What what [2:03] problem is it solving? Yeah, the biggest thing  is trying to get artificial intelligence being [2:09] used within the tools that our users are using  every day. So rather than having a separate [2:14] tool that you need to bring in to bear, you can  have it embedded inside of the tool you've got. [2:19] So it's more less disruptive for the stuff that  you're doing. A lot of what I saw in my previous [2:25] roles I I spent time both as an engineer as and  a designer also in selling software and creating [2:33] software solutions for engineers and designers is  that they spend a lot of time on nonvalue added [2:39] work and I think that some of the research out  there is suggesting that up to a day or a day [2:46] and a half a week is nonvalue added and I think  that also underestimates the type of work that [2:52] people think is value added that theoretically go  could go away. So the the the time the the time [2:59] that it takes in order to do things that you do  manually could make it go away and we're trying to [3:04] claw that back. So that a lot of the practical  applications are really trying to make these [3:09] solutions the things that I wish I had when I  was a designer and when I was creating solutions, [3:15] writing software, I wish I had them. These are now  in our hands now and we're making them available. [3:20] That's fantastic. Um, and tell us about I'm  going to go down and have each of you guys [3:25] tell me a little bit about how how many designers  do we have in the room here. Do we have a lot of [3:30] designers? Oh, we got a few. What about engineers?  Do we have have a good amount of engineers? Okay. [3:37] So I want to hear from each one of you guys  how you are using AI tools to to speed up the [3:44] iteration process in the beginning of your product  innovation cycle and tell us about a project maybe [3:50] where you've done that and how much efficiency  or what what advantages did it give you maybe [3:55] it isn't adding speed you want me to yeah go ahead  all right yeah I can talk a little bit about this [4:01] um we just uh undertook a a proof of concept  where um we built what we call a tiger team, [4:09] which is just a small, very purpose-built team  of makers um with less of a focus on specific [4:18] uh backgrounds, specific roles. Um more just  a a small team uh of a creative technologist, [4:28] designer, design engineers, a researcher. Um, and  in a span of of three weeks, we're able to put [4:37] forth a series of of prototypes and ultimately  a final working prototype uh for a an internal [4:44] negotiation tool that PepsiCo can can use uh  to make um more intelligent negotiations for [4:52] um different materials that it purchases on  a a regular basis. Um, and we were able to [4:58] produce um more in that 3-week period uh than  an an outsourced team was able to produce in a [5:07] 12-week period uh for a a much lower cost. So,  we're finding ways um to utilize technologies, [5:15] generative AI technologies, um to iterate more  and from a design perspective, design is all [5:23] about iteration, right? and how how much we can  evolve a design in a finite period of time that [5:29] we have to put something beautiful into the world.  Um, and this allows us to accelerate that greatly. [5:37] And that's really an area where we've leaned  heavily in is is in the iteration process because [5:42] we can on a Monday start iterating that day um  and have a proof of concept that is sharable, [5:50] a prototype that's sharable either that day or the  following day to get feedback on and then iterate [5:56] again. So it is rapidly condensed the iteration  timelines that our team is working with. And what [6:02] in your when you've got your tiger team of experts  put together for a sprint like this, what what are [6:07] your go-to top one or two generative AI tools  that you guys are are leaning on heaviest? Um [6:14] so so for us uh it's uh it's Claude. Claude is is  is a tool that we use. Um we use GitHub C-Pilot. [6:24] Um we're doing stuff with Figma make as well. Um,  and I think part of what I've really encouraged [6:30] my team to do is like we're consistently  evaluating new tools, new technologies, um, [6:35] especially on the research side. Um, looking at  synthetic personas and and how we can accelerate [6:42] our outcomes on the research side by um, tapping  into different tools as as well. And Feder Rico, [6:48] when you guys at Samsung are focused not so  much on the CX but more in physical product, [6:53] how does that what does that look like for you  when in the innovation cycle? Well, uh actually [6:59] we um a few years ago we started the new [sighs  and gasps] let's say discipline inside the company [7:04] which is computational design which is basically  how we apply AI, machine learning and advanced [7:11] computing for product development. So basically  what we do is we um uh we capture data 3D or 4D [7:20] data of humans and then we establish KPIs of what  we need to learn or understand. Um then we um [7:31] train our models and finally we start to simulate  and iterate intensively. So basically if you're [7:38] designing the next buds, if you're designing  the next watch and we do it most likely for [7:43] um wearable, if we need to improve how a buds fit  into your ears or a ring into your hands, what we [7:50] do is really we we duplicate um digitally uh and  synthetically hands of ears and then we simulate [7:58] variation of the same products. So we do digitally  we also of course uh capture the movement of of [8:05] humans using the products and we transfer those  movement into robots and we simulate robotically [8:12] and finally we do u user test with real people.  