Axios, Daimler Leaders Pivot Enterprise AI to Workflow Redesign

Axios与戴姆勒领袖推动企业AI转向工作流重塑 企业必须引导员工从孤立的聊天机器人生产力转向全面的工作流程重组与共享基础设施，以在不裁员的前提下释放可衡量的运营收益。 与会嘉宾概述了组织与专业人员如何从初级AI试验过渡到全面的工作流程重塑。核心讨论聚焦于结合组织上下文赋能模型、自动化日常行政任务以释放人力投入战略举措，以及通过集中式治理避免内部聊天机器人无序蔓延。嘉宾们还提供了评估供应商软件架构的实用框架，并敦促专业人员采用持久上下文工作区并坚持每日动手实践，以保持竞争优势。 企业推进 AI 应用的核心在于将其与业务目标紧密挂钩，推动实际工作流程的转变，而非单纯追逐技术新颖性。 (0:04) 生成式 AI 正充当知识工作的认知陪练与类似电子表格的提效利器；Axios 通过鼓励基层实验，让非技术财务员工将月度结账流程缩短了两整天。 (3:11) 企业落地需平衡自下而上的基层探索与集中式架构治理，避免各自为政导致的聊天机器人泛滥与低效 PoC 堆叠。 (11:08) AI 模型存在上下文盲区，唯有注入企业组织架构图等核心背景数据，才能让会议纪要分析具备真正的业务实操价值。 (12:02) 在评估第三方 AI 供应商时，企业买家应看穿表面包装、穿透审查底层基座模型，并优先签订灵活的一年期模块化合同以规避风险。 (21:36)

Enterprise AI adoption succeeds as an organizational and behavioral transformation anchored to core business outcomes rather than technical novelty, requiring leaders to balance grassroots experimentation with centralized architecture while teams cultivate daily, hands-on fluency. Three-Stage Maturity and Workflow Redesign (14:06) Enterprise AI adoption progresses across three distinct evolutionary phases: routine task automation, departmental process augmentation, and end-to-end workflow redesign. Andy Dingler compares generative AI to the introduction of spreadsheets, noting that while employees can immediately automate repetitive Monday morning tasks, true organizational value emerges only when workflows are re-architected. Instead of merely accelerating discrete steps within legacy ten-step processes, mature implementations position AI at the center to collapse handoffs and execute direct transitions between start and end states. Grounding Models with Structured Organizational Context (12:02) Transitioning models from generic chat tools into high-utility workplace instruments requires eliminating context blindness. Paul Farnsworth advises moving beyond isolated prompts by structuring persistent workspaces loaded with dedicated context documents, such as internal organizational charts. Grounded in structured role hierarchies and reporting lines, models can analyze raw enterprise meeting transcripts and deliver role-aware action items rather than generic summaries. Farnsworth forecasts that automated meeting transcription will eventually become ubiquitous across enterprises, enabling continuous corporate memory tracking across quarters. Balancing Grassroots Innovation with Centralized Governance (6:44) Scaling AI requires promoting decentralized exploration while preventing operational fragmentation. Allison Murphy highlights a breakthrough at Axios where an accounts payable specialist automated two-thirds of invoicing operations, cutting two full days off the monthly financial close and subsequently stepping into an enterprise-wide AI training role. However, unchecked grassroots activity carries substantial operational risk; Dingler explains that independent departmental initiatives at Daimler Truck North America created "chatbot sprawl" and credibility issues. Daimler addressed this by cataloging proof-of-concept outcomes into a centralized memory bank that shares validated solutions across business units. Vendor Procurement Risks and Modular Architecture (21:36) Enterprise buyers must exercise strict technical due diligence when evaluating commercial AI vendors at industry trade shows. The panel warns buyers to look beyond superficial user interfaces—often styled with generic purplish-blue gradients—and verify whether software relies on proprietary models or repackaged wrappers of standard foundational systems. Architecturally, organizations should decouple systems by differentiating between native features embedded in core records (such as SAP or Salesforce) and cross-platform orchestration layers. Because startup compute subsidies mask long-term pricing and vendor mortality remains high, the panel recommends flexible one-year contracts over three-to-five-year lock-ins. Cultivating sustainable AI adoption ultimately depends on psychological safety and cultural transparency rather than workforce reductions. The panel argues that automated efficiency should redirect human domain experts toward long-term strategic initiatives, though leaders must continuously manage technical hurdles including model hallucinations, rapidly shifting tooling, and governance trade-offs around employee monitoring.

Dice的Paul Farnsworth、戴姆勒卡车北美公司的Andy Dingler以及Axios的Allison Murphy指出，企业与个人落地人工智能的关键不在于追逐底层模型更迭，而在于将日常探索锚定于业务目标、重塑组织工作流并建立持续学习与治理机制。 从零散任务自动化迈向全流程重塑 (14:06) 企业推进人工智能并非单纯的技术升级，而是深度的行为变革。企业对技术的采纳通常经历三个阶段：初期以自动化单点常规任务起步（如周一早上的重复行政工作）；中期实现流程增强，为多步骤业务链条中的特定环节提速；最终迈向端到端的工作流重构，将原本冗长的多节点流转简化为点对点的直接转化，消除不必要的中间交接环节。Dingler将生成式AI比作当年的电子表格软件，认为它在认知劳动领域扮演着智力陪练的角色。 技术转型的推进需要明确个人效率与团队变革管理的边界。部署团队级AI属于经典的技术变革管理范畴，必须具备结构化培训、明晰标准和可量化的业务指标。如果团队无法量化现有流程，盲目引入AI往往收效甚微。在个人使用层面，Farnsworth建议脱离单次问答的浅层交互，转向在持久的项目工作空间中注入上下文文档（如组织架构与业务偏好），以消除模型因缺乏组织背景而产生的盲区。 基层草根探索与集中式架构治理的平衡 (6:44) 赋能一线非技术人员能够激发自下而上的技术突破。Murphy分享了Axios的实践：一名应付账款专员通过自主利用AI实现约三分之二的标准财务报告与发票处理自动化，将公司月度财务结账周期缩短了两整天，该员工随后转型专职负责企业内部的AI培训。Axios在500人的员工规模中抽调了3名全职员工专门从事内部赋能。然而，去中心化的探索若缺乏统筹，极易陷入各自为政的困境。 Dingler指出，业务部门初期各自寻找外部供应商或内部团队搭建独立对话界面，导致出现大量追踪困难、准确性存疑且界面混乱的“聊天机器人蔓延”（chatbot sprawl）。为遏制这些低质AI产物，戴姆勒卡车北美公司着手将概念验证整合至集中式的企业记忆库中，记录各部门的成败经验，促使跨部门复用成熟方案。此外，团队在实践中也发现部分问题用常规工程脚本解决比大语言模型更有效，证明自主迭代需与理性技术选型相结合。 甄别供应商虚幻演示并防范合约风险 (21:36) 企业采购商业第三方AI软件时，面临着类似传统SaaS的销售包装陷阱。许多供应商展示出看似功能完备的界面，实则只是缺乏技术深度的外壳（wrapper），甚至连UI都高度雷同地采用千篇一律的紫蓝色渐变设计。买方必须要求供应商明确披露底层基础模型的来源，区分该方案究竟是深度嵌入在SAP、Salesforce或Workday等核心记录系统内部，还是跨系统的云原生编排平台。 面对快速洗牌的技术环境，采购决策需防范长期绑定的结构性风险。由于初创公司常依赖风投资金补贴算力成本，买方需严格审查未来的计费模型，防止算力补贴退潮后出现多达20倍的隐性费用激增。为了维持系统的模块化与可替换性，企业应优先考虑签订灵活的单年期合同，坚决避免承诺3至5年的长期合约，同时对承诺“未来3个月内交付全功能”的路线图通胀保持高度审慎。 建立容错文化与融入日常的工作习惯 (28:38) 破除AI必然导致大规模裁员的误解是建立组织安全感的前提。Murphy与Farnsworth认为，简单依靠裁员缩减成本的企业将落后于那些主动将专业人才转向3至5年期战略性业务的竞争对手。在经历过行业转型的媒体领域，管理层更应保持坦诚沟通。早期试图在本地硬件（如Mac mini）部署开源助手的尝试曾暴露出上下文处理不足与微调繁琐的问题，而如今代码生成技术已使工程重心从编写具体语法转向系统设计与自然语言交互。 为构建持续的个人应用习惯，Dingler建议从业者每天至少投入15分钟与模型对话，并将其作为汇报前的演练工具，通过赋予AI跨部门主管的人设来预演潜在质疑与业务盲区。Murphy提倡将语音输入作为默认交互模式，通过连续数分钟的口头倾诉将复杂的组织摩擦和零散构想完整传达给模型，进而辅助管理规划；Farnsworth则提醒，即使是熟练用户也必须警惕模型产生的逼真幻觉，保持批判性验证。 讨论总结强调，大语言模型仅是不具备情感的计算工具，全员探索的时代窗口为非技术人员提供了通过主动学习重塑职业路径的机遇。然而，随着部分企业开始监控员工Token消耗量以评估效率，如何在防范失败惩罚与推行制度化考核之间维系健康的创新土壤，仍是企业管理层需要审慎拿捏的平衡点。

## Background 该内容由频道 Ai4 于 2026 年 8 月 21 日发布在 YouTube 平台。视频记录了一场面向 AI 新手的圆桌对谈，由索诺马州立大学教授 Emily Acosta Lewis 主持，邀请了 Axios 首席运营官 Allison Murphy、Dice 总裁 Paul Farnsworth 以及戴姆勒卡车北美 AI 战略负责人 Andy Dingler，旨在为职场人士与管理者提供一份从零起步、融入业务并实现价值转化的全方位实操指南。 ## TL;DR 对谈围绕技术如何切实服务于业务目标展开，梳理了从个人零散尝试到端到端重塑工作流的演进路径，并提出了涵盖产品级、工作助手与自主智能体的三层认知模型。嘉宾们结合制造业与媒体行业的真实案例，深入剖析了警惕企业内部“聊天机器人泛滥”、规避软件供应商“路线图虚标”及算力计费陷阱等治理难题，同时分享了将财务结账周期缩短两天的落地成效。针对日常实操，专家们重点推介了维护个人长期“上下文文档”、利用大模型做高管画像推演以及借助语音模式倾倒复杂思路等具体技巧，呼吁企业以赋能员工转型战略顾问代替盲目裁员。 ## Atmosphere & style 现场交流氛围轻松坦诚且充满幽默感，三十余年资深 CTO 自嘲先祖是卢德运动抗议者、回忆因拔错插头导致银行网站瘫痪的往事，引得全场笑声连连，迅速化解了技术带来的严肃与恐慌感。三位来自不同领域的嘉宾配合默契，既有媒体高管贴便签警示“它不是真的”这一审慎态度，又有资深工程师“不再写一行代码”的激进颠覆，在宏观战略避坑与微观生活妙用（如规划家庭夏令营）之间自如切换。整场讨论散发着务实且充满包容的求知欲，传递出一种鼓励大众重拾“边玩边学”心态、在时代红利中坦然探索的温暖力量。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=4s) AI 落地最艰难的部分在于推动组织内部的行为变革，促使员工从单纯的个人提效转向利用 AI 深度重构业务工作流。 - [03:11](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=191s) AI 应用可清晰划分为三大维度：基础模型研发与嵌入的“产品中的 AI”、办公协作提效的“工作中的 AI”、以及具备自主规划执行能力的“智能体 AI”。 - [06:44](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=404s) 会议转录的真正价值在于长期积累形成企业级历史语料库（Corpus），从而实现跨会议的任务对比与自动化行动项追踪。 - [06:44](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=404s) 企业在早期应警惕“聊天机器人泛滥”（Chatbot sprawl）与信息孤岛，需建立统一记忆库与可复用的中台技术底座。 - [11:46](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=706s) 通过向 AI 项目中注入组织架构、沟通风格与角色职责等结构化“上下文文档”，可大幅跃升输出的分析深度与执行精准度。 - [21:36](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1296s) 采购第三方 AI 软件时需穿透表面演示，优先签署一年期短约，警惕路线图虚标与用量补贴期结束后的算力成本飙升风险。 - [21:36](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1296s) 企业技术选型应注重“即插即用”的云原生与模块化架构，避免与特定平台或供应商深度绑定。 - [25:06](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1506s) 资深使用者容易产生对 AI 的“假性信任”，在日常工作中需时刻警惕模型幻觉并坚持事实核查。 - [28:38](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1718s) 单纯因 AI 提高效率而裁员的企业缺乏长远竞争力，企业应赋能员工转型为高阶战略顾问并投身长期创新项目。 - [34:07](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2047s) 利用“角色扮演”提示词将 AI 作为战略推演伙伴，可在正式汇报前全方位模拟不同利益相关方的高管视角与潜在质疑。 - [35:54](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2154s) 采用连续语音输入倾倒背景与思考细节，能够显著突破打字输入的信息过滤损耗，将 AI 提升为组织与决策的共谋者。 - [39:28](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2368s) 非技术人员可以通过将目标岗位要求与简历输入 AI，反向定制攻克盲区的学习路径，实现快速职业跃升与跨界转型。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=4s) 现场调研与嘉宾亮相：AI 旅程中的核心认知 讨论由主持人对现场观众的技术背景调研拉开帷幕。主持人依次询问现场有多少人是刚接触 AI、正在学习基础知识的“新手”（Newcomer），有多少人是偶尔在某些任务中尝试使用 AI 的“实验者”（Experimenter），以及有多少人是已将 AI 深度融入个人或工作流程的“实践者”（Implementer）。随后，三位嘉宾分别用简短的介绍分享了各自在 AI 探索中总结出的一条基本事实。 ```mermaid flowchart TD accTitle: AI 旅程的三大阶段 subgraph Journey["AI 探索与实践阶梯"] A["新手 Newcomers<br/>学习基础与概念"] --> B["实验者 Experimenters<br/>在零散任务中尝试工具"] B --> C["实践者 Implementers<br/>融入业务流程与工作流"] C --> D["持续进化 Continuous Learning<br/>跟进技术迭代与组织变革"] end ``` Dice 总裁 Paul Farnsworth 首先发言。他幽默地提到自己担任首席技术官（CTO）已有 35 年之久，并调侃大家“别把这事说得太大声”。见证了数十年技术浪潮更迭的 Paul 指出，技术的本质始终是变革，拥抱 AI 需要保持“边玩边学”（playing to learn）的心态并勇于承担一定风险；但他特别强调，在企业环境中，一切必须以“业务目标”为中心，脱离业务目标而单纯聚焦技术本身，很容易让人误入歧途。 