AES Rebuilds Corporate Structure as AI Cuts Coordination Costs

AI降低协同成本，AES重构企业组织架构 充沛的人工智能通过大幅压低协同成本，使传统企业官僚体制在经济上过时，迫使企业重新设计结构性工作流，而非仅仅自动化既有任务。 Erin Boyd立足于罗纳德·科斯（Ronald Coase）关于交易成本的经济学理论指出，企业官僚体系与职能孤岛最初是为了配给稀缺的人类智能而建立的。由于AI从根本上将智能从稀缺瓶颈转变为充沛的要素资源，企业转型不能局限于传统工作流的数字化，而必须重新审视企业本身的结构性目的。Boyd勾勒出了一场运营转变：迈向多智能体编排、保留必要人类审慎判断的有意裁决接口，并将合规职能重构为信任的构建者。最终，管理者应着眼于优化能力密度与专有上下文，而非员工人数，从而将人类人才从日常行政协同中解放出来。 以往的技术革命实现了体力、能源与计算的自动化，而人类判断力始终稀缺；AI则是首个直接挑战智能本身稀缺性的技术革命。 (0:04) 企业在历史上是为了降低公开市场的交易成本并克服个人认知局限而形成的，但充沛的AI智能正在冲击传统企业结构的经济学前提。 (5:06) 企业官僚体系曾是对稀缺智能进行配给的理性历史产物，但AI真正的机遇在于将知识型员工从行政流转和协同事务中解放出来。 (11:12) 不同于旨在优化传统工作流运作方式的数字化转型，AI转型迫使管理者重新思考组织最初为何要设立特定的部门壁垒与交接流程。 (16:28) 长期价值并不来源于构建孤立的单个智能体，而在于编排融合了共享组织记忆与上下文的多智能体架构。 (20:55)

Erin Boyd of The AES Corporation argues that because historical technological revolutions only automated physical labor or computation, modern enterprises were engineered entirely to ration and coordinate scarce human cognition. With artificial intelligence making intelligence abundant, organizations must radically redesign their foundational structures and Coasean transaction costs rather than merely automating legacy bureaucracy. Coasean Transaction Costs and Cognitive Scarcity (5:06) Across past economic revolutions—steam, electricity, computing, and the internet—muscle, energy, computation, and raw information successively ceased to be scarce, yet human judgment consistently remained the operational bottleneck. Drawing on Ronald Coase's 1937 theory of the firm, enterprises formed because open-market coordination carries heavy transaction costs: repeatedly sourcing talent, negotiating contracts, establishing trust, and aligning context. Organizations partitioned complex challenges into specialized functional silos to compensate for the cognitive limitations of individual minds. Because traditional corporations were constructed specifically to overcome the friction of coordinating scarce human intellect, competitive advantage historically stemmed from accumulating tacit, undocumented knowledge inside organizational routines. AI disrupts this foundation by providing specialized domain expertise on demand and retrieving deep context, challenging the economic necessity of traditional corporate structures. Reframing Bureaucracy to Stop Wasting Intelligence (11:12) Corporate bureaucracy—manifested in approval chains, status updates, and management hierarchies—was not an arbitrary creation, but a rational economic response designed to allocate scarce knowledge workers across large organizations. Over time, this architecture created systemic inefficiency by turning skilled engineers, attorneys, and project managers into administrative information routers. Boyd contends that the objective of enterprise AI is not replacing human workers, but eliminating the administrative waste of human cognition. Consequently, leaders must distinguish AI-driven redesign from the past two decades of digital transformation. Digitization converted paper into pixels, connected software stacks, and sped up human visibility, but it left the underlying premise intact: humans still bridged every functional seam. Abundant intelligence allows organizations to question why legacy processes exist at all, moving beyond inserting agents into inherited workflows toward eliminating unnecessary administrative layers. Balancing Autonomous Execution with Deliberate Friction (22:49) As software evolves from passive user interfaces to autonomous agentic systems that reason and execute independently, product design must shift its focus. Instead of reflexively removing user friction, designers must intentionally position friction where human ethics, accountability, and empathy are essential, such as personnel termination or high-stakes contracting. Technologists must also prioritize orchestrating multi-agent architectures with shared memory layers over pursuing isolated, single-agent models, warning that uncoordinated tools will merely replicate traditional bureaucracy on GPUs. Simultaneously, legal and compliance departments must transition from restrictive gatekeepers into architects of trust. Because enterprise acceleration requires transparent decision-making, explainable audit trails, and enforceable guardrails, governance serves as the prerequisite that allows organizations to pursue autonomous execution without descending into operational chaos. Capability Density and Enterprise Transformation Realities (27:40) Because frontier AI models are widely accessible commodities, enterprise differentiation no longer stems from raw processing power, but from proprietary context, institutional memory, customer trust, and authentic human relationships. Strategic leadership must pivot toward optimizing for "capability density"—maximizing the organizational value generated per individual through intelligent systems—rather than expanding headcount. In practical application, AES is testing these structural redesigns across its 10,000-person global workforce, targeting roughly 80% transformation within a three- to five-year horizon. This operational shift requires enterprise IT to move away from centralized roadblocks toward building enablement frameworks and security guardrails that allow non-technical, distributed teams to deploy solutions safely. The primary operational bottleneck in restructuring large enterprises is not technological capability, but the cognitive difficulty of helping employees entrenched in day-to-day operations envision work without bureaucratic scaffolding. In addition, organizations still face the unresolved challenge of developing mature quantitative metrics for judgment, confidence, and trust beyond basic speed and engagement indicators.

AES Corporation 战略专家 Erin Boyd 指出，随着人工智能使历史上最顽固的经济瓶颈——“智能”从稀缺走向充沛，现代企业的核心挑战不再是单纯在旧体制内局部嵌入技术，而是必须从根本上重构基于交易成本建立的科斯式组织形态，将企业重心从管理分配稀缺认知转向释放人类的高阶判断力。 经济稀缺性转移与科斯企业理论重估 (0:04) 两百年来，技术革命相继消除了体力、能源、计算与信息获取的稀缺，但企业形态始终锚定于唯一的顽固瓶颈：稀缺的人类智能与专业判断。罗纳德·科斯（Ronald Coase）在 1937 年提出的企业理论指出，企业之所以存在，是因为在公开市场上搜寻人才、谈判合约及对齐上下文的“交易成本”极其高昂；为了应对个人认知的物理局限，企业建立了职能孤岛，并通过层级审批、进度报告与冗长会议来管控分配稀缺的专业知识。 然而，这种科斯式设计导致高价值专业人才被大量消耗于低效的数据路由与行政协调中。Boyd 强调，人工智能的经济学本质不是淘汰人类劳动力，而是提供按需即得的跨领域智能，从而使繁重的协调性官僚机制失去存在基础。企业领导者不应仅关注在哪里增设智能体，而必须审视各项遗留审批与层级结构最初赖以成立的经济瓶颈是否已经消失，其终极目标是“停止浪费人类的智慧”。 超越数字化：从局部自动化到组织重构 (16:28) 过去二十年的数字化转型将纸质流程转化为数字记录，打通了系统连接并提高了全局能见度，但其底层假设未变——人类智能仍是每个业务关口的唯一协调者。当前的普遍误区在于试图优化继承自二十世纪的既有结构，仅把 AI 视作下游自动化工具。Boyd 明确提出“不要自动化它，而是重新设计它（Don't automate it. Redesign it）”，主张从询问“如何用技术完成任务”转向质疑“该组织结构与交接环节为何存在”。 在技术构建层面，软件开发的重心已从编写代码转变为组织智能的流动架构。鉴于前沿基础模型正迅速商品化，单独研发最聪明的孤立智能体是一场注定落败的竞争，技术主管的核心任务是打造“最佳交响乐团”——构建多智能体协同架构、沉淀组织上下文的共享记忆层并在系统中注入信任。如果不进行这种全局重塑，分散的部门级 AI 项目只会演变成运行在 GPU 上的新型企业官僚主义。 人机协作新范式：判断界面与信任治理 (22:49) 在自主智能体时代，产品设计正经历根本演变：以往的设计目标是通过减少点击和缩短等待来消除一切用户摩擦，而新的“判断界面”则关注人机关系的动态界定，核心问题转变为“人类究竟何时应当介入”。对于订购办公用品或安排日常会议等低风险常规任务，系统应当彻底消除摩擦以实现完全自主执行；但在涉及解雇员工、签署巨额合同或处理高风险客户关系等决策中，摩擦必须被审慎地保留与植入，以承载伦理、同理心、责任与关键判断。 与此协同，法律与合规部门的角色也从以往被动的“阻拦者”转型为“信任架构师”。市场和企业的运作完全依赖于可解释的决策、具有约束力的契约以及明确的追责机制，这些在经济学上属于维持信心的必要成本。治理的目标不是阻碍技术采纳，而是通过透明的决策逻辑、不可篡改的审计追踪和安全护栏来保障自主执行。正如演讲所指出的，缺乏信任的提速会导致混乱，而缺乏速度的信任会导致淘汰，严密的治理正是兼顾两者的基石。 能力密度战略与 AES 公司的转型实践 (27:40) 随着通用模型与基础算力走向平民化，企业的核心竞争优势将彻底转移至模型无法掌握的专属资产：专有业务上下文、组织记忆、客户信任、声誉与企业文化。高管层的战略重心必须从单纯管理员工规模，转向追求“能力密度”——即通过智能系统杠杆最大化单个人才所产生的战略价值，将精力聚焦于创新与复杂决策等非信息搬运类的关键活动上。 在拥有约 10,000 名员工的全球能源巨头 AES Corporation，管理层正在践行这种“推倒部门围墙”的重构理念。虽然早期评估的两年周期被证实过于激进，但团队认为在 3 至 5 年内实现大约 80% 的组织结构重塑是切合实际的目标。IT 部门在此过程中正转型为赋能中心，为全员提供安全护栏与开发基础设施；针对在转型中面临身份认同困惑的跨学科复合型人才，Boyd 建议不要用传统刚性职能标签自我设限，而应主动扮演“组织重构推动者”的角色。 将企业从分配稀缺认知的官僚体系重塑为以人类判断力和信任为核心的机体，面临的最大障碍并非算法算力或员工意愿，而是身处既有流程中的从业人员难以突破认知惯性去构想无官僚支架的新形态。如何建立针对判断力与信任成熟的量化度量体系，并在 3 至 5 年的重构周期内平稳化解组织摩擦，仍是包括 AES 在内的先锋企业正在应对的现实挑战。