So digital environment, robotic simulation and [8:21] physical or human uh iteration of products all  these participate in how we can quickly apply [8:29] AI in creating new iteration of the products  and uh exploring new ideas, new form factors, [8:37] new variations and quickly come up with solutions.  So that's mainly what we've been doing very [8:42] successfully in the past few years and definitely  we see how our products in particular buds or [8:48] rings they're more comfortable they fit better or  how we can improve the sensors um of our product. [8:55] So that's a an immediate translation from the high  level in a way desire of how we can design better [9:02] product for our consumer into how we apply AI and  diverse tools into experimenting and and advancing [9:10] the uh the product the production capabilities  that we have. So the the outcome is is greatly [9:17] improved having these tools available to you. What  is the if you compare your innovation cycle now [9:22] to what you were doing maybe 5 years ago before  you had these tools available to you? What's the [9:27] what's the biggest difference or how has that  changed? Um it changed uh at you know variety [9:35] of levels. First of all now we we really do  sciencedriven innovation. So we uh uh the way we [9:41] curate our data is unique in that sense that these  are uh data that we capture from uh real humans [9:49] highly curated. Um so we are you know we we trust  our data on one ends and applying AI is a unique [9:57] way to compute an an amazing numbers of of data  and data points uh in a very short period of time. [10:05] So this was not possible before and uh you know  for certain task we went from six weeks to uh um [10:16] couple of days and we are reducing now to minutes.  So you go from weeks to minutes in operating [10:23] task simulations um and iteration of the products  that were uh was unconceivable before. So time is [10:31] definitely one but two there is also this notion  of um putting the teams in a condition of having [10:38] informed discussions and quick iteration around  product advancement and uh development. So that's [10:44] was also another very important uh element of what  what we do. Amazing. One of the things I've seen [10:52] in the industrial design group at when I was at  Newell Brands just before I came to Americative [11:02] CAD tools, right? So, our industrial design  team was able to take sketches that they would [11:06] have had to be very selective on from a time  management standpoint on which ones they could [11:11] actually take to CAD and and take to 3D printed,  maybe prototypes to put in front of users. Um, [11:17] and now it really really has opened up that  possibility. you can try far more things in a [11:22] short amount of time uh to to explore and maybe  get to better solutions you would have had to [11:27] eliminate because they were riskier. What are you  guys doing at Dissult? Yeah, I'll answer from a [11:34] software provider perspective and I'll try to tie  in a couple of things that were mentioned here. [11:38] As Jordan said, it's all about iterations. The  more iterations you can do and the more different [11:43] varieties of things you can do, the better your  products are going to be and the faster you're [11:47] going to get it done. You you go back in time  before even my time, there was no computing [11:52] methods in order to do things. It was slide  rules and it was building things with atoms in [11:56] the universe and testing to see if it works. Very  few number of iterations in order to get a product [12:02] out there. You fast forward in time, punch cards,  that's before my time when people had to do that [12:07] stuff by hand. When I got involved was the dawn of  commercial off-the-shelf software for computerated [12:13] design and finite element analysis. And with that  it was awesome because you were able to do virtual [12:20] testing and you were able to go from maybe one  or two two iterations to maybe five or six and [12:25] you could do it without atoms. You could just  move ones and zeros in a computer system. It was [12:29] absolutely awesome. And on top of that we started  to expand further to be able to build optimization [12:36] routines that go around them. But what comes as  a part of that is there's a lot of skill that's [12:41] involved. You need to understand computerated  design, how to make something parametric. And [12:46] one thing that if anybody's done this for a  living, making something that's parametric [12:50] that survives when you make changes is not easy.  Very, very difficult to do. And that's a topic [12:55] if anybody's interested, I can talk more offline  about. But having that all work updates correctly, [13:02] gives you the result, and can iterate. You always  were wondering, did you really get the best [13:07] design? Because it took, even with that, you got  into the dozens of iterations and got something [13:12] better. But was there something even more? Well,  now with some of the tools we have for generative [13:19] capabilities as well as machine learning on  getting the performance results, rather than [13:24] doing a detailed finite element analysis that took  maybe hours in order to run back in the day, now [13:30] we could use machine learning models in order to  get answers much much quicker. So Grace, you were [13:35] talking about the how does this affect the early  design? That's where the the biggest impact on [13:42] your overall product is going to be. That's where  you're going to make the biggest dent on your [13:45] performance and cost is early on before you start  to unleash the rest of the organization doing the [13:52] detailed work. So, it's something that the more  that you can do that type of stuff up front, the [13:57] better you're going to be able to get the result.  