戴姆勒卡车北美公司（Daimler Truck North America）的 AI 战略负责人 Andy Dingler 介绍，他的核心职责是将业务构想转化为全公司范围内可复用的 AI 技术资产。他提出的核心认知在于：AI 落地最艰难的部分其实是“推动组织内部的行为变革”（behavioral change），让员工不仅将 AI 视为提高个人生产力的日常助手，更能利用它来深度重构业务工作流。Axios 首席运营官 Allison Murphy 则指出，AI 是一张不断延展的学习路线图。即便是已经进入落地阶段的团队，也绝没有到达终点，因为“下周技术又会发生新的变化”。面对如此迅猛的发展节奏，企业不必感到畏惧，关键在于主动参与并带领组织共同学习。 - **Paul Farnsworth（Dice 总裁）**：拥有 35 年 CTO 经验，强调技术要为业务目标服务，警惕脱离商业本质的技术盲从。 - **Andy Dingler（戴姆勒卡车北美 AI 战略负责人）**：负责半挂牵引车制造巨头的 AI 战略落地，核心攻坚点是推动组织从个人提效走向工作流协同的行为变革。 - **Allison Murphy（Axios 首席运营官）**：提出 AI 发展不存在真正的“终点线”，企业需要具备持续迭代的认知与组织协同能力。 --- ### [03:11](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=191s) 降维解析 AI：面向董事会与跨部门团队的三大分类框架 针对许多人被机器学习等复杂概念所困扰的现状，嘉宾们探讨了如何用最通俗易懂的语言向董事会和跨部门团队解释人工智能。Paul Farnsworth 提出了一个极其实用的三层认知模型：第一类是“产品中的 AI”（AI in a Product），涵盖基础大模型（如 Claude）的开发以及产品研发团队在应用中嵌入底层能力；第二类是“工作中的 AI”（AI at Work），即作为个人日常办公的提效助手。Paul 分享了自己使用 Excel 配合 Copilot 的真实体验，当他拖拽公式时，AI 能主动理解意图并自动补全，这正是人机协同的典型写照；第三类则是广受媒体关注的“智能体 AI”（Agentic AI），即由代理程序自主执行原本由人类承担的角色或复杂任务序列。 ```mermaid flowchart LR accTitle: AI 应用的三大核心维度 subgraph Framework["AI 应用分类模型"] direction TB P["产品中的 AI<br/>AI in a Product"] --- P_desc["基础大模型研发<br/>嵌入软件核心功能"] W["工作中的 AI<br/>AI at Work"] --- W_desc["认知副驾驶/助手<br/>如 Copilot 协同办公"] A["智能体 AI<br/>Agentic AI"] --- A_desc["自主规划与执行<br/>代理人类角色与任务"] end ``` Andy Dingler 对此深表赞同，并将 AI 的演进与电子表格的历史进行了类比。他指出，过去 Excel 的出现让人们从繁琐的计算与工程建模中解放出来，而今天的 AI 则是在“认知学习”层面扮演人类副驾驶的角色。Andy 在公司内部始终倡导“AI 优先”（AI-first）的理念，建议员工把 AI 当作 strategic leadership 演练中的“陪练伙伴”（sparring partner）。无论是梳理周一早晨最耗时的工作，还是利用 AI 自动化那些“自己最讨厌处理的任务”，都是驱动员工迈出第一步的极佳切入点。 Allison Murphy 从高管沟通与公司治理的维度进行了补充。她强调，在向董事会或领导层汇报时，完全不需要让每个人都成为深谙最新模型参数的技术专家，过度追逐技术细枝末节反而会分散核心注意力。企业真正需要的是让管理层和员工理解技术的整体能力边界，确保技术应用紧密围绕公司的战略方向展开，并具备足够的信息素养以在业务中做出明智的工具选择。 - **三层 AI 划分框架**： - **AI in a Product**：基础模型开发与软件内部深度集成的核心算法能力。 - **AI at Work**：办公协作与辅助提效（如 Office Copilot、知识总结等）。 - **Agentic AI**：能够自主规划、拆解步骤并代劳复杂任务的智能体系统。 - **高管认知原则**：警惕最新模型带来的信息过载，将精力聚焦于技术边界理解、战略对齐与智能化决策制定。 --- ### [06:44](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=404s) 从会议转录到财务自动化：企业落地的实践案例与避坑指南 在谈及具体的 AI 落地成果与经验教训时，Paul Farnsworth 提出了一个看似基础却威力巨大的切入点——会议工作流改造。他坦言自己现在几乎不再亲笔撰写会议议程。很多企业虽然开通了会议转录功能，但真正的秘诀并不在于处理单场会议，而在于“长期积累这些转录数据形成语料库（Corpus）”。通过大模型比对跨会议的历史记录，团队能够快速复盘两次会议之间的任务推进情况，带来巨大的时间节省。尽管部分企业初期会对记录未经加工的会议原话心存顾虑，但 Paul 认为全量记录与智能化提取行动项将成为不可逆转的企业趋势。 Andy Dingler 进一步拓展了这一思路。戴姆勒卡车北美正在构建基于“记忆机制”（Memories）的企业级用例库，以增强对历史经验的调取能力。Andy 直言，在生成式 AI 刚兴起时，企业内部出现了严重的“聊天机器人泛滥”（Chatbot sprawl）现象——几乎每个部门都各自聘请顾问或自行搭建聊天机器人，导致追溯困难、准确性堪忧、数据可信度下降，员工甚至不知道该去哪里提问。为了避免产生“内部 AI 垃圾”（Internal AI slop），戴姆勒正在整合所有功能模块，打造一个能够跨部门复用技术底座、基于统一记忆库调度的企业级门户。 ```mermaid flowchart TD accTitle: 聊天机器人治理与统一中台演进 subgraph Problem["早期乱象：Chatbot Sprawl"] A1["部门 A 独立采购/搭建"] A2["部门 B 聘请外部顾问"] A3["部门 C 零散自建工具"] A1 & A2 & A3 --> Err["问题：信息孤岛、数据失真、AI 垃圾 (AI slop)"] end subgraph Solution["演进方案：企业级复用底座"] U["统一前端接口 Unified Interface"] --> Engine["共享记忆库与中台底座"] Engine --> App1["跨部门复用用例 1"] Engine --> App2["跨部门复用用例 2"] end Problem --> Solution ``` Allison Murphy 则分享了 Axios 内部“百花齐放与规范治理并重”的转型故事。Axios 在早期鼓励全员探索 AI 工具，结果令人惊喜地在非技术部门发掘出领军人才。公司财务部负责应付账款（Accounts Payable）的一位女员工对 AI 产生了极大热情，凭一己之力将财务团队约三分之二的标准化报表与发票处理流程实现了自动化，直接使公司每月的财务结账周期（Monthly finance close）缩短了整整两天。这位员工如今已走出财务岗位，转型为负责全公司内部 AI 培训的领航者。Allison 总结道，这种赋能让不同部门在各自的周期中快速试错与迭代，财务团队在跑得足够快之后，甚至能主动识别出哪些流程“其实不应该使用 AI”，从而实现了技术认知的成熟蜕变。 - **会议语料库化**：超越单场会议总结，构建跨时间维度的会议数据语料库，实现自动化行动项追踪与执行对比。 - **警惕“聊天机器人泛滥”（Chatbot sprawl）**：避免部门各自为政导致的孤岛与数据失真，转向统一接口与可复用基础底座。 - **Axios 财务自动化实战成果**： - 自动化了财务部门近 2/3 的常规报表与发票流转。 - 每月财务结账时间直接缩短 **2 天**。 - 基层业务人员成长为全员 AI 培训带头人，完成职业轨迹升级。 ### [11:08](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=668s) 平衡基层创新与集中治理 在讨论内部工具开发时，嘉宾指出许多任务并不一定非得动用复杂的 AI，有时让工程团队编写一个快速脚本就能解决问题。部分团队之所以能够迭代四到五次工具，正是因为组织为他们提供了探索的空间与挑战。然而，企业必须高度重视治理问题。基层创新的成功往往建立在单一业务团队对自身工作流程了如指掌的基础之上；如果缺乏全局把控，企业很容易陷入“产生了一千个表现平平的聊天机器人”的混乱局面。最有效的做法是在鼓励一线员工自发创新的同时，保持中央层面的战略审视与治理平衡。 --- ### [11:46](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=706s) 业务领导者寻找低风险切入点与上下文管理 当企业高管决定引入 AI 却不知从何下手时，首要任务是明确区分使用场景：是个人用来提升自身效率，还是要求整个团队进行变革。如果是推动团队采用，这本质上与任何技术变革管理项目没有区别，核心依然是建立培训体系、明确标准并锚定业务目标——“如果某项指标可以被衡量且能够被改进，那么你就走在正确的轨道上”。 ```mermaid flowchart TD accTitle: AI 组织落地演化路径 subgraph S1["阶段一：个人自动化"] A["个人任务自动化"] --> B["消除日常琐碎重复劳动"] end subgraph S2["阶段二：流程增强"] C["团队流程增强"] --> D["将 AI 嵌入现有10步业务流"] end subgraph S3["阶段三：流程重塑"] E["工作流端到端重塑"] --> F["AI 驱动整体流程，缩减多余环节"] end S1 --> S2 S2 --> S3 ``` 针对个人日常使用的切入点，嘉宾分享了一个构建深度上下文的实用技巧： - **组织架构上下文注入**：单次对话往往缺乏背景信息。高管可以从公司内网下载包含整个组织架构的 PDF 文件并上传至项目库，让 AI 提取每个员工的姓名、职位、角色以及汇报汇报线。 - **会议纪要质量跃升**：在处理会议转录文本时，AI 不再只是输出“Kevin 说了某句话”，而是能结合 Kevin 的具体职责与业务范围进行深度分析和上下文总结。 - **推动行为与认知转变**：AI 本质上是一个需要被正确引导和设定角色的白板。在业务团队中，许多非技术人员通过与 AI 的持续互动，逐渐理解了底层逻辑，进而成为懂业务又懂技术的复合型人才。团队应用应遵循渐进路径：先从周一自动化单项个人任务做起，再过渡到团队现存流程的局部增强（例如为原本包含 10 个步骤的流程赋能），最终发展为彻底重塑工作流，让 AI 直接主导端到端运转并裁撤冗余步骤。 --- ### [15:46](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=946s) 从高层战略看 AI：COO 如何用 AI 规划职业转型 除了自下而上的流程优化，业务领导者还可以从宏观视野出发。Axios 首席运营官 Allison Murphy 分享了高管团队在大约一年前做过的一项实验：在 ChatGPT 或 Claude 中输入自己的职位并提问：“我是公司的首席运营官，因为 AI 的发展，我的工作在未来两年和五年内会发生什么变化？请告诉我届时我的职责应该是什么样。” 模型给出的反馈为个人技能演进指明了“北极星”方向： - **管理职能的转变**：高管将不再需要耗费精力主持路线图优先级排序会议，年度规划周期将被大幅压缩，传统滞后的回顾性报告也将全面转变为实时数据分析。 - **成为系统架构师（Systems Architect）**：COO 的核心价值不再局限于维持现状运转，而是转型为连接业务运作、先进技术与自动化流程的系统架构师。 - **普适性的职业规划法**：任何层级的员工都可以使用这一方法，让 AI 充当头脑风暴与职业发展的思考伙伴，从而减轻面对新技术时的焦虑感，清晰规划个人技能学习路线。 --- ### [17:57](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1077s) 追踪 AI 前沿的实用日常习惯 面对日新月异的技术术语与动态，嘉宾们分享了非技术业务人士保持敏锐度的日常习惯： - **精选播客与权威简报**：无需深陷底层技术细节，每隔几周收听高水准行业播客即可把握大势。Allison Murphy 推荐了 Axios 旗下的每日通讯《AI+》（聚焦前沿实验室、商业与科技动态）以及播客《The AI Daily Briefing》（每日新闻速递配合深度解析，并附带免费技能实操培训链接）。 - **自动化定时追踪与目标对齐**：通过设定自动化定时 Prompt，让 AI 根据个人或业务的核心目标，按天或按周自动执行深度研究，生成包含具体操作步骤（第 1 至 4 步）的行动路线图，充当高度定制化的动态研究伙伴，而无需为层出不穷的基础模型更新耗费心力。 - **团队内部的趣闻轶事**：团队中有成员编写了一个 Agent，每天早晨自动汇总前一天的所有互动记录并生成一段语音备忘录。Paul Farnsworth 幽默地透露，为了“戏弄”这个 Agent，他特意开始用莎士比亚风格的古英语给该员工发消息，引得全场大笑。 --- ### [21:00](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1260s) 评估第三方 AI 工具与警惕供应商陷阱 主持人向小组抛出采购决策问题：对于更倾向于采购第三方 AI 工具而非自研的非技术买家，在与 AI 软件供应商洽谈时应警惕哪些危险信号（Red Flags）？ 嘉宾指出，AI 软件采购在本质上与传统的 SaaS 软件采购并无二致。许多供应商惯于展示炫酷的演示 Demo 并许下宏大承诺，但在实际功能深度上却十分欠缺。买家在深入评估时，必须穿透表面演示，直接探究其背后的底层基础模型（Foundational Models）架构与真实交付能力。 ### [21:36](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1296s) 供应商评估与技术架构的“即插即用” 在探讨如何评估市场上的 AI 供应商时，专家们强调了深入技术底层的必要性。Paul Farnsworth 指出，企业必须穿透供应商展示的表面 Demo，探究其底层到底依赖自研模型还是第三方基础大模型（Foundation Models），并建议尽可能与已经落地的客户交流真实体验。Andy Dingler 随后分享了一个行业内有趣的视觉规律：许多仓促赶工的 AI 工具，应用界面全都是那种“AI 生成感十足的紫蓝色调”，连外观都没花心思打磨。对于 Daimler Truck 而言，技术评估的核心不仅在于能否带来业务价值，更在于避免平台锁定（Vendor Lock-in）以及与现有核心源系统（如 SAP、Salesforce、Workday）的集成方式——是作为嵌入式插件运作，还是作为云原生（Cloud-native）平台在底层跨系统进行统一调度。在 SaaS 快速迭代的“高速列车”上，保持架构的“即插即用（Plug-and-play）”与快速替换能力至关重要。 ```mermaid flowchart TD accTitle: AI Vendor Evaluation and Architecture Selection subgraph Evaluation["供应商评估核心维度"] M["底层模型架构<br/>(自研 vs 基础大模型)"] D["实际落地案例<br/>(穿透 Demo 调研)"] V["业务真实价值<br/>(非浮夸 UI 包装)"] end subgraph Architecture["集成与技术架构"] direction TB A1["嵌入式方案 (AI-Embedded)<br/>依附于 SAP / Salesforce / Workday"] A2["云原生方案 (Cloud-Native)<br/>跨源系统统一编排调度"] F["即插即用 (Plug-and-Play)<br/>具备随时剥离与切换能力"] A1 --> F A2 --> F end Evaluation --> Architecture ``` - **架构选型关键**：明确系统属于在现有软件（SAP、Salesforce 等）之上的外挂嵌入式方案，还是能跨系统调度的云原生架构。 - **避免锁定**：必须保障技术具有高适配性，确保在供应商掉队时能够迅速拔除并无缝切换到新技术。 --- ### [23:18](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1398s) 合约策略、路线图虚标与团队实战考察 Allison Murphy 深入剖析了采购 AI 工具时的商业防线。由于行业剧烈动荡，许多初创公司可能在 6 到 18 个月内便不复存在，因此 Axios 尽可能“只签一年期合同”。面对销售承诺，她给出了务实的忠告：“任何承诺能解决一切问题的公司，你都要保持警惕。”当前行业普遍存在严重的“路线图虚标（Roadmap Inflation）”现象，供应商口中“3 个月内上线”的 AI 功能往往遥遥无期。此外，计费模式与算力成本是另一大潜在地雷：许多初创公司依靠融资补贴掩盖了真实的算力成本，一旦企业产生技术依赖，后续调用费用可能会暴涨 20 倍。最后，Allison 建议将繁杂的厂商推介转化为内部员工的实战练兵场——挑选有兴趣的员工去对接四家展商，横向比对产品差异与真实落地水平，既能考察市场，又能激发团队的技术敏锐度。 - **合同周期控制**：优先签署 1 年期短约，严防厂商倒闭或技术落伍风险。 - **警惕算力暴涨**：高度关注用量计费条款（Usage Billing），防范补贴期结束后成本飙升 20 倍的困境。 - **路线图虚标陷阱**：对声称数月内交付全套 AI 能力的宣传保持清醒，不为未落地的 PPT 买单。 --- ### [25:06](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1506s) 早期认知的打破：从编程解脱到警惕“假性信任” 在谈及涉足 AI 过程中被彻底推翻的早期假设时，三位嘉宾给出了截然不同却极具启发的亲身经历： Paul Farnsworth 曾以为开源框架能一劳永逸解决个人生活的所有琐事，他特意买了一台 Mac Mini，部署了本地开源技术栈，试图通过手机无缝交互，结果发现 AI 经常在极度基础的逻辑上碰壁，想要真正好用需要耗费大量的上下文调优。 拥有 15 年以上机器学习和数据科学经验的 Andy Dingler 则经历了颠覆性的职业转变。他原本认为编码是不可或缺的核心技能，但从去年底到今年初，随着大模型代码生成能力的飞跃，他“从那以后再也没写过一行代码”。他将现在的开发视为一门“艺术”——从敲代码的工匠退后一步，转变为顶层设计者和构思者，用创意思维引导 AI 构建产品。 Allison Murphy 则推翻了两个认知：第一，AI 绝不可能仅靠阅读行业资讯来掌握，“除非你亲自上手使用，否则你永远无法理解它的细微差别与潜力”；第二，越是精通 AI 的资深用户，越容易产生“我绝不会被 AI 欺骗”的盲目自信。当 AI 越来越懂你的团队、会议甚至上下文时，用户往往会放松警惕。为此，她在自己的电脑屏幕上贴了一张便签，上面写着“记住，它不是真的（Remember it's not real）”，时刻提醒自己防范幻觉（Hallucinations）。随后，现场主持人也附和了一个生动例子：在威尼斯人酒店让 AI 推荐鸡尾酒吧，AI 给出的推荐在追问下才承认“那家店其实早就关门了”。 --- ### [28:38](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1718s) 组织文化、消除裁员恐惧与人才战略转型 面对初学者和普通员工对“AI 抢饭碗”的普遍恐慌，嘉宾们探讨了如何建立健康的组织文化与转型机制。Paul Farnsworth 坦言，变革过程中保留适度的危机感是不可避免的，关键在于企业的考核导向：如果管理层鼓励尝试并包容失败，员工就会主动探索；如果对失败施加惩罚，团队就会滑向保守。 Andy Dingler 直言不讳地指出：“那些单纯因为 AI 而选择裁员的公司，最终都会被时代抛弃。”他认为明智的企业应该解放生产力：让那些接受过高等教育的员工从日复一日的重复性事务中脱身，转型为所在领域的战略顾问，将精力投向 3 到 5 年后能为企业建立核心竞争壁垒的“大赌注（Big Bets）”方向。 ```mermaid flowchart LR accTitle: Workforce AI Transformation Strategy subgraph Legacy["传统模式 (易被淘汰)"] direction TB L1["员工从事日常重复性劳动"] --> L2["引入 AI 进行直接裁员"] L2 --> L3["丧失长期创新与竞争优势"] end subgraph Strategic["战略转型模式 (推荐路径)"] direction TB S1["AI 接管重复性底层工作"] --> S2["员工转型为领域战略顾问"] S2 --> S3["专注 3-5 年长远创新与战略项目"] end ``` Allison Murphy 从深陷技术冲击的新闻媒体业（Axios）视角分享了应对之道。面对行业因技术演进持续萎缩的历史阴影，Axios 采取了极致透明的沟通策略——开诚布公地承认行业正被 AI 颠覆，但管理层承诺竭尽全力提供资源为员工赋能。Axios 甚至直接从财务等业务部门抽调了 3 名优秀员工转型为全职 AI 培训师（在 500 人的公司规模中占比可观），用真金白银的投入帮助团队看清职业发展路径，让员工在安全感中拥抱技术演进。 --- ### [31:53](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1913s) 将 AI 融入日常工作的实操建议与寻根趣事 在回答观众关于“离开会场后如何迈出融入日常工作的第一步”时，Paul Farnsworth 没有直接抛出教条式的答案，而是讲起了自己父亲退休后的趣事作为铺垫。他提到英国人退休后似乎都有个共同爱好——研究家族谱系（Family Tree）。他的父亲在追溯家族历史的过程中，顺着线索一路查到了一位名叫 Brandreth 的先祖，从而引出了一段将复杂历史信息与现代检索工具相结合的探索历程…… ### [32:19](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1939s) 历史的回响与“上下文文档”的建立 在谈到人们对新技术的恐惧时，保罗·法恩斯沃思（Paul Farnsworth）分享了一段家族趣事：他的家族曾经追溯族谱，却发现某一分支无法向前查找，后来才发现原来是家族改过姓氏拼写——因为原姓“布兰德雷思”（Brandreth）的祖先是英国卢德运动（Luddite movement）中抗议纺织机并最终被处决的最后一批人之一。保罗借此感叹道：“这种对技术的恐惧从来都不是什么新鲜事。” 随后，保罗分享了一个极具实操性的 AI 协作技巧：维护一份“上下文文档”（Context Document）。这一灵感源于他过去的一位上司，那位上司在共事之初给了他一份文档，详细列出自己的沟通偏好、业余爱好、管理互动方式以及每周报告要求。如今保罗也沿用了这种做法，并进一步将其移植到 AI 工具中。他建议用户不要只停留在零散的聊天对话框中，而是应该在工具里创建一个专属“项目”（Project），并将一份不断迭代的个人上下文文档植入其中。文档中可明确说明你的行文风格偏好、任务复核要求与思维习惯。保罗还透露了一个让安迪·丁格勒（Andy Dingler）直呼“可怕”的细节——他甚至为全家人也建立了一份上下文文档，记录孩子们的名字、妻子的偏好以及旅行住宿喜好，以便在制定家庭计划时 AI 能够无缝调用这些背景信息。 ```mermaid flowchart TD accTitle: AI 项目化上下文工作流 subgraph ContextSetup["构建长期上下文 (Context Document)"] A["个人/团队工作风格与偏好"] B["任务输出格式与复核标准"] C["业务/生活专属背景信息"] end subgraph AIWorkflow["AI 项目化落地"] D["在 AI 工具中建立专属 Project"] E["注入并动态更新 Context Document"] F["开展具体任务对话与策略推演"] end ContextSetup --> D D --> E E --> F ``` ### [34:07](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2047s) 将 AI 作为战略推演伙伴与高管画像模拟 安迪·丁格勒（Andy Dingler）建议初学者每天至少花 15 分钟与 AI 展开对话，借此熟悉其能力边界。随着使用深入，他逐渐将 AI 转变为自己的“战略合伙人”。例如，在需要分别向工程部门（Engineering）和售后市场部门（Aftermarket）的高管做汇报时，他会利用 AI 进行角色扮演（Persona Engineering）与推演：“假设你是售后市场的高级领导层，我正在主持这项议题，请帮我预判他们可能会提出哪些疑问？”通过这种提示词工程，他能在进入真实会议前，站在不同业务条线领导者的视角全面掌握对方关心的痛点与沟通方式。 艾莉森·墨菲（Allison Murphy）则在此基础上建议“更进一步，直接将 AI 设为你的默认选项（Make it your default）”。艾莉森起初也怀疑 AI 的价值，因为她作为首席运营官，日常 90% 的工作都涉及复杂的人际动态和组织协调。然而，当她尝试将一次组织架构重组（Org Redesign）的难题交给 AI 时，输出的结果令她大为震惊。她提醒大家不要把 AI 仅仅当成一个执行具体小任务的下属，而是要跳到更高维度，先问一句“这件事情你会怎么做”，把所有难题第一时间抛给它。 ### [35:54](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2154s) 语音输入的高效倾倒与家庭场景起步 艾莉森特别强调了语音模式（Voice Mode）对于深度思考与沟通的颠覆性价值。如果还在费劲地打字，就会错失巨大的效率红利。在处理前述组织重组项目时，她直接对着 AI 开启了 10 分钟的连续倾述，把要解决的所有问题、难点、阻碍以及所有灵感毫无保留地讲了出来。她指出，口头表达能传递出文字过滤后所缺失的大量背景和真实语境，“不仅要把 AI 当作助理，甚至可以把它当作指导你该做什么的‘管理者’”。 对于那些在职场中对 AI 望而却步，或者受限于企业数据隐私政策的员工，艾莉森建议不妨先从家庭生活切入： - **日常膳食规划**：生成定制化的一周食谱与采购清单。 - **家庭财务预算**：辅助梳理开销结构与理财计划。 - **子女夏令营日程协调**：将复杂的暑期安排交给 AI 进行冲突排查。 通过在安全的生活场景中建立熟悉感与掌控力，员工能更自然地将这些经验反哺到职场工作中。 ### [36:58](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2218s) 结语与核心寄语：享受学习者的时代红利 在座谈会的最后结语环节，主持人艾米莉·阿科斯塔·刘易斯（Emily Acosta Lewis）邀请各位嘉宾留下最想让听众记住的一条建议： - **保罗·法恩斯沃思**：“它只是一台没有感情的计算机。”用户完全可以随心所欲地提问，即使觉得不妥也可以随时删除对话记录，最重要的是找回“玩耍”（Play）的心态，享受探索的乐趣。 - **安迪·丁格勒**：AI 赋予了每个人前所未有的超级大脑，他现在能在一天内从零搭建并上线完整的 Web 应用程序来快速验证想法。他提醒道，当前技术的迭代周期正在从过去的“每 6 个月翻倍”加速至“每 3 个月翻倍”，必须将 AI 融入贯穿全天的思维流程中。 - **艾莉森·墨菲**：呼吁大家“享受成为一名学习者（Enjoy being a learner）”。在职业生涯中，人们极少能在不被既有知识和技能评判的情况下与全公司、全团队乃至全家人一起站在同一起跑线上共同学习新事物，这种时刻弥足珍贵。 - **艾米莉·阿科斯塔·刘易斯**：从传播学角度强调“输入垃圾，输出必然是垃圾（Garbage in, garbage out）”，获得糟糕的回答往往是因为给出的上下文和提示词不到位。同时，企业内部需保持透明沟通。艾米莉提到，目前已有部分企业开始监控员工的 Token 使用量，甚至将不使用 Token 的员工视为“效率低下”而列入裁员观察名单。 ### [39:28](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2368s) 容错的勇气与 AI 驱动的职业跨越 保罗回忆起早期在银行工作时的趣事：当时他们为了给刚从印度带回来的泰姬陵发光纪念品插上彩灯电源，拔掉了作为前沿展示的银行网络摄像头插头，结果导致整个银行网站瘫痪，因为摄像头的代码被直接硬编码嵌入在网站中。他笑着说道：“我们花了几分钟才搞清楚发生了什么，虽然道了歉，但我现在依然活得好好的。年轻人在职业生涯中现在正是实验的黄金期，学习的代价是美好的，千万不要害怕变革。” 安迪与艾莉森最后补充指出，AI 正在重塑职业上升路径。安迪分享了一位非技术背景的项目经理（Program Manager），她直接将心仪职位的 Job Description（岗位描述）和自己的简历丢给 AI，让 AI 为她量身定制了一套攻克技术盲区的学习路径图（Roadmap），最终仅用了两个月时间就成功追平技术人员的认知差距。艾莉森对此高度认同，并强调现代企业在招聘时最渴求的正是这种能够洞察前沿方向、主动利用工具并展现出极强敏捷度与灵活性的自驱型人才。整场对谈在热烈的掌声中圆满落幕。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=4s) 技术的本质始终是变革，拥抱 AI 需要保持“边玩边学”（playing to learn）的心态并勇于承担一定风险；但在企业环境中，一切必须以“业务目标”为中心。 — Paul Farnsworth > [00:04](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=4s) AI 落地最艰难的部分其实是“推动组织内部的行为变革”（behavioral change），让员工不仅将 AI 视为提高个人生产力的日常助手，更能利用它来深度重构业务工作流。 — Andy Dingler > [11:46](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=706s) 如果某项指标可以被衡量且能够被改进，那么你就走在正确的轨道上。 — Andy Dingler > [23:18](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1398s) 任何承诺能解决一切问题的公司，你都要保持警惕。 — Allison Murphy > [25:06](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1506s) 除非你亲自上手使用，否则你永远无法理解它的细微差别与潜力。 — Allison Murphy > [25:06](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1506s) 记住，它不是真的（Remember it's not real）。 — Allison Murphy > [28:38](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1718s) 那些单纯因为 AI 而选择裁员的公司，最终都会被时代抛弃。 — Andy Dingler > [32:19](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=1939s) 这种对技术的恐惧从来都不是什么新鲜事。 — Paul Farnsworth > [35:54](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2154s) 不仅要把 AI 当作助理，甚至可以把它当作指导你该做什么的‘管理者’。 — Allison Murphy > [36:58](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2218s) 它只是一台没有感情的计算机。 — Paul Farnsworth > [36:58](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2218s) 享受成为一名学习者（Enjoy being a learner）。 — Allison Murphy > [36:58](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2218s) 输入垃圾，输出必然是垃圾（Garbage in, garbage out）。 — Emily Acosta Lewis > [39:28](https://www.youtube.com/watch?v=r1Z4yXY8eXs&t=2368s) 年轻人在职业生涯中现在正是实验的黄金期，学习的代价是美好的，千万不要害怕变革。 — Paul Farnsworth ## Mentioned Emily Acosta Lewis, Allison Murphy, Paul Farnsworth, Andy Dingler, Ai4, Sonoma State University, Axios, Dice, Daimler Truck North America, Claude, Excel, Copilot, ChatGPT, AI+, The AI Daily Briefing, SAP, Salesforce, Workday, Mac Mini ## Video description Emily Acosta Lewis - Professor of Communication & Media Studies, Sonoma State University Allison Murphy - Chief Operating Officer, Axios Paul Farnsworth - President, Dice Andy Dingler - Head of AI Strategy, Daimler Truck North America This panel is designed to provide newcomers to AI with a comprehensive playbook for getting started on their AI journey. Panelists will break down the foundational concepts of AI, explain common use cases, and offer practical advice on how to navigate the rapidly evolving AI landscape. Learn how to identify the right opportunities to integrate AI into your business, understand the challenges and risks, and discover the critical steps needed to derive value from AI. Whether you’re an executive, entrepreneur, or professional, this discussion will equip you with the essential knowledge and confidence to begin your AI journey. [Source](https://www.youtube.com/watch?v=r1Z4yXY8eXs)

[0:04] Paul Farnsworth: 技术的本质始终是变革，拥抱 AI 需要保持“边玩边学”（playing to learn）的心态并勇于承担一定风险；但在企业环境中，一切必须以“业务目标”为中心。 [0:04] Andy Dingler: AI 落地最艰难的部分其实是“推动组织内部的行为变革”（behavioral change），让员工不仅将 AI 视为提高个人生产力的日常助手，更能利用它来深度重构业务工作流。 [11:46] Andy Dingler: 如果某项指标可以被衡量且能够被改进，那么你就走在正确的轨道上。 [23:18] Allison Murphy: 任何承诺能解决一切问题的公司，你都要保持警惕。 [25:06] Allison Murphy: 除非你亲自上手使用，否则你永远无法理解它的细微差别与潜力。 [25:06] Allison Murphy: 记住，它不是真的（Remember it's not real）。 [28:38] Andy Dingler: 那些单纯因为 AI 而选择裁员的公司，最终都会被时代抛弃。 [32:19] Paul Farnsworth: 这种对技术的恐惧从来都不是什么新鲜事。 [35:54] Allison Murphy: 不仅要把 AI 当作助理，甚至可以把它当作指导你该做什么的‘管理者’。 [36:58] Paul Farnsworth: 它只是一台没有感情的计算机。 [36:58] Allison Murphy: 享受成为一名学习者（Enjoy being a learner）。 [36:58] Emily Acosta Lewis: 输入垃圾，输出必然是垃圾（Garbage in, garbage out）。 [39:28] Paul Farnsworth: 年轻人在职业生涯中现在正是实验的黄金期，学习的代价是美好的，千万不要害怕变革。

[0:04] Paul Farnsworth: play to learn [3:11] Andy Dingler: the task that you hate doing [6:44] Allison Murphy: letting a thousand flowers bloom [11:08] a thousand different chat bots all doing a mediocre job [15:42] Allison Murphy: I'm the COO of my company. How is my job going to change in two years and in five years because of AI? Tell me what my job is supposed to look like. [15:42] systems architect for the company [25:06] remember it's not real [34:07] Paul Farnsworth: Don't just use chat. Use a project and add in your context document in there and you enhance that as you go and it will help you greatly. [36:58] Allison Murphy: enjoy being a learner. [36:58] Emily Acosta Lewis: communication is key [39:28] Allison Murphy: When was the last time that we weren't being evaluated based on skills we already had and knowledge you've acquired? We're at this unique moment for a little bit of time where everyone in your company, everyone on your team, everyone in your family has a chance to learn something new together.