## Background 该视频由 Ai4 于 2026 年 8 月 23 日发布在 YouTube 平台上，内容为 The AES Corporation 首席数字战略与 AI 官 Erin Boyd 的主题演讲。页面简介指出，本次分享跳出了 AI 智能体部署、扩展与治理的常规讨论，转而探讨当智能本身变得极其充裕且廉价时，现代企业最初赖以建立的组织边界与分工逻辑将如何被根本性重构。 ## TL;DR 演讲从经济学与技术史的第一性原理切入，指出历次技术革命虽降低了能源与信息门槛，但现代企业依然建立在“管理稀缺的人类智能”与科斯交易成本之上，导致大量高素质人才沦为官僚链条中的“信息路由器”。随着 AI 让专业智能走向充沛，企业的核心命题不再是机械地在既有流程中插入智能体，而是彻底反思现存架构的合理性并转向追求“能力密度”。未来组织的胜负手在于构建多智能体协同的组织记忆交响乐，并在关键节点保留体现人类审慎价值的“判断界面”，使企业真正升华为凝聚信任与专有上下文的智慧容器。 ## Atmosphere & style 整场演讲兼具宏大的商业史纵深与严密的经济学思辨，主讲人 Erin Boyd 的语调冷静从容、洞见深刻且极具说服力。她以层层递进的批判性叩问打破了常规技术布道的叙事惯性，给人一种直击本质的认知震撼；在随后的问答环节中，她结合万人跨国企业的真实转型阵痛坦诚交流，既有宏观战略的远见，又充满务实落地的温度，现场互动真实而发人深省。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=4s) 以往历次技术革命（蒸汽、电力、计算机、互联网）消除了体力、能源和信息的稀缺，但人类智能与判断力始终是整个经济系统中最恒定稀缺的资源。 - [05:06](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=306s) AI 颠覆了智能本身的经济学属性；企业界不应只问“如何在企业中部署 AI”，而应反思“现代企业这种组织形式是否还适配 AI”。 - [07:17](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=437s) 科斯定律指出企业因降低市场交易成本而生；现代企业本质上是人类为克服个体认知极限、降低智能协调成本而发明的制度工具。 - [11:12](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=672s) 企业官僚体制是过去管理稀缺智能的产物；AI 的真正目的不是取代人类智能，而是停止将人类专家当作“信息路由器”的巨大浪费。 - [14:05](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=845s) 当前最大的战略误区是在为旧组织做渐进修补；管理者应基于第一性原理，重新审视现有审批流与管理层级背后的经济学假设是否依然成立。 - [17:34](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1054s) 数字化转型与 AI 转型的本质区别在于：前者改变了企业的“运营方式”（提高效率），而后者要求重塑企业“为何以这种方式运作”（架构重组）。 - [20:55](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1255s) 技术壁垒不在于构建单个最聪明的 Agent，而在于打造协调多个智能体的“交响乐团”；若各部门孤立部署 AI，最终只会制造“运行在 GPU 上的官僚体系”。 - [22:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1368s) 产品设计正从“减少摩擦”转向“设计判断”；软件不再只是被动工具，关键在于设计“判断界面”，决定何时将关键摩擦留给人类审慎裁决。 - [25:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1548s) 法务与合规应当转型为“信任的架构师”；面对 AI 落地，核心使命不是阻止自主系统，而是构建可解释、可审计的信任系统，兼顾速度与合规。 - [27:38](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1658s) 企业高管的考核指标应从追求人力规模转向追求“能力密度”；当顶尖大模型平民化后，真正的护城河是模型所不具备的专有上下文与组织记忆。 - [31:21](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1881s) 在智能泛滥充沛的世界中，信任成为最稀缺的资源；企业的最终形态将从“协调交易成本的工具”升华成“容纳人类智慧与判断力的容器”。 - [34:44](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2084s) AES 实践表明，大企业落地 AI 最大的瓶颈在于员工深陷过往流程细节，极难跳出思维惯性去想象没有官僚体系的工作流应该如何运转。 - [41:41](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2501s) 在智能泛滥的转型期，跨学科复合型人才不应强行给自己贴传统岗位标签，而应利用未被定义的弹性优势作为跨界协同的催化桥梁。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=4s) 技术革命的历史规律与人类智能的持续稀缺 演讲开篇抛出了一个颠覆性的前置假设：如果说以往每一次重大技术革命都改变了人类工作的方式，那么当前的 AI 革命是否正在改变组织从一开始为何需要人类员工的底层逻辑？若是如此，围绕人类员工构建起来的现代企业又将何去何从？ 回顾过往历史，每当重大技术变革降临，伴随而来的往往是对“人类工作终结”的末日预言，但每一次这些预言都被证明是错误的。在工业革命时期，蒸汽机彻底重塑了制造业，机器的转速、起重能力和生产效率远超人类，无数传统行当消亡，数百万人被迫转岗。然而人类并没有被淘汰，因为机器无法提供经济体系依然依赖的关键要素——判断力、适应性、复杂问题解决能力，以及在面对从未见过的陌生情境时理清头绪的能力。稀缺资源不再是体力与肌肉，而是人类的智慧。 随后的电气化革命让工厂不再受限于单根蒸汽传动轴，生产线得以重组，生产变得更快、更安全、更高效，工人再次凭借不可替代的人类判断力存活了下来。当计算机时代来临，人们曾断言会计师和分析师将成为历史遗迹，然而企业雇用的知识工作者反而比以往任何时候都多，因为计算机替代的是“计算”（Computation）而非“理解”（Understanding），人类依然是信息的解释者。到了互联网时代，信息几乎变得完全免费，任何人只要打开浏览器就能获取以往需要整座图书馆或多年经验才能掌握的知识。但信息不等于智能，知道某件事并不等同于知道该怎么做，人类依然扮演着翻译者、综合者和决策者的角色。 每一次革命都在急剧降低某种要素的稀缺性——体力、能源、计算能力和信息，但有一项要素始终顽固地保持着稀缺：可靠获取的人类智能，即专业知识、判断力、在充满不确定性的环境中进行推理并在缺乏规则手册时做出决策的能力。 - **历次革命消除的稀缺要素**： - 蒸汽革命：消除了对人类基础体力和肌肉力量的稀缺依赖。 - 电气革命：消除了能源传输与动力分配的空间限制。 - 计算机革命：消除了大规模数据运算与处理能力的稀缺。 - 互联网革命：消除了获取信息与知识的门槛与高昂成本。 - **贯穿历史的恒定核心**：人类智能（专业经验、不确定性下的逻辑推理、非结构化决策）始终是整个经济系统中最稀缺的资源。 ```mermaid flowchart LR accTitle: 技术革命演进与稀缺要素转移 subgraph S1["传统技术革命（工具重塑）"] A["蒸汽与电气革命<br/>（体力/能源丰裕）"] --> B["计算机与互联网<br/>（算力/信息丰裕）"] end subgraph S2["根本分水岭"] B --> C{"稀缺边界转移"} C -->|"唯一恒定稀缺"| D["人类智能与判断力<br/>（企业围绕其构建）"] end subgraph S3["AI 革命（底层逻辑颠覆）"] D --> E["AI 时代<br/>（智能本身走向丰裕）"] end ``` --- ### [05:06](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=306s) 核心命题：当智能变得充裕，企业还适用吗？ 一个多世纪以来，技术不断重塑工作的工具，蒸汽改变工厂，电力改变生产，计算机改变日常办公，互联网改变沟通协作，但现代企业本身的组织架构变化却微乎其微。原因在于，以往的技术只是改变了人们工作的方式，而 AI 可能是第一个不仅改变工具，而且动摇现代企业立足基石的技术——它颠覆了“智能本身的经济学属性”。 这引出了整个演讲的核心问题："当智能变得充裕时，企业会发生什么变化？"（What happens to the enterprise when intelligence becomes abundant?）两百多年来，技术革命不断推进稀缺性的边界，却从未跨入智能的领地；而 AI 正在跨越这道界限。 当前绝大多数企业在 AI 战略上的思考方式都存在一个巨大的盲区：大家都在问“我们该如何在企业中部署 AI？如何实现规模化应用？”这种提问方式默认了一个核心前提，即企业本身的组织形态是不变的。但如果 AI 不仅仅是另一个可供部署的工具，我们更应该反问："企业这种组织形式，是否还能适配 AI？"（Does the enterprise still fit AI?）如果战略仅仅停留在寻找 AI 的落脚点，企业可能会忽视真正令人彻夜难眠的危机——AI 是否正在让现代企业这种组织结构在经济学层面上走向过时？ --- ### [07:17](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=437s) 企业存在的本质：科斯交易成本与专业分工的极限 要理解为什么企业组织可能会被颠覆，必须回归一个基础而本质的经济学问题：如果自由市场真的如此高效，为什么所有人不直接作为独立自由职业者工作？为什么人们要自愿聚集到庞大、官僚、繁琐且迟缓的组织机构（即企业）中？ 诺贝尔经济学奖得主罗纳德·科斯（Ronald Coase）在近一个世纪前给出了经典解答：市场的力量极其强大，但在市场中协调一切却令人精疲力竭。想象一下，如果要在没有任何固定雇员的情况下创办一家软件公司，每开发一个新功能都要去市场上寻找合适的工程师；每次安全审查都要临时雇佣安全专家；遇到法律问题要重新物色懂业务的律师；每次产品发布都要重新组建全新团队。紧接着是无休止的合同谈判、背景资质验证、系统架构解释、权限授予、建立信任与工作协调。当项目一结束，一切又得推倒重来。经济学界将这种令人疲惫不堪的过程统称为“交易成本”（Transaction Costs）。 企业本质上是人类为了解决交易成本而发明的一种经济创新。企业不再每天清晨重新组建团队，而是一次性建立团队：招聘员工、沉淀专业经验、制定共享流程、构建内部信任、设立共同目标并积累组织制度性知识。最关键的是，企业让内部协作取代了每一次交互的商业谈判。因此，企业的本质不是员工的简单聚合，而是一种大幅降低“智能协调成本”的制度工具。 与此同时，企业扩张的另一大驱动力是个体心智能力的极限。无论一个人多么优秀，都不可能同时精通云原生架构、网络安全、产品设计、财务风控、劳动法务以及大客户销售。面对复杂的商业目标，企业必须将大问题拆解到不同专业领域（即职能部门），让工程师负责研发，安全团队负责防御，财务评估商业回报，法务把控合规边界，产品团队连接技术与客户需求。企业的价值不仅在于雇佣了许多聪明人，更在于将无法装入单个人类大脑的海量跨领域知识整合在了一起。 单纯的专业分工仍不足以解释为何优秀的企业极难被抄袭复制。竞争对手完全可以挖来同等履历背景的工程师、设计师和销售人员，却依然无法瞬间复制一家同样成功的公司。这其中的核心资产，正是人与人之间随着时间沉淀下来的“组织智能”（Organizational Intelligence）与共享上下文（Shared Context）： - 团队清楚哪些捷径暗藏杀机，哪些系统架构极其脆弱； - 明确知道哪些特殊客户需要定制化关注，哪些看似合理的方案过去曾经尝试过但最终被证明不可行； - 沉淀了深度的信任、工作默契、组织内部专用语言以及预判彼此行动的能力。 这些隐性知识绝大多数从未被完整写入文档，而是深植于员工关系、工作习惯、实战经验与企业文化之中。 --- ### [11:12](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=672s) 官僚体制的经济学根源：AI 是为了停止浪费人类智能 过去一百年间，要办成大事就必须建立大公司。企业是人类将稀缺智能转化为可靠、协同且可规模化调用的核心机制。然而 AI 正在打破这一运行了上百年的规则：它能按需提供各领域的专业知识，能够保存并检索复杂的上下文，能够跨领域协调工作推进，能够监控任务执行，并接管以往必须依靠增加人力才能完成的工作。 这并不意味着企业或人类员工会瞬间消亡，但它引发了一个尖锐的问题：我们是否正在试图把一项 21 世纪的技术强行塞进一个 20 世纪的组织制度中？当 AI 极其剧烈地降低了获取和协调智能的摩擦力时，企业会走向过时吗？所有人都会被智能体（Agents）取代吗？ 答案是否定的。人本身不会过时，真正开始走向过时的是人类过去为了“管理稀缺智能”而不得不建立起来的庞大官僚体制。 企业内部的官僚体制——包括无休止的汇报会议、层层审批、问题升级机制、状态同步报告和治理流程——从来不是因为组织天生喜欢复杂性而诞生的，而是为了解决特定的经济学问题。在一个多世纪里，这些层级机制是极其理性的设计，它们降低了协调拥有稀缺知识的人才时的沟通成本，推动智能在组织内部艰难流转。 然而，这种设计导致了巨大的悲剧：企业往往将极其出色的人才当成了“信息路由器”（Information Routers）。 - 雇佣了顶尖的工程师，却让他们把半周的时间浪费在各处翻找信息和拉通对齐上； - 雇佣了卓越的法务专家，却让他们深陷在繁琐日常的合同审批与文书归档中； - 雇佣了富有天赋的项目经理，却让他们沦为编写周报和汇总状态的协调机器。 人类的价值并未丧失，真正的悲哀在于我们一直在让人类从事本该由机器承担的工作。"AI 的目的不是替代人类智能，而是停止对人类智能的浪费。"（The purpose of AI isn't to replace human intelligence. It's to stop wasting it.） ```mermaid flowchart TD accTitle: 官僚体制在 AI 时代的重构 subgraph OLD["传统企业架构（管理稀缺智能）"] direction TB O1["专业知识高度分散且稀缺"] --> O2["构建官僚层级与汇报链条"] O2 --> O3["高昂协调成本<br/>（会议 / 审批 / 状态报告）"] O3 --> O4["人类专家沦为‘信息路由器’"] end subgraph NEW["AI 赋能架构（智能按需供给）"] direction TB N1["AI 智能体按需调度上下文与知识"] --> N2["消除跨职能流转与路由摩擦"] N2 --> N3["极简自适应协作网络"] N3 --> N4["人类精力聚焦非结构化判断与创造"] end ``` --- ### [14:05](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=845s) 停止提问“在哪里部署下一个智能体”，回归第一性原理 面对技术拐点，管理者必须彻底摒弃行业中最常见、也最令人厌倦的提问惯性："我们还能在哪里再部署一个 AI Agent？" 回到办公室后，领导者应当穿透企业内部的日常表象，审视每一个审批节点、每一场例会、每一个管理层级、每一处工作流流转与组织边界，并带着第一性原理提出一个最简单的经济学问题："这个机制究竟为什么存在？它最初是为了解决什么经济问题而设计的？" - 是因为当时专业知识必须高度集中在特定人员手中吗？ - 是因为智能在当时难以跨部门调用，导致决策必须在专家之间逐级流转吗？ - 是因为当时协调数千名员工是解决超出单一大脑处理极限的复杂问题的唯一可行方式吗？ 如果答案是肯定的，那么绝不能想当然地认为这些流程与层级在今天依然应当原封不动地保留。AI 带来的真正历史机遇，绝不是去把“昨天的企业”自动化，而是去深刻质疑“昨天的企业模式是否依然是最优的经济解决方案”。 任何一家组织都建立在一整套底层假设之上：关于专业知识应当驻留在何处、决策应当由谁制定、工作应当如何流转、信任机制在何处不可或缺，以及部门为何必须这样划分。大部分现代管理者只是继承了这些假设，极少主动质疑过它们，因为在过去智能稀缺、专业昂贵、协调极难的世界里，这些假设是完全经济理性的。但经济学规律揭示了一个不可动摇的真理：当底层的经济学条件发生根本逆转时，最适配的组织机构形态也必然随之重构。 当前许多企业正在犯下的最大战略错误，就是在为一个本该被彻底重新构想的组织形式做局部修补与渐进式优化。诸如“AI 可以自动化哪些任务？”“AI 能取代哪些岗位？”“在哪里部署下一个 Agent？”这类问题虽然合情合理，但它们全部属于极其下游的战术问题。这类问题的前提依然是企业结构保持原样、AI 只是作为插件接入其中。下一代企业的核心竞争优势绝不会诞生于这种修修补补之中，它属于那些敏锐意识到“AI 不仅在改变工作内容，更在颠覆百年来组织工作的底层经济逻辑”的领导者。 --- ### [17:34](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1054s) 致数字化转型负责人：从“数字化流程”到“重构组织架构” 在需要彻底重塑企业的历史节点上，不同专业角色承担着截然不同的使命。针对在场主持和推动企业转型变革的领导者们，回顾过去二十年的职业生涯具有极大的启发意义。 在过去二十年里，转型团队的核心使命是“企业数字化”（Digitize the enterprise），而这一使命在当时是完全正确的。行业付出了巨大努力： - 将纸质单据全面转化为数字像素； - 打通烟囱式的孤立软件系统，推动业务流程标准化； - 自动化重复性机械劳动，沉淀出前所未有的企业级业务数据。 这一轮数字化进程带来了非凡的成果：企业管理者第一次拥有了穿透全局的视野，能够交叉比对来自不同系统的数据，识别以往肉眼不可见的业务模式，做出更快、更好甚至以往完全无法想象的全新决策。技术极大地提升了组织认知和协调的速度与质量，帮助企业捕获了巨大的套利空间与市场价值。 然而，在过去二十年数字化的繁荣表象之下，一个核心底层假设始终稳固如初：人类智能依然是整个系统中最稀缺的资源。技术所做的只是加速信息流转、自动化例行常规流程、提炼洞察线索，但重大决策依然必须由人做出，跨部门的专业知识依然必须由人串联，当既定流程走到边界时依然必须由人行使最终裁决。 **两代转型的本质差异**： - **过去 20 年的数字化转型**：改变了企业的“运营方式”（How the enterprise operated）。重点在于消除信息孤岛，用像素取代纸张，在既定架构边缘提升运作效率。 - **当前的 AI 转型**：要求我们重新审视企业“为何以这种方式运作”（Why it operates the way it does）。重点在于质疑现存架构是否依然是最优设计，推动组织结构的根本性重塑。 因此，转型负责人的岗位职责已经发生了根本转变：工作不再是简单地将业务流程数字化，而是重新设计工作本身的发生方式。面对每一项流程，必须停止询问“哪里可以塞进一个 Agent”，而是质问：“这一工作流建立在关于工作的哪些既定假设之上？它之所以被设计成在专家之间层层流转，究竟是因为过去的知识壁垒，还是因为今天依然必要？” ### [20:55](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1255s) 重新设计而非自动化：构建协作交响乐，拒绝“GPU上的官僚主义” 在智能成本高昂时代设立的假设一旦改变，现有的组织流程也必须随之动摇。Erin Boyd 建议在座的管理者下周走进公司，审视每一个会议、每一项审批、每一条工作流、每一张电子表格、每一次手动交接和管理层级，并扪心自问：“如果我们今天掌握了现有的 AI 技术去从零构建这家组织，还会把它设计成现在的模样吗？”她认为所有人的答案必然是否定的。既然答案是否定的，那就绝不能仅仅对旧流程进行自动化，而是必须彻底重构——“不要把接下来的十年花在对昨天的企业进行数字化上，而要花在设计明天的企业上”。 对于技术构建者而言，其职业本质已在悄然间发生转变：工程师不再只是编写软件，而是在设计“智能在企业内部流动的路径”。在智能日益充沛的时代，最稀缺的资源变成了“协同”（Coordination）。 Boyd 强调，不要试图去耗费十年构建所谓“最聪明的独立智能体”，因为市场上总有人会训练出更聪明的模型；真正的壁垒在于“打造最出色的交响乐团（Orchestra）”，让专业化的智能体协同运作，建立能够沉淀并保留上下文的组织共享记忆架构，并将信任嵌入每一次交互。她警示道，如果各业务部门各自为政构建孤立的 AI，最终只会催生昨天的官僚体系——“唯一的区别仅仅在于，这一次的官僚机构是运行在 GPU 之上的”。 ```mermaid flowchart TD accTitle: 从孤立智能体到智能企业交响乐 subgraph S1["传统自动化陷阱"] A["孤立部门需求"] --> B["孤立的专有智能体"] B --> C["碎片化流程与冗余协调"] C --> D["运行在 GPU 上的官僚体系"] end subgraph S2["智能企业重构设计"] E["充沛的基础智能"] --> F["专业化智能体协同编排"] F --> G["共享组织记忆架构"] G --> H["敏捷自适应的企业操作系统"] end ``` ### [22:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1368s) 从“减少摩擦”到“设计判断”：产品设计与产品经理的范式转移 过去十年间，产品设计的核心使命几乎完全围绕着“减少摩擦”（Friction）。产品经理与设计师对用户体验极度痴迷：用户需要点击多少次？按钮应该放在哪里？如何砍掉一秒钟的等待时间？这种执着曾成功将软件从人们勉强忍受的工具变成了令人愉悦的体验。然而，每一代人都只能解决时代技术允许他们解决的问题。在智能体世界中，软件不再被动等待人类点击交互，而是能够自主推理、行动、协调并执行。产品经理的职责因而彻底颠覆——不再是构建“人类完成工作的地方”，而是“设计人类判断与机器执行之间的关系”。 过去几十年的产品设计始终在追问“我们如何协助人类完成任务”，而在智能体时代，核心问题转变为“人类究竟何时才需要介入”。这意味着设计师的任务不再是一味消除摩擦，而是艰难而克制地决定“摩擦应当留存在何处”。Boyd 指出，并非所有摩擦都是负面的，摩擦往往代表着人类的审慎判断、道德伦理、同理心与终极责任： - **消除摩擦的场景**：智能体预订会议、采购办公耗材等高频、低风险的日常琐事。 - **植入摩擦的场景**：做出解雇员工的决定、审批数百万美元的巨额商业合同，或是回应一位企业苦心维系数十年的核心客户。 在这些关键节点，必须通过精心设计的“判断界面”（Judgment Interfaces）而非单纯的用户界面，让人类的深度思考无缝切入工作流。衡量产品成功的指标也必须随之重写：不再是用户在应用界面停留了多长时间，而是“正确的智能（无论是人类还是人工智能）是否在最恰当的时刻介入”。 ### [25:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1548s) 信任的架构师：合规与法务的使命重塑 面向合规官和法务团队，Boyd 幽默地指出，尽管几十年来他们常常被企业其他部门戏称为“只会说‘不’的部门”（Department of No），但他们真正的身份应当是“信任的架构师”。经济学将此类机制称为交易成本，而在本质上，这正是“信心的代价”。任何经济体与企业的运转都仰赖信任——确信合同会得到遵守、系统行为具有可预测性、决策能够被合理解释，以及在出现偏差时有人能够承担责任。 面对 AI 浪潮，合规与法务的核心使命绝非延缓技术采纳，而是让 AI 的落地具备可信度。不能把精力放在“如何阻止自主系统”上，而应聚焦于“如何让自主系统承担责任”。Boyd 给出了清晰的行动准则：“不要去建立对模型的信任，要去建立对系统的信任。”这就要求团队搭建具备可解释性的决策系统、保留完整审计追踪（Audit Trails）、确立让人人都能理解的治理框架。在当下的 AI 经济中，“失去信任的速度是混乱，而失去速度的信任则是被淘汰”（Speed without trust is chaos, but trust without speed is irrelevance），合规架构的目标正是让组织不必在速度与信任之间做单选题。 ### [27:38](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1658s) 丰富智能时代的战略：能力密度与不可替代的稀缺资产 对于肩负战略重任的企业高管而言，战略决策的本质始终是分配稀缺资源——资本、人才、时间与技术。但在商业史上，核心生产要素的属性首次发生了根本性颠覆：智能本身正变成充沛且廉价的商品。然而，智能的充沛绝不等于判断力的充沛、信任的充沛，更不意味着竞争优势随处可见。高管的职能正在从“基于稀缺智能设计组织”转向“围绕充沛智能重构组织”。 在评估企业各项业务活动时，高管们必须告别以往单纯的“自建还是外采”、“外包还是留在内部”的思维，转向提出更本质的质问： - **这项活动是在创造竞争优势，还是仅仅在协调信息？** 如果只是协调信息，AI 必然会以更低成本、更高质量将其替代。 - **管理指标转向“能力密度”**：不再为了追求人力规模（Headcount）而扩张，而是追求“能力密度”（Capability Density）——即每一个个体在智能系统深度放大下所能爆发出的综合价值产出。 当世界上每一家企业都能平等接入前沿顶尖大模型时，模型本身便不再构成护城河。真正的竞争壁垒全面转向“模型所不知道的一切”：企业的专有业务上下文、组织记忆、真实的客户关系、企业文化、商誉声誉以及长期沉淀的直觉判断。Boyd 警告高管切莫将 AI 当成另一次普通的“技术推广项目”（Technology Rollout），历史不会铭记部署了最多智能体的高管，只会铭记那些拥有勇气彻底重塑企业形态的领袖。 ### [31:21](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1881s) 科斯之问的终极答案：企业因何而存在？ Boyd 将论述引回了现代组织理论的起点。1937年，经济学家罗纳德·科斯（Ronald Coase）提出了看似极简的命题：“企业为何存在？”科斯给出的答案深刻统治了近一个世纪的管理学实践——企业之所以存在，是因为市场交易成本昂贵，跨市场协调与获取专业知识的摩擦极其巨大。为了克服这些摩擦，人类招募雇员、设立层级、划分部门、建立严密的管理与流程体系。在过去百余年中，这种组织形态在“协调稀缺智能”方面展现出了无与伦比的高效性。 然而，当获取智能的门槛全面归零，交易摩擦不再是企业存在的核心阻碍时，企业究竟是为了什么而存在？Boyd 指出，企业的最终形态必将升华为“人类智慧的容器”： - 在这里，解决切中要害的真问题重于日常琐碎的行政流程； - 人类的审美与判断重于层层汇报与任务协调； - 彼此建立的信任重于冰冷的死板规章。 她畅想，未来的子孙后代审视当下的企业运作时，或许会像我们今天打量中世纪纸质账簿一样，感到不可思议与荒谬：为什么当年顶尖的工程师要把大好下午浪费在手动更新电子表格上？为什么项目经理整周都在催促进度报告？为什么高薪法务要耗费几千小时比对几乎雷同的文本？每一代人都继承着基于昨日经济学构建的制度，但极少有哪一代人能获得重新发明制度的历史机遇。 > “在智能日益充沛的世界里，信任成为了最稀缺的资源。而经济学告诉我们，价值永远栖身于稀缺之中。” ### [34:44](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2084s) 现场问答：AES的落地重构实践与组织挑战 在演讲结束后的观众互动环节中，听众针对落地实施细节进行了深入提问，Boyd 结合 AES Corporation 的实践给出了坦诚的分享： - **AES 正在推行的变革实践**：Boyd 透露，AES 内部正在践行一种近乎“拆除所有墙壁”的归零思考模式。作为一家拥有能源重资产的企业，团队抛开现有官僚流程包袱自问：“如果我们今天从零开始建设与运营能源资产，在没有任何历史遗留系统与流程的前提下，数字化、AI 和人类员工各自应当扮演什么角色？”在实施策略上，团队从最初宏大而分散的“试图煮沸整片海洋”，收敛聚焦到少数关键核心业务领域进行深度重构。 - **角色重构与员工抵触心理**：针对岗位职责的重塑，Boyd 明确表示所有既定角色都在重构范围内。但重塑的核心绝对不是“消灭人类岗位”，人类的核心价值在于终极判断力；变革的目标是“消灭官僚流程”，精准找出人类最该身处的位置，让智能系统成倍放大其工作产出。 - **量化信任与判断的指标探索**：当有产品策略背景的观众询问如何量化“信心、判断力与信任”时，Boyd 坦言整个行业目前依然处于摸索规则的“狂野西部”（Wild West）阶段。目前虽然能借助用户的持续回流度、交互响应速率等行为指标进行侧面观察，但科学、统一的量化评估体系仍在摸索构建中。 - **IT 部门定位的根本转变**：针对大模型普及对传统 IT 职能的冲击，Boyd 指出 IT 部门必须与法务、合规紧密协同，其核心职能正在转向：为非技术背景业务人员扫清一切门槛，提供安全、极简的工具链与清晰的安全护栏，让全员能够极其无缝地进行创新成果的部署、生产化应用与全司共享。 - **全球万人企业的落地时间表**：面对跨国企业“大船难调头”的实施周期疑问，Boyd 坦承自己在半年前曾乐观估计需要 2 年，但在深入面对全球 10,000 名员工的庞大体量后一度感到极其艰难。她目前将目标调整为：在未来 3 至 5 年内实现 80% 核心流程的重构。最大的落地瓶颈往往不在于员工的技术接受度，而在于业务人员深陷过往流程细节太久，极难跳出既有惯性去想象“没有官僚体系的工作流程到底该长什么样”。 ### [41:41](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2501s) 跨学科转型与重塑自我：在智能泛滥时代打破职能标签的束缚 现场观众提出了一个在组织变革中极为普遍的身份认同困境：他在公司内部正致力于重塑自身的工作方式与定位，却发现自己很难被塞进现有的传统岗位头衔体系中。他既不是传统意义上的产品经理（Product Manager），也不是单纯的软件工程师，而是一个深度结合领域知识、专注于构建 AI 产品的业务专家（Subject Matter Expert, SME）。面对这种兼具多学科复合特征的全新形态，他向 Erin Boyd 请教：当个人身处如此深刻的范式转变当中，该如何向外界清晰阐释自己的工作内容，又该如何概括并定义这种融合了多元视角的跨领域新型角色？ 针对这一困惑，Erin Boyd 给出了一个反直觉却极为务实的建议——“不要试图定义你自己”（"don't try and define yourself"）。她结合自身的经历指出，尽管自己拥有经济学博士学位、本质上是一名经济学家，但在当下的创新环境中同样处于这种复合角色之中。从切身体会来看，一旦过度追求一个精准的职能标签，外界反而容易用固有的思维定势将你“框住”（"box you in"），从而限制了跨界协同的施展空间。 Erin 建议，与其纠结于头衔，不如坦然接纳并驾驭这种“未被定义”的优势（"own it and run with it"）。在与跨团队成员沟通合作时，与其一开始就亮明特定工种的预设立场，不如以更开放的姿态去协助对方“重塑工作流程与业务想象”（"I'm here to sort of help you reimagine"）。这样能够让对话自然展开，促使业务伙伴毫无防备地畅所欲言；而作为融合了多重技能与视角的复合型人才，完全可以根据不同合作对象的具体痛点与语境，动态调整自身各项专业能力的权重组合，在组织中发挥不可替代的桥梁与催化作用。 - **传统角色定义的局限**：在 AI 驱动组织变革的背景下，垂直领域专家（SME）与技术构建者的界限日益模糊，强行套用产品经理或工程师等既有标签容易窄化个人的实际价值。 - **避免被标签禁锢**：过度明确定位往往会导致团队协作中的认知偏见与壁垒，保持角色的弹性与非定义状态反而能打开更广阔的探索空间。 - **动态权重适配**：复合型创新者应将自身拥有的多元能力视作一个自适应的工具箱，面对不同利益相关者时灵活调配沟通视角与专业侧重点。 ## Notable quotes > [05:06](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=306s) What happens to the enterprise when intelligence becomes abundant? — Erin Boyd > [05:06](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=306s) Does the enterprise still fit AI? — Erin Boyd > [11:12](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=672s) The purpose of AI isn't to replace human intelligence. It's to stop wasting it. — Erin Boyd > [20:55](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1255s) 不要把接下来的十年花在对昨天的企业进行数字化上，而要花在设计明天的企业上 — Erin Boyd > [20:55](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1255s) 唯一的区别仅仅在于，这一次的官僚机构是运行在 GPU 之上的 — Erin Boyd > [25:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1548s) 不要去建立对模型的信任，要去建立对系统的信任。 — Erin Boyd > [25:48](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1548s) Speed without trust is chaos, but trust without speed is irrelevance — Erin Boyd > [31:21](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=1881s) 在智能日益充沛的世界里，信任成为了最稀缺的资源。而经济学告诉我们，价值永远栖身于稀缺之中。 — Erin Boyd > [41:41](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2501s) don't try and define yourself — Erin Boyd > [41:41](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2501s) own it and run with it — Erin Boyd > [41:41](https://www.youtube.com/watch?v=Z0Bm8L0jysY&t=2501s) I'm here to sort of help you reimagine — Erin Boyd ## Mentioned Erin Boyd, The AES Corporation, Ai4, Ronald Coase (罗纳德·科斯), GPU ## Video description Erin Boyd - Chief Digital Strategy & AI Officer, The AES Corporation Most discussions about AI agents focus on deployment, scale, and governance. This session explores a different question: what happens if they succeed? Modern organizations were built on a simple premise: bringing people, expertise, and decision-making together inside a company was more efficient than coordinating them through the market. As AI agents reduce the cost of accessing knowledge, coordinating work, and making decisions, leaders may need to rethink not only how work gets done, but why organizations are structured the way they are in the first place.Rather than asking how to scale AI agents, attendees will be encouraged to ask a more fundamental question: what would they design differently if they were building their organization today in a world of abundant intelligence? [Source](https://www.youtube.com/watch?v=Z0Bm8L0jysY)

[5:06] Erin Boyd: What happens to the enterprise when intelligence becomes abundant? [5:06] Erin Boyd: Does the enterprise still fit AI? [11:12] Erin Boyd: The purpose of AI isn't to replace human intelligence. It's to stop wasting it. [20:55] Erin Boyd: 不要把接下来的十年花在对昨天的企业进行数字化上，而要花在设计明天的企业上 [20:55] Erin Boyd: 唯一的区别仅仅在于，这一次的官僚机构是运行在 GPU 之上的 [25:48] Erin Boyd: 不要去建立对模型的信任，要去建立对系统的信任。 [25:48] Erin Boyd: Speed without trust is chaos, but trust without speed is irrelevance [31:21] Erin Boyd: 在智能日益充沛的世界里，信任成为了最稀缺的资源。而经济学告诉我们，价值永远栖身于稀缺之中。 [41:41] Erin Boyd: don't try and define yourself [41:41] Erin Boyd: own it and run with it [41:41] Erin Boyd: I'm here to sort of help you reimagine