You'll be able to find something that's more of an [14:01] overall better result rather than one that it's  it walked and found something close by. These [14:07] are all very important things and the artificial  intelligence technology will help us to generate [14:11] new ideas that are evaluating and then evaluate  them so that you can get better answers sooner and [14:17] make better products. Yeah, let's talk a little  bit more about that process and how that is going [14:24] to look in the future. So if you have friends  maybe in your marketing teams or your executive [14:29] teams that suddenly can now generate uh a product  idea image that looks fairly uh fairly realistic [14:38] um but that they really don't understand how that  turns into the manufacturing process right or even [14:44] um the the CX process. what what do you see the  role of the designer uh in the future and how [14:52] does the product designer's role either in the in  the CX space or in the physical product space look [14:57] in five or 10 years? So, so I can weigh in here a  little bit. Um, my point of view on this is that [15:07] the designer is probably going to spend a lot less  time focused on the pixels and a lot more time [15:14] focused on the value and strategy and storytelling  around those pixels. Because anything, pixels, [15:22] code, documentation, it can all be generated  within seconds by any of these tools. These [15:29] tools are only going to get better, but they're  tools. Um, and I feel very strongly that when [15:34] you take a talented, capable individual and you  pair them with the tools, then it's basically [15:39] like rocket fuel. It rapidly expands what they are  able to do singularly as as as a as a contributor, [15:46] as a a leader. Um, and so my feeling is that  these tools are are going to help talented [15:58] people accelerate and expand the aperture of what  they are able to do and that their role is going [16:05] to be much more as a facilitator, as a connector,  as a strategist. Um, and I think that design plays [16:12] a really unique role. Um because as a lot of this  stuff is democratized and I think we need to feel [16:17] comfortable with saying we're not going to in  all cases own the pixels um we need to be the [16:24] uh the people who are advocating for the quality  and making sure that we're setting a high bar on [16:28] craft and storytelling um and and that anything  that's put out is is matching that quality bar [16:35] that we're establishing and and that we are sort  of the the purveyors uh and the gatekeepers of [16:41] that. Amazing. Fed Rico, you at Samsung have have  moved into almost another genre of design and and [16:50] kind of relabeled it in your organization.  What how do you see the the product designer [16:55] uh role evolving from that space or are you guys  already there? Well um for me the let's say the [17:03] our mission hasn't changed and the mission is how  we can understand people and culture how we can [17:09] understand lifestyle and then how we can uh once  that you understand that how we can imagine how [17:16] new products and experiences can leverage on  emerging technologies to uh provide end user [17:23] amazing products or experiences. So this notion of  human- centered design still remain. So we design [17:30] around humans not humans around our designs. And  given that the role of designer somehow is still [17:38] in a way um how you connect the dots between human  needs and the products that we bring to market or [17:47] the experiences that we bring to market. So um  there is a tremendous push on our organization [17:53] in using at our best uh the latest technologies um  from AI to machine learning to any type of tool. [18:01] So definitely we are advancing the use of how we  integrate those tools in our processes but I think [18:08] that the role of designers is really how do you  really understand and empathize with the people [18:12] you're designing for and this is something that  AI cannot simply tell you. So actually we have a [18:18] formula which is AI equal EI plus HI which  basically means artificial intelligence is [18:27] emotional intelligence plus human imagination. So  somehow this formally encompassed this uh desire [18:34] to um bring together advanced tools with strong  empathetic and imaginative capabilities from our [18:44] teams that need to bring those capabilities  into everything that we we design. So the [18:50] role of designer is even more central because  to some extent they're the one who understand [18:57] um not only societies at the social level um even  psychological level but trying to also bridge [19:05] uh economy society engineering and embody  everything into an artifact a product and [19:12] experience a design UX and so on. So role  of designers nowadays for me is particularly [19:17] interesting because exactly can help to to  connect the dots of a fastmoving u economy, [19:24] fastmoving uh innovation landscape. Yeah, I think  empathy and and sort of consumer understanding is [19:32] one of those things that's going to be harder for  our synthetic intelligence to bridge, right? And [19:37] it is a lot of what designers do