[0:04] So, first we wanted to start off and do a poll to  see how much of a beginner you actually are. So, [0:09] if you're a newcomer, like brand new to AI,  you're learning the basics and have very little [0:14] experience, can you raise your hand? Okay.  And then if you're more of an experimentter, [0:21] like you've played around with AI a little bit  and use it for some tasks but not too much, raise [0:26] your hand. Okay. And then how many of you are  implementers where you feel like you've integrated [0:32] AI into your personal lives or your professional  work lives? Great. Okay. Well, we're going to [0:39] start off with a two sentence introduction from  each of our panelists. Um so, if you could just [0:44] introduce yourself and share one fundamental truth  about AI and um what you think everybody in this [0:50] room should know in term while they're starting  their AI journey or continuing their AI journey. [0:56] Yeah, go for it Paul. All right. Hi, I'm Paul  Farnsworth. Um, I'm a a career technologist. So, [1:02] I've been a CTO um for 35 years. Let's not say  that too loud. Um, so I've seen a lot of change. [1:09] Uh, currently I'm president of dice.com and we're  a talent marketplace for AI talent. And the the [1:16] thing I'm going to um talk about AI is not really  about AI, which is um technology has always been [1:21] about change, continues to be about change. Um,  I think that if you are, um, playing to learn and [1:29] you're willing to take some risks, um, you can you  can figure out how you adopt that into what you [1:34] do. The thing to remember with technology is that  if you're in a business, it's about what the goal [1:39] of the business is, not just the technology.  You can get taken down a pretty windy path if [1:44] you're just focus on the technology alone. Yeah,  great. Thanks, Paul. Andy, so I am Andy Dingor. [1:50] I'm the head of AI strategy in North America for  Daimler Truck. We build the semis on the road. [1:56] You've probably seen them. And what I am doing  is taking business ideas and how do we create [2:02] them into reusable tech across the company using  AI. And the one fundamental truth that's very [2:08] difficult is that people have to adopt behavioral  change of using AI to help run workflows other [2:15] than just doing everyday assistance and helping  make individuals productive. Great. Allison. Good [2:21] morning. I'm Allison Murphy. I am chief operating  officer at Axios, which is a news media company. [2:28] And I think the truth, you all just highlighted  it, which is it's incredible to see people at the [2:34] very beginning of their journey, people who are  all the way to implementation. And by the way, [2:39] if you're at implementation, you're not at the top  of the journey either because next week something [2:45] will be different. So the fact that this is an  expanded road map of learning is both daunting but [2:53] also an incredible opportunity because no matter  where you are, you're on the track because you've [2:58] chosen to engage to learn about this and to bring  your organizations along and I think that's the [3:04] the fact it's so fast moving and we are nowhere  near the end of this journey. Thank you. So [3:11] for audience members who might feel a little bit  overwhelmed by concepts like machine learning, how [3:16] do you personally define artificial intelligence  in simple terms when explaining it to your board [3:21] or cross functional teams? Um so I we we we've  worked on a model um I hope this resonates with [3:29] all of you whether you've just started to play  or you have a little bit more experience and [3:32] we and I think about AI in three big chunks  as you start to explore how you might use it. [3:38] One is there is um AI in a product. So if you're a  product engineer or you're part of a product team, [3:45] how do you implement AI inside that product  and that is also the foundational models, [3:50] the people working at the core of AI, the ones  making claude and all these kind of big engines. [3:54] That's AI as a as a product. Then there's AI at  work. So if you are how many people use Excel? [4:00] Probably a lot of you. Um, I had a my first great  experience with Copilot last week where I put a [4:07] formula in and I dragged it down and C-pilot went,  "Oh, it looks like you're doing this. This is a [4:10] formula for you." And I'm like, "That's great. AI  was working with me and there's AI at work." And, [4:16] um, you know, there's a lot of opportunity  for us to continue to use AI at work. Um, [4:21] we'll talk about what the good and the bad is of  that in a little bit. Um, and then the last bucket [4:26] I'd say is there's aentic AI and that's the one  which you you read a lot of headlines around which [4:31] is where agents are acting and doing the roles  that a person or or a task that a person would do. [4:38] And I think if you can decide to begin with where  you want to be focused, it will definitely help [4:43] you narrow down what you're actually starting to  uh starting to explore. So it is AI in a product, [4:49] it's AI at work, how are you using as an  assistant? And then there's aentic AI which is a [4:54] little different and that's another whole um whole  topic. Andy, do you have anything to add? Yeah, [4:59] that that what he said aligns with what I think of  directionally. The one part that I really liked is [5:05] Excel. In Excel, you did a lot of computations  and modeling and that helped do the engineering [5:11] and that was a step forward in doing computational  work. And when I look at AI, it's your human or [5:17] assistant helping you do the cognitive learning  in your everyday work and figuring out where you [5:22] want to go. And I constantly say at work, being an  AI first, use it as your spar partner to learn to [5:28] develop to to for your strategic leadership or  want to do. And it's it just helps you run the [5:34] workflow. And there's more two different ways  to do that is everyday AI. And that's what you [5:40] could do. go come into work and on Monday decide  what I spend most time on. Previously some people [5:46] might have create macros or done something to make  it repetitive and reusable. But now you go in AI [5:52] you you could use it to automate a process that  you do every day or a task either the everyday [5:58] task or the task that you hate doing which is  also a really powerful motivator. When I think [6:04] about talking to leadership or board, uh I think  one of the things we try to emphasize as well is [6:10] we don't need everybody to know and be indepth on  the very very latest thing that came out with a [6:17] model. In fact, that can be a huge distraction. So  recognizing where you need your employees and your [6:24] leaders to be, you need them to understand the  the broad scale of what the technology can do. [6:30] You need them to understand that it's directed  by our own focus and strategies. And you need [6:35] them to be informed enough and intelligent  enough to make smart choices in how they're [6:39] using the tools and a lot of that is familiar  with what we already do with technology in our [6:44] companies. Great. Thank you. Can you each tell me  about one AI initiative that you either crossed [6:51] the finish line with or one that's in currently in  development and highlight the biggest lesson you [6:56] learned from that project that would be sort of  relevant to a beginner and AI? I mean I think um [7:03] I'm going to we we have a broad set of initiatives  which encompass product and using AI at work and [7:11] um and agentic AI. I I'll start with something  I think which um is is simple um but is powerful [7:19] which is um I don't really write agendas anymore  for meetings. Um you know I think if you haven't [7:25] used the tools around meetings and in particular  um if you have the ability to record a meeting I [7:31] know you all probably see the transcriptions that  come back but just digging in and grabbing that [7:35] transcription and sticking it in an engine which  can actually read that is incredibly powerful. Um [7:41] the secret though I think isn't just to do it on  the one meeting. It's to start to actually keep [7:46] that corpus of data so you can go back and start  to compare did we do something between the last [7:52] one and that's a huge um timesaver in terms of  actually using AI. So if you haven't had a chance [7:58] to kind of experiment with that a little bit um  I certainly know I'm part of an organization who [8:02] has been um they've been concerned about using  transcription um because they want the essence [8:08] of the meeting and not the raw meeting. But I I  think at some point um all businesses are going to [8:13] end up transcribing a lot of the calls that we're  on. So I would start there. I'd start with some [8:19] basic transcription and action items and agendas  for meetings. So it's a big it's an easy win. An [8:24] easy win. I I could spin off what he said a little  bit is taking your meetings. What I do is instead [8:31] I take my use cases, our PC's or products we're  developing across the company and with AI you [8:37] have something called memories and that helps  have augment augmented recall of what is done [8:43] in the past as a human you can't remember exactly  what use case you did why it failed why it worked [8:48] but AI can if it's correctly documented and so  then we go to across the