[11:12] Erin Boyd: The purpose of AI isn't to replace human intelligence, it's to stop wasting it. [11:12] Erin Boyd: Why does this exist? [20:55] Erin Boyd: If I were building this organization today, knowing what I know about AI, would I design it this way? [20:55] Erin Boyd: Don't automate it. Redesign it. [22:49] Erin Boyd: When should the human be involved at all? [34:44] Erin Boyd: burning the walls down [34:44] Erin Boyd: everything is on the table [41:41] Audience Member: I found it very difficult to even fit myself in the role titles that are here because I'm not a product manager. I'm not an engineer. I'm essentially a subject matter expert building AI products, which is—I don't know what I am now. [41:41] Erin Boyd: Don't try and define yourself. [41:41] Erin Boyd: Own it and run with it and really just sort of go to people and almost just say, 'I'm here to sort of help you reimagine.'

[0:04] I want to sort of start with a with a premise.  What if every major technological revolution has [0:11] changed the way humans work? But what if the AI  revolution changes why organizations need human [0:20] workers in the first place? And if that's true,  what happens to the enterprise built around them? [0:28] So let's start with a worker. Take it back a  little bit. No matter the way the change occurred, [0:34] we have still always needed humans for work. And  I'm I'm talking the the work that you actually [0:39] get paid to show up to do, right? We we all  come to our jobs for many of us because we [0:43] get paid. So people people have always thought  otherwise, right? that it seems that with every [0:50] major technological revolution came and comes a  prediction of the end of human work and every time [0:59] they were wrong or rather they were mostly wrong  and apologies as I'm going to get a little bit [1:04] historical here. I'm just going to run through  a few things. So if you I think if you sort of [1:08] consider the industrial revolution right steam  engines transformed manufacturing machines could [1:15] spin faster than people lift more than people  produce more than people entire trades disappeared [1:21] and with the placement and with that the placement  of millions of workers and yet the workers [1:29] survived. Why? Not because humans were stronger.  The machines had already won that contest. [1:36] Humans survived because the economy still depended  on something machines couldn't provide. Judgment, [1:43] adaptation, problem solving, the ability to  make sense of situations the machine had never [1:49] encountered. The scarce resource wasn't muscle  anymore. It was human intelligence. And then came [1:57] the revolution of electricity. Factories no longer  needed to be built around a single a single steam [2:03] shaft. Production lines were reinvented. Work  became faster, safer, more productive. Entire [2:10] industries were reorganized. Again, people feared  for the worker. Again, the worker survived because [2:19] the instance of electricity amplified human  capability, but it didn't replace the need [2:24] for human judgment. And then computers arrived. We  were told office workers would disappear, right? [2:31] Accountants and analysts would become relics of  the past as computers would automate calculations [2:36] that once took days. And yet, organizations hired  more workers, right? More knowledge workers than [2:43] ever before. Why? Because computers replaced  computation. They didn't replace understanding. [2:51] They processed information. Humans interpreted  it. And intelligence remained scarce. And then [2:59] the internet arrived and suddenly information  became almost free. Knowledge that once required [3:05] libraries, experts or years of experience  was available to anyone with a browser. But [3:13] information isn't intelligence. Knowing something  isn't the same as knowing what to do. Humans [3:21] remain remain the translators, the synthesizers,  the decision makers. And once again, right, we [3:28] all know the worker survived, right? So, I think  you notice the pattern here, right? Every one of [3:34] these revolutions made something dramatically less  scarce, though. And that's what I want to point [3:38] out. The the worker survived, but something became  less scarce. Physical labor, energy, computation, [3:46] information, and the worker yet survived.  Because every one of those revolutions, one [3:52] thing remains stubbornly scarce. Not information,  not computation, not energy. Human intelligence [4:01] or perhaps more precisely reliable access to  human intelligence. I'm talking about again here [4:08] expertise, judgment, the ability to reason through  uncertainty, the ability to make decisions when [4:14] there isn't a rulebook. And that's where the story  changes because if intelligence remained scarce, [4:22] organizations had to organize around it. For more  than a century, technology kept reinventing the [4:29] tools of work. Steam changed factories.  Electricity changed production. Computers [4:36] changed work. The internet changed communication.  Yet the architecture of the enterprise changed far [4:43] less than we imagine because every one of these  technologies changed how people worked. AI may [4:51] be the first technology not merely to change the  tools of work but to change one of the assumptions [4:58] on which the modern enterprise itself was built  specifically the economics of intelligence itself. [5:06] So I really I want to throw a premise out here  a question. What happens to the enterprise when [5:12] intelligence becomes abundant? Again, for over 200  years, every technological revolution shifted the [5:20] boundary of scarcity, but it never crossed  the boundary into intelligence. AI might, [5:26] and it probably will, right? We're all here today  in this track talking about how to deploy AI into [5:32] our enterprises. Um, again, that's what this track  is all about. How do we add AI to the enterprise? [5:38] How do we deploy it at scale? It's a natural  question that we sort of rarely start to stop [5:44] to examine it, right? But it assumes a big  assumption that the enterprise stays the same. [5:52] But what if AI isn't simply another technology we  deploy? The answer to the question, how does AI [5:59] fit the enterprise? Maybe the question we should  be asking is, does the enterprise still fit AI? [6:07] Because every AI strategy I've seen starts from  the assumption that the organization is fixed. [6:16] We ask ourselves where should we use AI? But the  question that keeps me awake at night is actually [6:21] a different question, right? What if AI makes the  modern enterprise economically obsolete? Maybe it [6:30] should. What if it should? And for this, sorry  again, we need yet again to talk about a more [6:36] fundamental, a probably much more weirder concept  for a tech conference um for this audience. Right? [6:42] I'm going to ask a sort of theoretical question.  Why do any of your companies exist in the first [6:47] place? Right? If free markets are so incredibly  efficient, why don't we all work as independent [6:54] freelancers? Why do we voluntarily group  ourselves into the same enormous, bureaucratic, [7:01] often frustrating, and slowmoving institutions we  call enterprises? It's a serious question. I know [7:07] I'm sort of laughing as I as I go through it a  bit, smirking, but right, if markets allocate [7:11] resources so well, why does the enterprise exist  at all? And I'm sorry, I'm going to geek a little [7:17] bit on economics because I am I am an economist,  right? If economists have been wrestling with [7:22] this question now for nearly a century and one  of the most influential answers actually came [7:27] from Ronald Kosh. You might have heard of him.  His insight was actually surprisingly simple. [7:33] Markets are extraordinarily powerful but they're  exhausting. Right? They're not imperfect. They're [7:40] not expensive. They're exhausting. Imagine trying  to build a software company without employees. [7:48] Every feature you have to find the right engineer,  right? Every security review you have to locate [7:54] the right expert. Every legal question you have  to find an attorney who understands your business. [8:00] Every product launch you assemble an entirely  new team. Then you negotiate contracts, verify [8:06] credentials, explain the architecture, grant  access, build trust, coordinate everyone's work, [8:11] and then when the project ends, you start all  over again. That's exhausting. Economists have [8:17] a much less interesting name for it. We call  it transaction costs. And the enterprise was [8:22] essentially humanity's answer to that problem.  