intuitively. We  don't even even talk about that. That's what we're [19:42] doing when we're taking [clears throat] uh ideas  and translating them for execution. Uh so there's [19:48] this statistic and I'm going to butcher it because  I didn't look it up before this. I should have [19:53] uh that from the advent of photography, you know,  when there were tin types in the 1800s to I think [19:59] it's around 2015 um when digital photography  entered kind of the smartphone ecosystem. [20:05] that total number of photographs that were  captured is now approximately equivalent to the [20:11] amount that's captured every minute of every day  today. Right? So if you think about 150 years of [20:16] photography is now the amount that we're getting  every minute with the number of people who are [20:22] are capturing photos everywhere. It's kind  of equivalent to the scale of design that we [20:28] might see with all of these AI native tools being  able to create all these things designers uh have [20:34] made forever. But if you think about that and the  way sort of experts in photo and video are still [20:40] needed every day and their expertise is incredibly  valuable and very useful. I know we have someone [20:46] actually videotaping us right now. Um how do you  think uh that that is kind of related to the way [20:53] we'll see the product designer uh continue? And  I'm going to go to you at the end Jeff. Okay. Oh, [21:00] I think there's going to be a limit on that. the  throttle will be reality. That images are easy [21:07] because it's just ones and zeros. It's pretty  straightforward and you can consume them. But [21:12] if you're making a product, you have to bring  it into existence and that requires lots of [21:16] different things. It requires the upfront product  manager to be able to understand what's important [21:22] and what will sell in the marketplace. Whenever  you have a product, it needs to be produced. [21:27] Sometimes you can 3D print it, Grace, like in  the example you had, but a lot of times that's [21:32] not going to be economical. You're going to need  to do something with more subtractive methods, [21:36] more traditional type of methods. When you've got  that, you need to be able to bump those designs up [21:41] against reality. And that's where the more that  you can have your data connected in your system [21:47] in order to connect. what are the requirements  and did you meet those requirements when you did [21:52] that virtual testing and then physical testing  and record that when you do that design is that [21:58] something that's able to be produced or is it  something that needs to be adjusted in order [22:01] to do that so I think that you're going to get  a lot of help from the artificial intelligence [22:05] tools if it can traverse across the sort of  knowledge graph of your enterprise but it's [22:11] going to be grounded in reality so your your  designs will be grounded in the digital data [22:15] that you've got and then the actual thing you  produce will be grounded in the reality of what [22:20] actually sells and what you can produce. Yeah,  there's definitely a limit in uh in time and [22:26] how much it takes to make a manufacturing tool and  to set up uh the actual manufacturing process for [22:33] uh not only a time limit but an expense limit and  absolutely that's true. What do you guys as design [22:39] leaders uh how are you handling this transition  with your teams? How are you helping your team [22:44] members either scale up or be comfortable with  this amount of transformation that we're kind [22:50] of living through day-to-day and also seeing uh  coming into the future. Yeah. Um so we are very [22:59] much trying to in embrace uh a change in in  philosophy uh a change in our ways of working [23:08] uh a value consistently evaluating and bringing  on new tools into our arsenal. Well, I mean, [23:12] that's what I love about design is it is  consistently evolving and so you you can't [23:17] sort of rest on your laurels as a designer.  So, this is just sort of to me the next wave [23:23] uh of of technology and I think you know for the  right individuals that is a very empowering thing [23:32] um because we're I think establishing a lot of new  paradigms. Going back to your your last question, [23:37] like, yeah, it's possible now for anybody to put  something out into the world. It's it's possible [23:42] for somebody to go on a tool right now and make  a song or make an app or or or create an image [23:49] or a piece of artwork. Um, but that doesn't mean  that it's going to be good. Um, and and it doesn't [23:57] mean that it's going to come from an authentic  place. And I think people as human beings value [24:02] authenticity. They value craftsmanship. they they  value a really fantastic story. And that's why [24:09] even though we all have cameras in our pockets and  we can all be making movies and taking pictures, [24:15] uh, photographers still exist and my wife pays  a lot of money for one to take our family photo [24:21] every year and people still create movies and  art and music. Um, and all of that is going to [24:30] continue to exist. It's just how can we utilize  these tools and give them to the right people so [24:35] that you know that kid in a basement uh who is  really you know I I go back to to music right [24:44] somebody who wants to create music but doesn't  have the the budget to enlist a full orchestra [24:50] but can they use these tools to bring that  musicality into what they're creating and enhance [24:55] what they're able to do with with