company to figure out [8:54] how do I want to use AI what use cases you have  this vault of all your use cases and AI will say, [9:00] "Hey, you did this in your function here. It could  work in function B over here. How can we make a [9:05] reusable tech underneath to enable both use cases  at the same time?" And so that's what we're doing [9:10] is across Dynamo right now is trying to figure out  how we build that reusable foundation to support [9:15] use cases across the whole company instead of  doing siloed use cases. Because when AI first [9:20] came out, we had a chatbot sprawl. You I'll say  every single team wanted a chatbot. So they hired [9:28] consultants or they tried to do it themselves but  it became a very problem for tracing it accuracy [9:34] credibility of the chat bots and adoption because  they didn't know where to go and so now we're [9:38] trying to figure out how can we create this one  chatbot interface that everyone goes to that pulls [9:43] from the same memory bank to support the whole  company the risk of internal AI slop is real I'll [9:50] give a a different example which is similarly  we encouraged early on at Axios all employees [9:57] to experiment with the technology and we wanted  to see what they came up with and it really was [10:03] surprising because we found employees in  unexpected places really seize up the tech. [10:11] So one of our best examples is the woman who was  responsible for accounts payable in our finance [10:16] department got obsessed. She automated probably  twothirds of the standard reporting in our finance [10:24] team invoicing. so many different things. She's  now running uh our internal AI training for our [10:32] other employees. She's moved out of finance and  she's now on a different career trajectory because [10:38] she took it upon herself and our monthly finance  close has dropped by two days. So all of these [10:44] uh gains have been felt. But the other great  thing about letting a thousand flowers bloom with [10:50] governance, so you're correct in that, uh, is you  can get your learning going on different cycles. [10:57] The finance team got so ahead so quickly that they  were the first ones to say, "Oh, wait, you know, [11:03] we could do this more efficiently, or maybe we  shouldn't use AI for this. Maybe we should just [11:08] have our engineering partners write a a quick  script for this. This doesn't need to be AI." [11:14] And so they're now on iteration four or five of  some of these tools because we gave this space [11:19] and the challenge to do it. Now you do have to  pay attention to governance. You you you know [11:24] this worked because it was within the confines  of one team who knew their work well. Um because [11:29] it is absolutely true that you can end up with a  thousand different chat bots all doing a mediocre [11:34] job. Um but finding that balance and encouraging  a mix of the on the ground innovation as well as [11:40] keeping your eyes out centrally um is I think the  balance that ends up working best. Great. Thank [11:46] you. Um you touched on this a little bit with your  your your previous answer, but when a business [11:52] leader decides that they want to start using  AI, but they have no idea where to begin, what's [11:57] your framework for identifying the very first  low-risk opportunity within their organization? [12:02] think there's um a difference between are  you using it for yourself to make yourself [12:06] more efficient or are you asking your team to do  something because you read an article uh top 10 [12:12] uses of AI in enterprise today just making it  up or five best prompts for a finance manager [12:18] that kind of thing um you you got to decide is  it you or is it your team um if it's the team I [12:23] think it's the same as any technology um change  management program you got to put some training [12:29] in place potentially make sure you understand what  the standards are and what you're actually asking [12:33] for. It's back to what is the business outcome you  after. Um it the the the core of business hasn't [12:39] changed. Um which is if you can measure something  and you can improve it, you're on the right track [12:44] as the as the core of that. Um I will tell you  though if you're going to start using it yourself [12:49] um and I'll expand a little bit on you know  maybe you're you're just starting to experiment [12:54] a little bit. I'll give you a top tip for me  from me recently that I discovered which was [12:58] um amazing uh is um once you get past chatting  like a single chat lacks some context and you're [13:05] moving into into something where you're creating  a project and putting documents into a project [13:10] so that there's some continuity in terms of what  you're doing. For me at work um I actually went [13:16] onto our internet. I downloaded the PDF of the  entire organization and I stuffed it into my [13:22] um project and said build me a text list of every  individual what their title is and what their role [13:28] is and who they report to. And that actually  meant that when I transcribed meetings rather [13:33] than it just saying Kevin said this, it had some  context around what Kevin's role was, who Kevin's [13:39] boss was, what Kevin was supposed to be doing. And  then it really helped with the context of that. So [13:44] as you think about iterating, the more context you  can give these projects, the better. And there's a [13:50] lot of great things you can do to actually enhance  the the understanding that the these agents have. [13:56] It's easy to go in thinking because it's AI,  it has all this amazing context, but actually [14:01] it's pretty dumb at the core of it unless you've  trained it. So I the context part is very crucial [14:07] because when I look at AI and chatting with it,  it is all you're doing is when you're providing [14:11] that prompt is the direction like what role should  you use what you want to do and the quality and [14:18] how I try to push AI within the company is it's a  behavior change it down to ground seem to because [14:26] when I look in IT you have some people that adopt  it that are extremely technical but then you have [14:31] some people in the business that actually  start doing AI first and thinking about how [14:35] can we critically use this in our business. They  actually understand the technical by you talking [14:40] to AI and they become a technical person in the  business and doing that. And so to me AI is a [14:46] journey. Everyone goes at their own pace. Every  team goes at their own pace. Every company goes [14:50] at their own pace. And to get there you you have  to do it. You can't just read about it. You can't [14:54] talk about it. You can't figure out what model  is best for this or what's best for that. It is [14:58] just getting in there and starting like I'm going  back to I said like on a Monday figure out how you [15:03] automate a process or a task and then you go from  automating your individual task to within my team [15:09] or domain how can I do a process augmentation  and what I mean by that is take a process that [15:15] you do today how can you use AI to help automate  that process and then eventually once you get [15:20] comfortable with augmentation of a process you go  okay now that we know how to augment our existing [15:26] process which might be 10 steps steps. How can I  make this faster and actually reshape the process [15:32] where I put AI as running the whole workflow?  Instead of a 10-step process that's automating, [15:36] it automates one step from A to B and eliminates  the other other steps in the process. And so [15:42] there's a whole journey that with teams get  there. I think that part of the question is [15:46] where do you start if you're sitting in this  room and I think these are fantastic ideas. [15:52] The other way you can come at it is from the the  big picture and this is something that our senior [15:58] leadership team did probably close to a year ago  now which is we went into chat GBT quad take your [16:06] pick and I asked I'm the COO of my company how is  my job going to change in two years and in five [16:12] years because of AI tell me what my job supposed  to look like and it was an incredible response [16:20] because it helped give a direction to how I needed  to evolve my own skills. So, absolutely everything [16:27] the guys have said is spoton with how you can  find ROI and look at processes. I think can [16:33] also help to give people the north star. And this  is something every employee can do. It doesn't [16:38] matter what level you're in. you go in and it  it says to me like you're going to stop managing [16:44] uh roadmap prioritization meetings and your annual  planning cycle is going to get shorter and all [16:51] the reports you have that are backward-looking  will become real time. But what it really said [16:56] is you're going to have to start to become a  systems architect for the company. Your job is [17:00] already how things work. Now it's going to