Instead of rebuilding a team every morning, [8:29] we built one once. We hired people, developed  expertise, created shared processes, built trust, [8:37] established common objectives, accumulated  institutional knowledge. Critically, we [8:42] stopped negotiating every interaction and started  collaborating. In other words, the enterprise [8:49] wasn't simply a collection of employees. It  was an economic innovation. It dramatically [8:55] lowered the cost of coordinating intelligence.  But there was another reason enterprises grew [9:01] the way they did. The limit of the individual  mind. No matter how capable one person may be, [9:08] she she absolutely cannot simultaneously master  cloud architecture, cyber security, product [9:15] design, f financial controls, employment law, you  know, customer special customers, their behavior, [9:21] enterprise sales, right? Complex outcomes required  specialized knowledge. So, enterprises divide [9:28] large problems into domains, right? We all know  about our silos in our companies. Engineers focus [9:34] on building the product. Cyber security protects  it, right? Finance evaluates its economics. Legal [9:42] interpret interprets its obligations. Product  teams connect technical capabilities to customer [9:48] needs. The enterprise creates a structure through  which these different forms of expertise can [9:54] contribute to the same objective. Its value does  not come merely from employing many smart people. [10:01] It comes from combining knowledge that could  not fit inside any one person. But again, [10:07] specialization alone does not explain why a  successful enterprise is actually difficult [10:12] to copy. A computer might hire hire or company  might hire equally talented engineers, designers, [10:19] lawyers, salespeople. Yet pulling together a  comparable resumes doesn't instantly create a [10:25] comparable company. Right? What is missing is  the intelligence that develops between people [10:31] over time. Right? We all know this in our  companies. Teams accumulate shared context. [10:37] They learn which shortcuts are dangerous, which  customers require sort of special attention, [10:42] which systems are fragile, right? We all have  those probably, which apparent problems have [10:47] already been tried and actually abandoned.  We develop trust in enterprises, routines, [10:53] language, judgment, an ability to anticipate one  another, right? We we build that sort of common [10:59] bond. And most of this knowledge is never fully  documented. We also know that as well, right? It [11:04] lives in relationships, habits, experience, and  organizational intelligence. Which brings me to [11:12] AI. For the last 100 years, to do big things,  you had to build a big company. For more than [11:20] a c a century, enterprises have been one of our  primary mechanisms for making scarce intelligence [11:27] reliable, coordinated, and available at scale. But  AI is changing the rules. It can make specialized [11:35] knowledge available on demand. It can preserve and  retrieve context, coordinate work across domains, [11:43] monitor execution, and perform tasks that  previously required additional human capacity. [11:49] Now, that does not automatically eliminate the  enterprise or the worker, but it does raise a [11:56] fundamental question. Are we trying to fit a  21st century technology into a 20th century [12:03] institution? Right? If the enterprise exists to  reduce the friction of accessing and coordinating [12:09] intelligence, what happens when AI radically  reduces that friction? Does the enterprise [12:16] become obsolete? Do we replace everyone with  agents? I'm going to say no. Uh but I do think [12:23] that something does become obsolete, not people.  I want to make that very clear. I don't think that [12:28] people become obsolete. The bureaucracy we built  to manage scarce intelligence is what's begun to [12:36] hopefully come become obsolete. Right? And think  about what bureaucracy actually is in a company in [12:42] an enterprise. It's meetings, it's approvals, it's  escalations, it's reporting, it's all of those [12:49] things, right? Governance routing work, right?  None of those emerged because organ organizations [12:55] enjoyed complexity, right? None of us enjoys  that. But it just it it sort of became necessary [13:00] to run them. It emerged because it solved an  economic problem, right? And for over a century, [13:06] those structures were rational. They reduced the  cost of coordinating people who possessed scarce [13:14] knowledge. They helped intelligence flow through  the enterprise. Right? We've taken brilliant [13:20] people and used them as information routers.  Right? We've hoard hired extraordinary engineers [13:26] that asked them to spend half their week looking  for information. We've hired brilliant lawyers [13:32] then buried them in routine you know that  asked them to document and review all kinds [13:37] of contracts. We've hired gifted project managers  then turned them into status report coordinators. [13:43] Right? The tragedy isn't that people aren't  valuable. The tragedy is that we've often used [13:49] them for work machines for for work that machines  should have been doing all along. So I would argue [13:57] that the purpose of AI isn't to replace human  intelligence. It's to stop wasting it. So what [14:05] should we do? Stop asking a question I hear all  the time. Please, I beg you to stop asking this [14:10] question. If there's one thing you you walk away  from, it's it's never to ask this question again. [14:16] where can I deploy another agent? Right? Rather,  when you return to your office, walk through [14:23] your organization, look at every approval, every  meeting, every management layer, every workflow, [14:30] handoff, organizational boundary. Then ask one  simple question. Why does this exist? Right? [14:38] What econ economic problem was this designed to  solve? Was it created because expertise had to be [14:45] concentrated? Because intelligence was difficult  to ask access because decisions had to move from [14:51] one specialist to another? Because coordinating  thousands of people was the only practical way to [14:58] solve a problem that exceeded the capacity of  any single mine. Right? If the answer is yes, [15:04] don't assume that it should survive unchanged  because that's the real opportunity AI presents. [15:11] not simply to automate yesterday's enterprise,  but to question whether yesterday's enterprise [15:18] is still the best economic solution. Every  organization is built on a set of assumptions. [15:24] Assumptions about where expertise lives, about who  should make decisions, about how work should flow, [15:30] where trust is required, why the organization  is structured the way it is. Right? Most of us [15:36] inherited those assumptions. Very few of us have  ever questioned them. We haven't really had to, [15:41] right? Not because they were wrong, but  because they were economically rational, [15:46] right? They reflected a world in which  intelligence was scarce, expertise was expensive, [15:53] right? Coordination was difficult, and for  more than a century, those assumptions gave [15:58] us the modern enterprise. But economics teaches us  something important. When the underlying economics [16:06] changes, the optimal institution often changes  with it. AI is giving us one of those moments, [16:14] not because it makes people less important, but  because it changes the economics that shaped the [16:21] enterprise in the first place. So if the economics  have changed, right, what should we do? Right? [16:29] Again, I think the biggest mistake organizations  are making today is that they're optimizing an [16:35] institution they actually should be reimagining,  right? They're asking which tasks can AI automate? [16:41] Which jobs can AI replace? Where should we deploy  another agent? My least favorite question on the [16:47] face of the earth right now, right? Those  are those are all reasonable questions, [16:51] although I would argue the last one's I don't  want to hear it again, but they're they're also [16:55] downstream questions, right? They all begin with  the assumption that the enterprise stays the same [17:01] and AI simply plugs into it. Right? That's not  where the next generation of competitive advantage [17:08] will come from. Competitive advantage will come  from leaders who recognize that AI is not simply [17:15] changing work. It's changing the economics on  which work has been organized for more than a [17:22] century. Right? And if redesigning the enterprise  is the challenge before us, then every discipline [17:29] in this room has a different responsibility.  