with what they  have and I think that's really the opportunity [25:00] that we have and that's what I'm trying to get my  team to lean into is like if you have these tools [25:05] with what you have in in your arsenal in your  background how can you enhance what you're able [25:10] to produce um and that's that's been my challenge  to them and I think something that they've very [25:15] much embraced because I feel very strongly  that again the people plus the tools the sum [25:22] is greater than the parts so the person and the  tool is going to be greater than either of those [25:28] things individually and that's what I want to be  able to facilitate with my team is connecting them [25:32] with the right tools to be able to have better  outcomes. Absolutely. I mean, I do feel old when [25:38] I have to tell team members, you know, back  in my day, if I wanted to clip out a person, [25:42] I'd had to hold down the option key and go around  with the lasso for 25 minutes, right? But now you [25:47] guys can click a button. What do you think? How  do we manage the the transition? I mean, for us, [25:54] it's very easy. Um, I hire very good people.  I hire very is a great start. So it's very [26:01] easy for me. Uh so my role is how to protect  them and enable them. I I have an incredible [26:08] uh small efficient team of very smart people. So  for example in computational design we said before [26:15] was discussing before um our teams they have they  have been written uh in the past couple of year [26:23] between 15 20 AI specific programs that they need  to develop their own um practice. So on one end [26:34] what they do is really they trying to bend techn  technological development for for what they need [26:40] and um and that's one strong point and I truly  believe in that. Uh the second one actually what [26:46] you were saying before about the authenticity.  This is something that we discussed tremendously [26:50] because in a in a moment where uh product the  production the output is no longer a a matter of [27:00] of time or quality because you can really produce  a video in a in a slap of a finger nowadays. Um [27:07] what mattered the most is the meaning of it. So  while producing and creating become almost instant [27:17] uh understanding the meaning is important. So what  the team needs is really time to to uh process to [27:24] sediment to discuss um go through uh argument. So  having a in a sense a safe place to enable those [27:32] type of engagement where designer can really go  through uh um long discussion and and and thinking [27:40] because at the end of the day we try to um to  solve problem and we don't have an answer yet. [27:45] I mean for all these new products, new experiences  that we are seeing developing under our eyes. Um [27:52] it is an open domain and there are a lot of um  possible path that we need to take. So uh having [27:59] strong teams that uh work and discuss together  uh we hire team from a variety of disciplines. [28:06] uh this is something I've been doing since um  I've been teaching at MIT for about 20 years [28:10] and and the first things we used to do is to teach  classes across disciplines and at one point at the [28:16] media lab we have this mot to uh basically destroy  um uh go against disciplines to have only hybrid [28:24] disciplines so I'm not saying that we need to uh  go this path but bringing individuals and talents [28:31] that have very strong disciplines working together  is definitely one area an area that is extremely [28:38] extremely uh interesting for us. What are you  guys seeing uh from a design leader standpoint as [28:47] uh kind of the biggest challenge that your teams  are facing? Is it getting the the tools that they [28:53] need? Is it getting the skills that they need?  Or is it is there are you feeling reluctance? [28:59] I have not seen in the design group as much sort  of um reluctance and adoption as in some of the [29:06] other parts of the company, right? It's it's been  pretty quick for designers to see when something's [29:13] valuable and oh, I want that. Oh, I want to use  that. Oh, I tried that. Um cuz it's kind of a a [29:18] native uh trait of designers. Um but what what are  you guys seeing? What's the the biggest challenge [29:24] that your teams are bringing to you kind of at  the on the ground level? Well, I can start with [29:29] one thing. I guess that relates to the previous  question as well with where's the resistance, [29:33] where it's coming from. And I think we heard that  on stage for those who were in the plenary session [29:38] this morning. A lot of people are afraid of their  job going away. And I think the thing that really [29:45] will resist that or will fix that is a lot of the  stuff that's going to go away is the drudgery. [29:51] And a lot of the time I spent in my career was  noticing people having a difficult time doing [29:57] something that is very repetitive and not really  value added and making it go away. I've yet to [30:02] have one person say, "Geez, thank you for doing in  five minutes." That used to take two days, right? [30:08] So I I think removing the drudgery is one way that  you can get past the any sticking point with some [30:14] of the adoption of that. And I think the other  the other piece of it is when you're the problems [30:20] never stop. There's always going to be things that  you want to solve and there's kind of a hierarchy [30:25] of need related to those. And when you're dealing  with limited time that you can spend in order to [30:32] get to a result, you're going to