be how  that works in a much tighter way with technology [17:05] and how that works through people and through  automation. And all of a sudden that helped [17:10] me kind of conceive of what do I need to learn  about? What do I need to be good at with this [17:13] technology? So there's still incredible day-to-day  efficiencies, but for me and I think for something [17:20] truly anyone at any level can do, that gives  you a set of just like learning priorities [17:24] that might make everything seem a little less  daunting. Uh when you can paint that picture [17:29] of here's how I want to use this technology to  grow my career and the impact that can have. [17:35] I always say when in doubt, ask AI, right? So if  you're not sure where to start, ask AI. Use AI [17:42] as a thought partner. Ask it what kind of prompts  you should be asking or what you're trying to do, [17:47] right? Use it as like a brainstorming tool  and then that'll help you sort of get started [17:52] on your journey too, especially if you're  feeling really overwhelmed by the technology. [17:57] Um, what is one practically practical daily  habit you recommend for business professionals [18:02] to stay updated on real AI capabilities without  getting bogged down in the technical jargon? Uh, [18:08] so this is a a short answer and something you  probably do which is um there is literally and [18:16] really literally millions of YouTube videos and  podcasters a podcast about AI every day. Um, [18:21] I think if you can find a couple of those that um,  talk about the higher level of the trends without [18:28] getting bogged down into the detail, it's a pretty  good way. And, you know, depending on, um, how you [18:33] consume media, um, YouTube might work for you or  a podcast in the car. Um, just making the effort [18:39] to listen to one of those every couple weeks is  probably enough. There's a lot of information and [18:43] a lot of change. I think you can stay more  broadly in touch with the what's happening [18:48] more generally and not not get too bogged down.  I'll you mind if I jump in the last one? I'll [18:53] give a plug here which is Axios has a handful of  fantastic publications uh truly that are are meant [19:00] to summarize and really tell you what matters.  So, we've got a great daily newsletter called AI [19:06] Plus that covers, especially if you're a business  leader, which most people in this room are, talks [19:12] about what's going on in terms of the Frontier  Labs, the business, the tech, but we've also been [19:16] weaving it into our other newsletters. Um, second,  the point on podcasts. There's one I love called [19:22] the AI daily briefing. Uh, that is it does a a  news brief every day of just like what's come out, [19:29] models, etc., but it also has a deep dive. uh  and they've started launching training programs [19:34] around it. So they'll talk about a new capability  and there's usually a free training program that [19:40] links that you can then go and participate in to  try the skill yourself. So axios.com and AI daily [19:48] briefings. So I I do something very similar.  I listen to podcasts and everything but I'll [19:54] I'll take it a different uh area. What I also  do is in AI when you run a prompt, you can also [20:01] schedule that prompt to run every day or every  week at what you want. And so what I also do is [20:06] I have my goals or different projects I'm working  on across the business and personal and I have AI [20:12] run that prompt on a schedule related to my goals  and then it goes and does re hyper research focus [20:17] to where I want to go and creates a road map of  what's new or what came out how to get there and [20:22] has a actual plan of step 1 2 3 4 that I could do  within that week to establish my goals within the [20:28] week and it's your own research partner that's  dynamic and always updated with what's coming [20:32] out. I don't keep track of the models and all that  stuff. Those will continue to change. I care about [20:37] how it's impacting what I'm actually working  on. I'm stealing that. That's great. One of [20:41] my team members went deep on this and and created  an agent that looked at all of the previous day's [20:46] interactions and created a little voice memo for  him in the morning. So, I started uh communicating [20:52] with him in Shakespeare and English just to mess  with his agent. [laughter] That's great. Um, [21:00] this audience will likely evaluate thirdparty  AI tools rather than building their own. What [21:05] is a massive red flag or a common trap  you want you warn non-technical buyers [21:09] to watch out for when taking uh when talking  to AI vendors over the next few days? I mean, [21:14] I've been in IT procurement long enough to know  that this is not a new phenomenon. Um, lots of [21:20] companies promising lots of things and flashy  demos without real depth. So I mean it's not a [21:25] different it's no different really than buying any  SAS platform. I would say um the the one thing you [21:31] probably want to ask if you want to dig a little  deeper is what are the foundational models that [21:36] the products are built on so you understand how  much change there's likely to be as a result of [21:42] that. Some products are built on their own models.  There's risk with that. Um some products products [21:47] are built on other foundational models. You should  really understand kind of what the dynamics of [21:51] that business there are. Um but apart from that  it's all the other usual diligence which is you [21:56] know you've got to dig one level deeper than  just a demo and actually understand and then if [22:01] you have the time highly recommend talking to an  implementation that has already been done so you [22:07] can get kind of a sense of their experience. Yeah,  this this one's a very hard one because we get [22:13] presented with new products multiple times a week  and they're all trying to do the same thing in AI [22:20] and there's one cue that you know that is a lot of  is AI generated like in their apps they have the [22:25] little purplish blue color throughout the whole  app and then you know AI generated they didn't [22:29] even take time to make it pretty they just went  with the AI guy. Uh but in what we really look [22:35] at is more business value that the use case is  bringing and not just the business value but also [22:40] is it a lockin to that platform and how does it  work with our source systems is it like a what I [22:46] call a AI embedded solution where it works on top  of a source system like SAP Salesforce workday or [22:53] is it more cloudnative where it runs in the cloud  orchestrates across all source systems and works [22:57] as a unit and you got to make sure it's what I  call is um adaptability or plugandplay of the tech [23:03] as they adopt because this SAS environment is on a  hyper train and it's going to continually evolve. [23:08] So you want to make sure that you can quickly pull  out one thing and go to something new if you need [23:13] to. So contracting becomes very critical whether  you sign a threeyear, one year or 5year and how [23:18] you want to handle it. Yeah, we are really only  signing one-year contracts as much as possible. Uh [23:24] because first of all this is wildly changing. So  some of these vendors won't be here in 6 or 12 or [23:30] 18 months. Um there's a few things we think about.  One, which is just always a good adage, is anyone [23:36] who's promising to solve everything, just be  afraid. And the version of that that's happening [23:41] now, I think is roadmap inflation. So it's like,  oh well, with AI, we'll have all these features [23:47] shipped in the next 3 months. They're not going  to be shipped in three months. They're not going [23:51] to be out. So that's the version um I'm seeing  now. The other thing I think that we have to pay [23:56] really close attention to which is similar part  of contracting is how will usage be buil. We're [24:04] at a moment when the cost of compute is changing  and all of these companies are also funded which [24:11] means the economics are pretty out of whack. So  what we all have to be on our guard for is you get [24:17] dependent on a technology and then suddenly it's  going to be costing you 20x more. And if every AI [24:24] power system you brought into place does that, you  know, you're really up a creek. And the last thing [24:29] I'd say that's just useful, we are all going to  be meeting tons of vendors here today. U this is a [24:34] great way to engage teams. Uh we're inundated with  uh different pitches. Uh this is a great chance to [24:43] say to someone who's interested on a team, I want  you to meet with four of these vendors. Tell me [24:48] which ones seem to jump out. How similar are they?  how are they different? Who seems to really have [24:54] their act together? And that's a chance to get  your teams more educated and excited about what's [24:59] going on as well and to be a bit more on the  forefront and edge of what's being built. Great. [25:06] So, looking back at your own journey on a with  AI from starting off at nothing to where you're [25:11] at now, what's one assumption you had at the  beginning that turned out to be completely wrong? [25:16] Um I think uh well true for those of you  that are are implementing a little bit more [25:25] um sophisticated technology um I assumed  OpenClaw was going to solve all my personal [25:31] life problems. So I have a little Mac Mini  just like lots of people went out and bought [25:34] a Mac Mini and I stuck Open Claw on there and  I put the frameworks on and I did all the work [25:39] um to to talk to it on my phone. And what you  discover of course like lots of technology is [25:45] the promise is there. It you can you can taste it.  