So let me spend a few minutes speaking to each [17:34] of you in the room. Right? For those of you who  design transformation, I want you to look back [17:41] at the last 20 years of your career. Right? Your  mission was to digitize the enterprise. And it was [17:47] exactly the right mission. We digitized paper. We  connected systems. We standardized processes. We [17:54] automated automated repetitive work. We created  data that had never existed before. And something [18:00] extraordinary happened. For the first time,  leaders could truly see across the enterprise. [18:06] We could triangulate information triangulate  information from multiple systems. We could [18:13] identify patterns that had been invisible. We  could make better decisions, faster decisions, [18:19] and sometimes entirely new decisions that simply  weren't possible before. Technology didn't just [18:27] make work faster. It expanded what organizations  were capable of knowing. I've spent much of my [18:34] career leading digital transformation, and it  accomplished something remarkable. I think all [18:38] of us have accomplished something remarkable,  those of us that have designed it. It didn't [18:42] just replace paper with pixels. It connected  processes that had never been connected. It [18:48] gave us visibility across the enterprise. It  allowed information to move instantly instead [18:54] of sitting in filing cabinets or inboxes, and  gave leaders the ability to triangulate data [19:00] from across the organization, uncover patterns  we couldn't previously see, and make faster, [19:06] better decisions. It transformed the speed and  quality of coordination. It allowed us to seize [19:14] opportunities we would have never seen and allowed  us to arbitrage value. But underneath all of that, [19:22] one assumption remained remarkably constant. The  enterprise was still organized around the idea [19:29] that human intelligence was the scarce resource.  Technology could move information faster. It could [19:37] automate routine work. It could surface better  insights. But people still made the important [19:43] decisions. People still connected specialized  knowledge across functions. People still exercised [19:49] judgment when the processes reached its limit.  In other words, digital transformation changed [19:55] how the enterprise operated. AI gives us the  opportunity to rethink why it operates the [20:03] way it does. That's a fundamentally different  question and it's a fundamentally different [20:09] kind of transformation. It isn't simply about  making today's enterprise more efficient and [20:15] grabbing more value at the margin. It's about  asking whether today's enterprise is still the [20:21] optimal design. Which means for those who design  transformation, your job description has changed. [20:28] You are no longer simply digitizing work. You  are redesigning how work gets done and that [20:35] requires a different set of questions. Don't ask  where I can deploy another agent. Please ask what [20:42] assumptions about work is this process built on.  Was this workflow designed because expertise had [20:49] to be concentrated? Because decisions had to move  from specialist to specialist because coordinating [20:55] intelligence was expensive. If those assumptions  have changed, perhaps the process should too. Walk [21:02] through your organization next week, every meeting  again, every approval, every workflow, every [21:07] spreadsheet, manual handoff, management layer, and  ask one simple question. If I were building this [21:14] organization today, knowing what I know about AI,  would I design it this way? I think probably all [21:21] of us would say no. So, because if the answer is  no, which I think it is, right? Don't automate it. [21:27] Redesign it. That's your next transformation  opportunity. Don't spend the next time, [21:34] the next decade digitizing yesterday's enterprise.  Like, spend it designing tomorrow's. Right now, [21:42] for those of you that design technology, I think  your profession has quietly changed. You're no [21:48] longer just writing software. You're designing  how intelligence flows through an organization, [21:55] right? and what and that changes what matters.  Intelligence is becoming abundant. Coordination [22:02] isn't. So don't spend the next decade trying  to build the smartest best agent. Someone else [22:09] will always build a smarter model. I promise you.  Instead, build the best orchestra. Right? Build [22:16] systems where specialized agents collaborate.  Build shared organizational memory architectures [22:23] that perver preserve context. Build trust  into every interaction. And one final thought, [22:30] if every department builds its own isolated  AI, you've created yesterday's bureaucracy, [22:36] right? You've created another layer of corporate  bureaucracy. Only this time, it's running on GPUs, [22:42] right? Don't build intelligent agents. Build  an intelligent enterprise. Right? If you design [22:49] products for much of the last decade, your  your mission has been to reduce friction, [22:55] right? You obsessed over the user experience. How  many clicks? Where does the button go? How do we [23:00] simplify the workflow? How do we remove one more  second of waiting? And that was again the exactly [23:07] the right mission. You transformed software from  something people tolerated, right, into something [23:12] they actually enjoyed using. But every generation  solves the problems its technology allows it to [23:20] solve. Yesterday your job was to design great  interfaces. Tomorrow your job is to design great [23:28] decisions. Because in an agentic world software  no longer waits for a human to click a button. It [23:36] reasons, it acts. It coordinates. It executes.  That changes the role of the product manager [23:43] completely. You are no longer designing the place  where people do the work. Right? You are designing [23:49] the relationship between human judgment and  machine execution. Think about that for a moment. [23:56] Right? Think. For decades, product design asked  one question. How do we help the human complete [24:02] the task? Tomorrow the question becomes when  should the human be involved at all? Right? That's [24:08] a completely different question. It's a completely  different design problem. So your job is no longer [24:14] to simply remove friction. Your job is actually  now to decide where the friction belongs. Right? [24:20] That's a sort of an uncomfortable place I think  for a lot of uh product managers. Right? Because [24:25] of designers because not all friction I want to  tell you is bad. Sometimes friction is judgment. [24:31] Sometimes friction is ethics. Sometimes it's  empathy. Right? Accountability. If an agent is [24:38] booking a meeting, remove the friction. If an  agent is ordering office supplies, remove the [24:43] friction. But if an agent is about to terminate an  employee, approve a multi-million dollar contract, [24:49] or respond to a customer whose trust you've spent  decades earning, that's exactly where you want [24:56] thoughtful human judgment to enter the workflow.  You're no longer designing user interfaces. You're [25:04] designing judgment interfaces. You're designing  the moments where humans and AI collaborate to [25:12] make better decisions than either could make  alone. And perhaps your metric should change as [25:17] well. Success now isn't changed uh isn't measured  by how long someone spends in your application. [25:23] Right? In fact, your agents, if they're doing  their job so well, your you your users may [25:28] hardly see the application at all. Right? So  the real question is did the right intelligence [25:34] human or artificial show up at exactly the right  moment. Right? That's the product of the future. [25:42] Don't design better screens. Please design better  decisions. Now for those of you who design trusts, [25:50] my compliance officers, my lawyers, right? For  decades, people have called you the department [25:54] of no. And I I don't think that's what you are.  