have really hit  the big hitters first. But once you can automate [30:38] away the lower level pieces and you can start to  focus on that upper level, more like an editor at [30:43] a newspaper rather than the one that's writing  the article, you can then start to go after [30:48] some of those things that weren't the lowhanging  fruit. They were the more challenging problems, [30:52] but the ones that might be very interesting, the  ones that might be better for green technology or [30:57] better for accessibility or various different  things that you would never probably get the [31:03] time to get to work on, but now you have the time  to do it. And so that is the type of thing that's [31:08] exciting for designers and engineers is to be able  to solve new and exciting problems. For anyone in [31:15] this room that's a design leader that maybe  hasn't started on uh a deep AI implementation [31:22] journey in their team, what would be the  first thing you would recommend that they do? [31:29] Uh well, I think it's great that  you're attending this conference. Um, [31:35] so that is like a really great first step. If if  that's all you've done is attend this conference, [31:38] then I think you're getting a lot of really  great perspectives and information there. Um, [31:44] I think evaluation of the get get people using  tools and experimenting and figuring out how this [31:52] can augment what they're really good at and what  they're passionate about. And uh to Jeff's point, [31:58] you know, uh take those things that were just  timeconuming that they aren't passionate about [32:02] and make those something that they can get out  of the way quickly so that they can focus on the [32:06] things that are are most valuable to them and the  and the work that they're they're doing. Um and [32:12] just have an open mind. Um and don't be afraid  of the technology. I think that is something [32:18] that I've seen in my team in in where maybe some  practitioners have struggled is being so afraid [32:28] that the technology is is going to somehow replace  you that you sort of completely put blinders [32:35] on and and just sort of say I don't need that or  that's not helpful to me. Um, absolutely have have [32:42] an open mind and and a willingness to learn and  and I think that will help guide you into finding [32:47] ways to make this work within the structure of  what you're trying to do um and the way in which [32:53] you're trying to lead your team. Well, um I  would say this in uh let's say two ways. On [32:59] one end of course when you're trying to implement  new tools in any larger corporate corporation you [33:04] need to you know make sure that uh security  is addressed uh proper training is in place [33:12] uh cost those tools they are pretty expensive. So  there are you know all the functional things that [33:18] need to be taken care of but most importantly  and for me this is the first things that we pay [33:23] attention uh nowadays those tools they do not  only change the speed of the work that we do is [33:30] a dramatically change in the mindset. So changing  how people thinks the what they do in the way they [33:38] do it is the first and the most important things  to do for me. So it's not only because you can [33:43] u expedite certain tasks which is absolutely  very valuable but how do we think really how [33:50] those tools empower designer to do the job uh in  the best possible way. Actually in uh in Greece [33:57] you have like two ways to uh define time. One  is kronos which is the chronological succession [34:04] of time. So definitely you need to shorten the the  time that you need to do certain task but you also [34:10] have the word chyros which is the right moment to  make decisions. So you know it's it's not speed [34:17] doesn't mean quality output doesn't mean meaning.  So in a way again we uh we use our another motto [34:26] um technology without humanity is perfection  without purpose. So how do we bring this humanity [34:33] into the design process and this require not  only the tools but the proper mindset? Love [34:42] that your question was how do you get if you  were a leader what would you tell them to do [34:48] first? Yeah. To get I get the the way I would look  at it is start small and grow. I I like to think [34:54] of my progression through using some of the large  language models as a part of my job is the first [35:01] and most obvious thing is just use it to replace  Google. When you're doing a search, you can do a [35:06] search and it can find things a lot faster rather  than just bringing up a bunch of things that you [35:11] then have to click through. The progression goes  from there to maybe you can group things that you [35:17] do similarly into a project and then be able to  reuse things from there. And then going further, [35:23] you've got things like co-work or what you but  probably started with code where you've got a [35:29] project on your machine and then it's doing a  bunch of things coordinating and creating a lot [35:34] of different files at once whether it's you're  writing code or you're doing knowledge work. So [35:38] start small and then you can progress upward  as you do more and more of it and you'll see [35:43] that the value will grow as you do that. Yeah. And  just just to add to to what they've they've said, [35:49] it don't don't change your standards. Um don't  feel like you need to implement something because [35:54] of the fervor of what's going on around you.  