You can taste the future but it gets stumbles it [25:51] stumbles over very very simple things. So I think  um I was very optimistic that this was going to be [25:57] the thing where there was less context needed but  like everything it's going to require much more [26:02] tuning. Yeah. What changed for me is so a little  background I was a principal data scientist and [26:09] I coded for 15 plus years doing machine learning  and all that work but AI got to a state I'll say [26:15] in December January where the models actually were  very good at writing code and developing and since [26:20] that point in time I actually haven't written one  line of code anymore AI dissolved for me and so [26:25] what I how I look at AI now is I call it an art I  I'm able to take my creative instincts what I want [26:32] to do and to build it I need to be able to design  it and communicate with AI to get to my goal. And [26:37] so it's being able to take the step back even  though I majored in programmed in coding to say, [26:43] "Hey, I don't need the skill anymore. AI could  take it." What I really like was the creativity [26:47] of building the product I wanted to get there.  And so it changed how I think about work now. I'll [26:52] give two going to break the rule and give two.  The first is that I thought I could learn about [26:59] AI without necessarily having to be in it. And  you really can. I mean, you should and can read [27:07] about everything that's going on and be educated  about the industry and the momentum, but this is [27:12] not technology that you can grasp the nuance and  potential of unless you're using it yourself. [27:19] And then the second thing is I think when you  do become adept and a pretty sophisticated user, [27:26] it is really easy to assume you're not going to  get duped by the AI. You know what to look for. [27:33] You know how to spot the biases. And that's when  it's important to remember that the technology [27:38] keeps getting better. So building in the systems  that will check your thinking and ensuring that [27:45] you keep skepticism around the outputs you're  getting is so so important and it gets harder [27:53] the more comfortable you are with the technology.  Um, so it's I have a little post-it note now on [27:58] my monitor that's like remember it's not real  and [laughter] it's like you because it can so [28:04] quickly you you build a level of trust especially  as you layer in all this valuable context and it [28:10] knows your team and your meetings like this is  great. No, it can still create uh massive errors [28:16] and hallucinations that can be really problematic.  Yeah, it happens all the time. Happens to all of [28:22] you all the time. I asked it for a recommendation  for a cocktail bar since I was staying at the [28:26] Venetian and it suggested something. I'm like,  "Would you double check that?" And it said, "Oh, [28:31] you're right. That's closed." I'm like, "Okay,  well, thank you." Yeah. Not perfect. A big concern [28:38] for beg beginners is the human element. How do  you practically foster a culture where employees [28:43] feel excited to experiment with AI rather than  fearing that they're going to lose their jobs? [28:49] And I think there's and I'm going to I'm going  to say you probably need a little bit of fear [28:53] in there as well if I'm going to be really  honest. Um you know we're driving through a [28:57] lot of change and and um you've there's this you  know pretty common um position which is there's [29:04] lots of jobs that will exist um with AI and the  jobs that don't exist won't be with AI. So I I [29:11] think encouraging adoption um I think it comes  back to the culture of the organization. Um, [29:16] if you are rewarding some experimentation and  some failure, you're likely to see better results [29:21] from employees wanting to try it. If you are  penalizing them for experimentation and failure, [29:28] you're likely to see conservatism on the change.  So, I you it's a it's back to technology change [29:33] and the culture that you're you're providing your  employees. Yeah. So this one's interesting and [29:40] how I how I take it is I have a opinion that the  companies that choose to do layoffs because of AI, [29:46] they're going to get left behind. And what I  think the better thing is to do is take those [29:51] individuals that went to school to do strategic  thinking. They're currently doing repetitive tasks [29:56] that AI is currently doing day in day out. Have  them start to do more strategic advising within [30:01] their domain or across the company. that use cases  that will more your big bets use cases that are [30:06] like three to five years out that will get you the  competitive advantage and start working on that. [30:10] So it's moving your resources from the repetitive  task that you don't want them to do anyways to [30:15] where can we get our competitive advantage. I  mean I work in the business of journalism so it [30:20] will not be surprising that transparency is a huge  [snorts] feature of our company. There is also a [30:26] long and really dismal uh history of contraction  of journalism because of technology. We can have [30:33] a whole conversation about that later if you want.  You can find me. Uh but what it means is we have [30:38] already constantly employees who fear for not just  whether their job but whether our industry will [30:45] will persist in a changing information age. So  what we have done is be completely upfront about [30:53] all we know and all we don't know which is that  our industry will be absolutely upended by AI. [31:00] But we intend to do everything that's in our power  to equip our teams and our employees to succeed in [31:06] that environment. And that relates to our business  strategy. That relates to the resources we are [31:11] giving them. You know I mentioned we moved  this fantastic woman from our finance team [31:15] to a training role. We've moved three employees  into those roles, which we're 500 people. So, [31:20] that's like a pretty good number. Um, and kind of  putting our money where our mouth is on it because [31:26] at the end of the day, when we look at technology  for our industry at least, these aren't axios [31:31] choices. This is how the world is evolving.  And so, we can either equip our employees and [31:37] help them see where their career is going and  help them get there, or the company as a whole [31:42] can't succeed. and being really really upfront  has worked for us. Granted, journalism company, [31:47] so there's a different uh kind of culture than  some other places, but it's uh it's hasn't failed [31:53] us yet. Great. Um if an attendee leaves the room  today and wants to take one immediate practical [31:59] step to bring AI into their daily workflow, what  exactly should that action be? I'm not going [32:04] to answer that question straight away. Going  to tell you a story first. My father retired [32:08] um uh quite a few years ago and did a family tree.  