Although sometimes I feel that way. I think you're [25:59] the architects of trust. That's what I would  like to call you now. Right? Every economy, [26:04] every market, every enterprise depends on trust.  Trust that contracts will be honored. Right? [26:11] Trust that systems behave predictably. Trust that  decisions can be explained. Trust that someone is [26:18] accountable when things go wrong. As I mentioned  earlier, economists, we call this transaction [26:24] costs. And even though I'm an economist, I think  of them as the price of confidence. For decades, [26:31] your job has been to reduce uncertainty. AI  doesn't eliminate that responsibility. It [26:38] amplifies it. Because the faster organizations  move, the more important trust becomes. Your [26:45] mission is not to slow AI adoption. Your  mission is to make AI adoption trustworthy. [26:53] Don't ask how do we how do we stop autonomous  systems. Don't ask that. Ask how do we make [27:00] autonomous systems accountable. That changes  everything. Don't build trust in the model. [27:07] Build trust in the system. Build architectures  that explain decisions. Build audit trails that [27:13] preserve accountability. Governance that people  can understand. organizations that are confident [27:19] enough to move quickly because they know exactly  where the guard rails are because in the AI [27:25] economy speed without trust is chaos but trust  without speed is irrelevance right so your job [27:33] now is to ensure we never have to choose between  them and finally if you design strategy right my [27:40] executives my leaders in the room your job has  always been to allocate scarce resources capital [27:46] people time, technology, every strategic decision  you've ever made has really been an exercise in [27:53] econ economics. I'm an economist, I can say  that, right? Um, where do we invest? Where do [27:58] we build ourselves? Right? What do we outsource?  What capabilities truly differentiate us? Those [28:05] questions don't disappear in the age of AI. But  again, one of the variables has changed. For [28:12] the first time in history, intelligence itself is  becoming abundant. That doesn't mean judgment is [28:20] abundant. It doesn't mean that trust is abundant.  And it certainly doesn't mean competitive [28:26] advantage is abundant. It simply means one of the  fundamental inputs into your business has become [28:34] dramatically cheaper. So your job changes, right?  not from managing people to replacing people. It [28:42] moves from designing organizations around scarce  intelligence to designing organizations around [28:49] abundant intelligence. That again is a very  different leadership challenge. For decades, [28:54] executives have been asking, should we build it  or buy it? Should we outsource it or keep it in [28:59] house? Tomorrow, I think you'll ask a different  question. I hope you'll ask a different question. [29:04] Is this activity creating competitive advantage?  Right? or is simply coordinating information [29:11] because if it's merely coordinating information  AI will increasingly do it better but if it's [29:18] creating trust creating relationships creating  strategy creating innovation judgment those [29:25] become even more valuable. So don't optimize  for headcount optimize for capability density. [29:35] How much value can this organization create?  Right? Not because it has more people, but because [29:42] every person is amplified by intelligent systems.  And remember something else. If every company has [29:50] access to the same frontier AI model, then the  models are no longer your competitive advantage. [29:56] It's important. Your advantage becomes everything  the models don't know. your proprietary context, [30:03] your organizational memory, your customer  relationships, culture, reputation, [30:08] accumulated judgment. Those are the the assets  that cannot be downloaded, right? And perhaps [30:15] the rarest of all, trust as AI fills the internet  with synthetic context, synthetic conversations, [30:22] and synthetic expertise, right? Authentic human  relationships become more valuable, not less. In a [30:32] world where intelligence is increasingly abundant,  trust becomes scarce. And economics teaches us [30:40] that scarcity is where value lives. Right?  That's important. And one final thought, the [30:46] biggest mista mistake an executive can make is to  treat AI like another technology roll out. Please [30:52] don't treat it like another technology roll out.  It isn't right. I've been in tech for a long time. [30:57] This one's different. Right? Don't ask how AI fits  into your existing enterprise. Ask what kind of [31:04] enterprise becomes possible because of AI because  history won't remember the leaders who deployed [31:09] the most agents. I promise you they won't. It  will remember the leaders who had the courage to [31:15] redesign the enterprise itself. Those are going to  be the people that remembered. Okay. So, I want to [31:21] bring this to a close. Every generation inherits  institutions designed for yesterday's economics. [31:29] Yours may be the first to design the next one. I  think that's exciting, right? And I so I'd like [31:35] to I'd like to end where I began. In 1937, Ronald  Coast asked a deceptively simple question, right? [31:43] Why do firms exist? His answer shaped nearly a  century of management thinking. Enterprise existed [31:51] because markets were expensive. Coordination was  difficult. Accessing accessing expertise was hard. [31:57] Right? So we built organizations. We hired people.  We created departments. We established management [32:04] layers. Developed governance. We built enterprises  because they were the most efficient way to [32:09] coordinate scarce intelligence. And for more than  a hundred years, right, it worked brilliantly. [32:17] But every institution is built around the  economics of its time. Today, one of those [32:23] economics is changing not because intelligence has  disappeared, but because access to intelligence is [32:31] changing fundamentally. So, let me leave you  with one final question. When friction is no [32:38] longer the defining problem, what is your company  actually for? Right? The answer isn't efficiency. [32:45] It isn't headcount nor organizational charts.  The enterprise becomes something much greater, [32:52] right? It becomes a container for human  intelligence. A place where people solve [32:56] problems that matter. Where judgment matters more  than administration. Where creativity matters more [33:03] than coordination. A place where trust matters  more than process. Perhaps one day our children [33:11] will look back and wonder why brilliant engineers  spent their afternoons updating spreadsheets. [33:16] Why project managers spent their weeks chasing  status reports. Why lawyers spent thousands of [33:22] hours reviewing nearly identical contracts.  Why leaders devoted so much of their energy [33:27] to routing information instead of imagining the  future. Right? Perhaps bureaucracy will feel as [33:34] strange to them as paper ledgers feel to us.  Every generation inherits assumptions designed [33:41] for yesterday's economics. Right? Most generations  improve them. Very few generations are asked to [33:48] reinvent them. Right? I believe ours is. So if you  design transformation, redesign work. If you build [33:56] technology, redesign coordination. If you design  products, redesign judgment. If you build trust, [34:06] redesign confidence. And if you lead enterprises,  redesign the institution itself. Right? Don't [34:13] spend the next decade asking how do we deploy  AI. Spend it at asking what kind of enterprise [34:19] becomes possible because of AI. Right? Economics  teaches us that every time scarcity changes, [34:27] the world reorganizes itself. The only question  is whether you'll help design what comes next. [34:33] And I hope that you do. And I hope I've  inspired you, too. So, thank you. [applause] [34:44] Right. So, we have I think we have a few minutes [34:46] for questions if anybody has  any questions. Any questions? [34:57] Sorry. So, what are you doing differently at  AES now? Uh so at AES we are going through a [35:05] process where we sort of are saying sort of  internally sort of burn the walls down but [35:11] you know if if we didn't do anything today like  if we had none of the bureaucracy or anything [35:17] we still had the