Continue to maintain your standards, your ethical [35:59] standards, your standards for human- centered  design and design thinking. Um and find ways to [36:05] implement the tools that fit within the frameworks  that you've already spent a lot of time and energy [36:10] and and passion establishing. Um, don't change,  don't throw the baby out with a bath water. One, [36:18] one thing just to quickly add to that one because  it's something that we've done within our software [36:22] is these existing software programs and they're  we're adding virtual companions to them. And when [36:30] you deal with those companions, you want that to  speak the language of what you do. If you're a [36:36] scientist, you want the chat to be around science.  If you're an engineer and a designer, you want it [36:40] to be around that. And if you're dealing with more  governance or business types of things, you've got [36:45] to run that. So we we created virtual companions  and we we gave them names. Leo is the engineer, [36:50] Marie is the scientist, and another one named Aura  is more business related. And so that's one thing [36:55] that we're trying to do is to help that is to make  it so that the conversation is more aligned with [37:01] what you've got and try to make it so it's less  disruptive. with the little bit of time that we [37:08] have left here, I would I want to ask each one  of you in your in your innovation cycle that has [37:14] been greatly accelerated or accelerated. What is  one thing you guys have seen your teams do that is [37:20] just like magic? So, I'll give you an example.  Uh at no brands, one of the brands we have is [37:26] um Marmet, which is outdoor apparel like um  jackets and and things like that. And one of the [37:32] things that the first thing that really astounded  me that we created was you could take the the [37:39] technical drawing for anyone who doesn't know  when you make a jacket there's a tech pack which [37:43] is basically just an outline in Illustrator of the  shapes that make up the jacket and you could take [37:50] that and turn it into a beautiful rendering  and 3D CAD and then put that straight into [37:56] uh some of the the photo studio software and put  it on a model. You can try 20 colorways and you [38:02] can you can have the model even walk around and  all of that you can now do in in about an hour, [38:08] right? And that used to be like a 6 to 12 month  process that involved getting sewn uh samples [38:13] that had to be mailed across the globe um to even  be tried on and then revised and iterated. Uh, [38:20] so it's it's just really something that adds to  the ability of the designers to to really express [38:26] their creativity because they're not bound by  sort of the cumbersome steps that used to be [38:31] entailed to really get from what their idea was to  what was getting produced. Is there any concrete [38:37] examples that you guys have seen that uh that will  help people get excited about this technology? [38:43] I can go with one that the there is an example  that we had where it was a automotive example [38:48] with a body in white and not only was the software  able to generate a bunch of fasteners that were [38:55] going along and fusing two things together but  it also was able to look at the manufacturing [39:01] process and determine the tooling that you're  doing will not be able to make one of those [39:07] connections and automatically adjust the position  of them so that it That's the very kind of thing [39:14] that is tremendously annoying in the real world  because what would happen is that you would put [39:18] your guess and then you would find that out after  you've already released things and downstream in [39:23] the process that oh no you have to now do a  rework loop and being able to get out of that [39:28] rework loop and get the answer right the first  time to me that was just mind-blowing. Yeah, [39:33] it probably made a lot of people happy that  you you didn't do that. [laughter] How about [39:38] at Pepsi? Have you guys seen any magic? So I have  to be a little bit care. There's some stuff that [39:44] we're working on that I think is really exciting  that um I can't really talk about right right now. [39:50] I think the magic for me that I have seen is how  the technology can empower an individual to make [39:58] something. We've had designers who have had an  idea and have been able to bring it to life and [40:04] conceptualize it in a prototype by themselves over  the course of a couple of days. And that is really [40:10] magical because that reduces so many barriers to  entry for them to bring thoughts and ideas to the [40:16] table and to tell a story about them and for us to  be able to share that as an artifact. Um, and that [40:22] I think is the most magical thing about it even  more so than than what we're we're creating. What [40:31] about you? Oh, I have a is an interesting question  actually. the magic into the into the process. One [40:38] thing that immediately come to mind is uh is this  ability to see and in a way manipulate millions [40:48] of data points in a in a very short period of  time. So that's in a way magic. I mean you can [40:55] uh you can really play with all these data and uh  iterate very fast very quickly and uh challenge [41:04] your your design improve your design iterate and  this was literally not possible uh some time ago. [41:12] So definitely this capability of having millions  of data points um that make sense for the type of [41:20] design you're designing at your fingertips and use  it for your design is amazing. Um if I only think [41:27] for example the the data we