It seems to be an English thing. You go and start [32:13] exploring your family. Uh got to a point in our  family tree um with a gentleman called Brandreth [32:19] and he couldn't find the the branch that went  further back. And it turned out the reason was [32:25] that the family had changed the spelling of their  last name because the original Brandreth was one [32:30] of the last people executed as part of the lite  movement in England protesting against looms. [32:36] So there you go. It's this is not new. None of  this fear is new. Um so back to your question um [32:43] uh my boss uh a a few bosses ago um when I when I  started working with him gave me a document that [32:49] basically said this is how you should work with me  and it was helpful. It's like this is the style of [32:54] communication I like. This is my hobbies. This  is the kind of way I think about um interacting [33:00] with my staff. I like a report each week. It was  very very helpful. Um I have since adopted that. [33:06] I have a little document I give to my staff. they  probably roll their eyes now and stick it in AI. [33:10] Um, but that's fine. I would say my one tip is  as you start to develop your um um relationship [33:18] with whatever AI model you're using is keep  that context document going. What style do you [33:24] look for? You're asking it to double check things  and then when you do the work in these um tools, [33:30] don't just use chat. use a project and add in your  context document in there and you enhance that as [33:36] you go and it will help you greatly because a lot  of the assumptions that it will make you can keep [33:41] that context document up to date. And I was just  saying this terrified Andy by the way when I told [33:46] him this I now also have a context document  for my family too which expresses what their [33:51] interests are. So when I'm using it for planning  um it already knows my kids' names and what my [33:56] wife's preference are and where I prefer to stay  all that sort of stuff. So it's context document [34:01] is something you should uh you should keep up to  date. It's very helpful. Yeah. I I would say to [34:07] really get comfortable with AI. I mentioned before  is just start doing it. Try to start talking with [34:12] AI at least 15 minutes a day and then you'll start  to get more comfortable about its capabilities, [34:17] what it could do. And how I eventually started  using it more is my strategic partner. And so [34:22] I say I go do a presentation in engineering on  a topic and I do one in aftermarket. I actually [34:27] can use AI to figure out what are the leadership  in engineering going to think about what are the [34:31] questions they're going to ask ask me and I can  do the same thing aftermarket I can do the same [34:35] thing across the whole company and so I already  know from their point of view what to speak to [34:40] how to talk to them and and relate and I do that  inside of AI you could create personas and going [34:46] back to the context engineering and prompting  say I am a senior leadership in aftermarket I [34:52] am presiding on this can you help me prepare  for what they're going to ask me about and so [34:56] is talking with it and use it as your assistant.  You have it right at your fingertips. I would add [35:01] to that and say go one step further and make it  your default. So, at first I thought to myself, [35:07] I'm not sure how AI can be all that helpful  because 90% of what I do is, you know, [35:12] human dynamics and complicated people stuff. I  thought, all right, well, let's just challenge [35:17] that. Let's really spend the time giving a problem  to um to an AI to see if it can help. And there [35:24] was an org redesign thing that um I threw in  there and I was astonished by the results. So [35:30] uh not just treating it as an assistant because  with an assistant you're thinking, oh here's a [35:34] task you could be good for. Go one level before  that and ask how would you do this thing and bring [35:41] everything to it first. Don't assume what the AI  will or will not be able to help you with. You [35:48] don't have to then choose to use it, but at least  ask, could you do this thing for me? Uh, and the [35:54] other really tactical tip I'll give is if you're  not using voice, well, you're missing out because, [36:00] uh, you save so much time by just speaking that  versus having to type things out. And that's what [36:07] I did with this org project is I was like, no,  let me just talk for 10 minutes about everything [36:12] we're trying to solve and everything that's hard  and easy and all the ideas. And that was what was [36:17] required. when you speak, you're also giving  much more context than you are when you filter [36:23] into writing. So, um, use it as your assistant,  but even more than your assistant, almost use [36:28] it as as your manager of like what should you be  doing, what should I be doing, uh, and then use [36:33] voice mode. I would say too, if you're intimidated  by using it in your workplace or you're worried [36:38] about data privacy and that kind of thing, start  using it at home. Use it for meal planning, [36:43] use it for budgeting, use it for your kids camp  schedules. Um, whatever you can do to just get [36:49] access to it and start playing around with it.  It'll help you to then kind of think about ways [36:53] that you can use it in a workplace. Okay. So,  to close out, what is one piece of advice or key [36:59] takeaway you want the audience to remember as they  continue their AI journey? It's just a computer [37:05] and it doesn't have feelings. So you can ask it  whatever you want and you can delete the the chat [37:11] afterwards. Um you know it's one of those things  where as I said at the beginning play is good. [37:16] So and it's also kind of fun. So just have fun.  Yeah. Mine is more it's you have this power this [37:25] brain that you could have to do your personal life  at work. You could have it do whatever you want. [37:29] And so I am able to build projects in a day and  build a whole application hosted on the web very [37:37] quickly and test ideas quickly in personal life  and at business. Just use it as your thought [37:43] thought process and throughout the whole day and  it's a journey. It's going to whatever what we [37:48] have today is going to get leaps and bounds in  a few months and it's instead of tech doubling [37:52] every six months it might be three months with  with the tech and how a is going right now. [37:57] Mine would be enjoy being a learner. You know,  when was the last time that we weren't being [38:04] evaluated based on skills we already had and  knowledge you've acquired? We're at this unique [38:10] moment for a little bit of time where everyone  in your company, everyone on your team, everyone [38:15] in your family has a chance to learn something new  together. And that's really special. That does not [38:21] happen all that often. That doesn't mean there's  not accountability and we're not on the hook for [38:26] ROI and priorities and all of this, but uh to  embrace this chance to get to have a little fun [38:32] and and explore something new is not something we  get all that often in our careers. So maybe, you [38:39] know, a little terrifying as well and dramatic,  but there's a flip side of that too, which is [38:43] really exciting. And as the communications  professor in the room, I would say communication [38:49] is key, right? So garbage in, garbage out. If  you're getting output that you don't like, it's [38:54] probably because you didn't put the right context  and prompt into it. And then also communicating [38:59] within your organization, right? Being really  open and honest and transparent about how it's [39:04] going to be used and how much you're going to  monitor it or not monitor it or check. Um I [39:09] know that some companies now are monitoring, you  know, token usage and if you are not using tokens, [39:14] then they are you're you're on the chopping block  in terms of your job, right? Because you're not [39:19] being as efficient as you can be in your role. And  so there's a lot of different ways that companies [39:24] are approaching it and you need to think about  what works for your company and your organization. [39:28] You if you're younger in your career, um, now  is the time to experiment. I worked for a bank a [39:33] long time ago. We had a big website when websites  were cool. Um, and that's what banks wanted to do. [39:39] We had a webcam that was cool because it was new  and you could put it on the internet and someone [39:43] brought a um a light up Taj Mahal back from India  and we unplugged the webcam and plugged in the [39:49] Taj Mahal cuz it had some fairy lights in there  and a lovely bank website went down because we'd [39:54] embedded the webcam in part of the code. It took  us a few minutes to figure out what happened and [39:59] we apologized but I'm still here so you have the  chance to experiment. It's not the the consequence [40:05] of learning is great. So, um, don't don't fear  don't fear change. I I would add we're at an [40:10] interesting point in tech and work right now that  you could almost get any job you want because you [40:16] have your assistant and roadmap to get you there.  For example, I have seen people I've recently had [40:22] someone where they were a program manager, not  a tech person, built a road map based off a job [40:28] description to learn AI and was able to do it in  like two months to catch up to tech people. Yes. [40:35] even for a program and understand how it works  project-wise. But was it just took the whole [40:40] job description, took her resume and was like,  how can I get to this position and I developed [40:45] the whole road map and she was able to do it. And  when I'm hiring people, when we're hiring people, [40:49] that's that's who I want with the person who can  can see here's where we're heading and I'm going [40:54] to be creative and nimble and flexible about how  we get there. We're going to be hanging out after [40:59] so if you have any questions for us, feel free to  meet us outside because there's another session [41:03] after this. But I just want to give a big round of  applause to our panelists and audience. [applause]