things we need to do build  energy assets you know all that kind of stuff [35:22] how would we do it today if we didn't have  anything else and so where does digital fit [35:27] in where does AI fit in where does you know  uh the human fit in and so sort of taking [35:32] that holistic approach to it. It's not an  easy thing to do. [laughter] Uh you know, [35:38] so I think we we started with boiling the ocean  and now we're very much sort of centering on [35:43] a few areas to sort of start there and then  sort of reimagine what it looks like. So yeah, [35:53] you you mentioned that redesigning the  organization. So when you can you elaborate [35:58] a little bit more when you're redesigning are  you thinking about redesigning the roles that [36:03] people are doing and what does it mean in  the AI era? Yeah. So I think it it's sort [36:08] of I would say sort of everything is on the  table right and so I again I think the idea [36:14] is not that you want to eliminate humans  right you don't you need them for judgment [36:18] all of that trying to understand where they  now actually best fit into that process and [36:25] so it very much should be a everything's on the  table approach but you know sort of recognizing [36:30] that we we don't want to eliminate the worker  right we want them to be there they're very [36:34] important to this you I do not believe human  work is going to go away as so [laughter] you [36:40] know but now design figuring out where they  are best placed and amplifying what they can [36:46] do and not hopefully having them do anything  bureaucratic anymore it's the goal [laughter] [36:54] hello coming from more of the product and  strategy space I had a question on you know [36:58] confidence judgment and trust how do you measure  those things and kind of give a quantifiable [37:05] reference for folks who are coming more from the  technical backgrounds. That's a I think that's a [37:09] great question. I think we're we're still sort  of figuring that out. So, if anybody has any [37:15] thoughts, I'd love to hear them. You know, I think  you can sort of say right now are are sort of [37:19] people coming back, you know, how quickly are they  engaging, but you know, like all metrics, I think [37:25] they'll change, but it's we're sort of in a a wild  west at the moment. [laughter] So, trying to sort [37:30] of figure out what the right right metrics are.  But it's a good it's a very good question. Yes. [37:38] Thanks for the nice presentation. I do have  a question when we uh look into the internal [37:43] organization redesign. I wanted to hear your  thoughts about the current IT department, [37:48] their roles in a large organization because AI  becomes so easily available for all the teams, [37:56] different you know groups and people doing their  own AI solutions anyway, right? So I would love [38:02] to hear your thoughts on how uh we can do a better  job to uh you know shifting the IT functionality [38:10] to support that. Yeah, I think you know from  my perspective I think it's I think that sort [38:15] of in in conjunction with the sort of compliance  and legal making things as easy and seamless as [38:21] possible for sort of what I would call the sort  of non-technical user for lack of a better term [38:27] right so how easy is it to actually now build  what they want to build have the tools that [38:32] they want to use you know I do think that's  where the sort of other side of it comes in [38:37] in the sense that if people know that they can  sort of do so safely what the guard rails are. [38:42] I I for I probably shouldn't say this for years.  I sort of been like governance I don't like you. [38:47] But um now I feel like [laughter] you know it's  it's really important but I you know it's not to [38:53] sort of get in the way but I think how do we  as as IT departments make things just as easy [39:00] and seamless to deploy productionalize uh share  across an organization is becoming more and more [39:07] important. Yeah. Do you have time for two more  questions? I I've got all days. Okay. [laughter] [39:16] Hi. Thank you very much. Very thoughtprovoking.  Um couple of one comment and a question. There's a [39:22] lot of nuggets of important information, but there  were no slides. So, I want to make sure that I [39:27] have a handle on what you shared so I can sit with  it. I if you if you drop me on LinkedIn, I'll I'll [39:33] I have the whole thing written. I'll share it with  you. Awesome. Thank you. [laughter] The question [39:37] is what's the time horizon you're actually looking  at to actualize these things. What I mean by that [39:43] like that there's different varied level of  maturity of all the things that need to come [39:47] together to truly have intelligence in abundance  in an organization and then move into more of that [39:54] focus on trust judgment and all of that and in  a large organization it's like the mothership is [40:00] so huge to turn. So like what is your thought on  the time horizon of these things coming together? [40:06] So it's it's a great question. If you had asked me  six months ago, I was gung-ho was two years then [40:13] we started and I thought this is never going to  happen because I'm as we're sort of 10,000 people [40:19] uh across the world. So you know I thought  oh we're never going to get there. I would [40:25] say now I hope that 80% within 3 to 5 years.  Um, I think it's, you know, getting to that [40:36] last bit is is like everything, the the sort of  long tail is going to be really difficult. But, [40:42] you know, I think as more as we're as  we're sort of engaging more and more, [40:46] people are getting more and more excited. I think  also as well, they're I think people are really [40:50] understanding that like we're not we don't want  to eliminate your job, right? we want to eliminate [40:55] the bureaucracy [laughter] and and so people now  are are very much sort of you know coming around [41:02] and excited and eager they want to sort of be  part of it. I would say the other the other sort [41:06] of just the biggest challenge is sometimes when  you're so in the weed of your process it's hard [41:11] to think how would I do this process if if if  we didn't have all this bureaucracy. So I think [41:18] that's kind of right now some of our our biggest  sort of um our biggest barriers. Thank you. [41:27] Sorry, last question that we can do in public  and then we will be lunch. Thank you so much. Um, [41:35] this was exactly what I was looking for in  this uh event because right now my day has [41:41] been essentially trying to reinvent what I do at  my company. So I uh I found it very difficult to [41:48] even fit myself in the role titles that are here  [laughter] because I'm not a product manager. I'm [41:54] not an engineer. I'm essentially a subject matter  expert building AI products which is I don't know [42:01] where what I am now. Um so curious just to to  see if you have any tips on how to explain to [42:10] people when you build when you are exactly in this  this change. Um what do you do and how do you how [42:17] do you can encapsulate like a a perspective  or a new a new sort of role that encompasses [42:25] multiple multi-disipline stuff. So, I'm gonna  say something completely unconventional and say, [42:34] don't try and define yourself. [laughter] I  know that's probably not what you want to hear, [42:39] but I think as somebody that that sits exactly  where you are, I'm a I, you know, I have my PhD [42:44] in economics. I'm an economist, but I I think um  what I would just coming from personal experience, [42:50] what I would say is that um sometimes when  you try and sort of define yourself too much, [42:57] people box you in. And and so I would say given  given how how much you have to offer and sort [43:06] of the sort of non nondefinition aspect of  your role, I would say own it and run with [43:13] it and really just sort of, you know, go  to people and almost just say, you know, [43:18] I'm here to sort of help you reimagine. And I  think without sort of saying what angle you're [43:25] coming at, you'll get people that will just  start talking. And you sound to me like you [43:29] have enough sort of various, you know, roles  within you that you can you can sort of figure [43:35] out which waiting of each one needs to come for  that person. So I know that's probably [laughter]