use to design buds and  and how we can improve the form factor in order to [41:35] provide a better fit or a better comfort. um this  was not possible before and now it's really um [41:42] uh available to to our design team. Storytell is  another thing I mean in in every in every design [41:50] the or every every work actually the importance  not only to create the content but being capable [41:56] to to share the content and engage in discussion  with other uh divisions of your organization with [42:03] other peers etc. This is a very important part of  the story. So we have increased not only the uh [42:10] ability to create a 3D printed or physical object  but also digitally how do we increase uh the way [42:17] we can tell the story about our ideas and and this  enable a quick iteration. So these two things for [42:23] me um having access to millions of data points  during the design process and improving the way we [42:30] tell the stories about the ideas that you want to  bring to market are fundamentally um magic. Yeah, [42:37] that is magic. I I can only imagine how many  great ideas were lost because the person who had [42:43] them didn't have the tools to tell the story about  what they were trying to to say. And now now that [42:48] there's really a lot more opportunity there. Um,  we have time to take maybe one or two questions. [42:54] Is there anyone in the audience that has a  question? Yes, right in front. Yeah, as we as [42:58] Jeff was talking about it, you were talking about  the lassu issue, but really is creativity becoming [43:05] the new mo, right? when you start to really  look at AI as opportunity to unlock creativity, [43:11] you know, and and democratize the opportunity for  people to ideulate and create because, you know, [43:16] it seems like part of what I'm hearing across each  of you is this, at least some of the things we're [43:22] reading, but from the research we're doing is  that four positive job outcomes for every one [43:27] negative. Um but what we're seeing is the increase  not in reduction in time but in the increase in [43:33] output and the increase in in creativity and  creative unlock. So just was curious how you as [43:39] designers really sort of see it taking out those  steps that prevented you from multiple designs [43:46] prevented you from doing more ideation creation  you know is is really the underpinnings of AI [43:51] you know unlocking creativity as a mode for large  enterprises. Uh for anyone who couldn't hear it, [43:57] the question was around is uh kind of the  the enablement of these technology tools, [44:04] removing some of the tedious pieces really an  unlock for creativity for some designers. Anybody [44:09] want to take that one? I I'll just take one shot  cuz it's something that came up in the past. I [44:15] remember we used to talk about oh this automation  is great because you reduce the cycle time. you [44:21] really don't res because [clears throat] the the  engineers and the designers, they're not going to [44:24] put the pencil down. They're going to keep working  until you tell them to stop because they're going [44:28] to constantly want to make it better. So, it's  really more about the creativity, getting those [44:33] ideas out there, evaluating them to see if they  work, and getting better results. The the cycle [44:38] time is really when you put the finish line.  It's it's when you say stop. Anybody else? Yeah. [44:50] success [45:02] it sounds. [45:10] All right. Well, how are we measuring uh  the success of these? and and we will also [45:16] uh all four of us will be out in the  hall after this if anyone wants to [45:19] continue to ask any questions. Anybody  got a a two second answer to that one? [45:26] Uh it's multiaceted. I'm watching the clock tick  down and it's freaking me out. Um it's going up [45:33] now. Uh it is bending the efficiency curve.  So you can do that in two ways. you can get [45:41] something done faster. Uh or you can expand the  aperture of what you're trying to do. Um and in [45:49] either of those cases, you can make an argument  that the investment is is is is worth it. Um the [45:58] time is the one that a lot of people focus on. So  reducing the timeline that it takes to get an idea [46:05] to prototype or get something to market. Um, and  you can push those a little bit, but you're also [46:11] again, you know, to the point of of what's already  been discussed here, the more that you iterate, [46:16] the the better the outcome is going to be. So,  you don't want to compress those too much. The [46:21] other thing is like the cost to learn, which is  really valuable to our company in understanding [46:26] which ideas have the most weight, have the most  impact. Um, and we found where we can get a lot [46:32] of value is conceptualizing and testing a lot of  different ideas and doing that at a much lower [46:40] cost to learn than our traditional uh methods.  And that's I think where a lot of we're deriving a [46:46] lot of value, but we're still very much in like I  guess early stages of defining this. And the cost [46:51] structures are constantly changing too, right?  Because credits now are heavily subsidized, but [46:57] that can't continue into the future. So, I think  like in the future as people walk out the door, [47:02] uh I guess I am boring them to death. Um  that is going to be something that we need [47:07] to evaluate is like the trade-off of the cost  relative to what we're getting from it and [47:12] being really intentional in how we're using  these technologies. Thank you guys so much.