Daimler Truck Replaces Legacy ERP With AI as CIO Bars Vendor Lock-In

戴姆勒卡车用AI替代传统ERP，CIO严防供应商锁定 随着内部AI实现事务性工作流自动化，企业软件正面临颠覆；但在半数企业数据尚未就绪的情况下，高管必须设法缓和董事会对立即裁员的要求。 该专家小组为引领企业AI转型提供了实操路线图，强调在仓促推进试点之前，赋能一线管理者并开展深度调研至关重要。与会嘉宾分享了如何在不采取一刀切禁令的前提下，通过内部沙箱、前瞻性专项工作组以及自下而上的员工技能提升，来平衡快速创新与严苛的安全要求。他们还审视了架构层面的演进，告诫各方警惕传统供应商锁定，着力化解企业数据就绪瓶颈，并倡导依托精心构建的生态合作体系打造模块化、解耦的系统。 企业必须将 AI 战略重心从核心产品扩展至深陷“AI 沙漠”的日常运营与一线业务经理，通过将工具直接无缝嵌入工作流来挽留关键人才并提升留存率。 (3:03) 在强监管行业中，企业可借助双轨创新模型在外部创新中心对高风险原型进行隔离验证，再经由结构化流程平稳过渡至核心业务；同时切忌一刀切禁用生成式 AI 以防催生影子 IT。 (7:50) 负责任 AI 原则与数据安全边界须在用例立项之初确立，企业应通过内部沙箱与自研智能体编排核心能力，避免商业服务条款利用专有数据训练第三方模型。 (11:08) 企业 AI 项目落地的首要防线是让终端用户深度参与前期需求探索阶段，以此避免斥巨资打造出无人问津的“烂尾软件”，并需采用追踪摩擦与工时削减的新型迭代指标取代传统 KPI。 (21:36) 随着自主内部系统在供应链等场景逐步取代传统 ERP，技术管理者须有效管理董事会对立竿见影削减 40% 人力的不切实际预期，正视约半数企业数据仍需深度清洗的现实瓶颈。 (33:52)

Enterprise AI transformation succeeds when organizations shift focus from executive hype to frontline workflow discovery, supported by modular architectures that prevent vendor lock-in. According to industry leaders across aviation, trucking, and enterprise software, sustainable execution requires pairing internal technical capabilities with structured workforce upskilling and proactive governance rather than reactive bans. Grounding Deployments in Frontline Workflow Discovery (3:03) James Briggs of AI Collaborator observes that while enterprises have spent years integrating AI into core products, operational managers outside those initiatives remain stranded in "AI deserts" and risk attrition without workflow support. To avoid costly, unused software, the panelists argue that leaders must prioritize an extensive discovery phase before launching technical pilots. Rushing straight into proofs of concept often fails because teams neglect to secure foundational user buy-in or align tools with everyday tasks. Payal Patel from Freshworks notes that corporate AI strategies frequently stall when deployments are not mapped to quantifiable outcomes. She advises organizations to replace traditional metrics with AI-aligned indicators focused on iterative time reduction and productivity gains. Involving frontline workers early to evaluate pain points ensures that deployments address real operational friction, creating the workforce ownership needed to achieve measurable returns on investment. Proactive Governance and Safe Internal Sandboxing (10:41) Addressing corporate risk, the panel emphasizes that responsible AI practices must serve as the primary baseline of any initiative rather than an afterthought. Reena Solomon of Airbus shares that abrupt enterprise bans on generative AI—prompted by employee leaks of proprietary code to public models—ultimately stifle innovation while driving shadow AI usage. Instead of blanket prohibitions, Airbus utilizes embedded task forces within business units to define clear operational boundaries and safe pathways for tool adoption. To safeguard sensitive enterprise data while enabling experimentation, companies use dedicated environments. Patel highlights Freshworks' internal sandbox, "Cloudverse," which allows personnel to test leading models such as OpenAI's ChatGPT and Google Gemini without risking corporate data leaks, supported by their customer-facing "Freddy AI Trust" framework. Briggs adds that effective risk management requires a centralized governance platform to monitor active models, tracking tools, and third-party vendors across the enterprise rather than relying on fragmented team-level oversight. Mitigating Automation Fears Through Continuous Upskilling (25:18) Organizational culture and employee apprehension around job replacement represent major hurdles to AI adoption. Solomon compares the advent of AI to the introduction of personal computers, arguing that successful integration depends on comprehensive education programs that teach workers why AI is vital to their careers and the company's future, rather than focusing solely on technical instructions. At Daimler Truck North America (DTNA), CIO Lutz Beck describes bridging the gap between an offshore delivery center of 200 technical specialists in India and regional staff through "HACK," an internal grassroots network of approximately 900 US employees focused on peer-led upskilling. Furthermore, the panel recommends introducing tools through collaborative beta releases rather than top-down mandates, framing software as an assistant that simplifies daily work to encourage constructive feedback instead of defensive resistance. Decoupling Architecture and Avoiding Vendor Lock-In (33:52) On technical infrastructure, Beck asserts that AI is fundamentally disrupting traditional IT environments, projecting that AI systems may drive up to 70% of operations between 2030 and 2040. DTNA achieved 90% automation in a supply chain and procurement workflow using internal agents entirely independent of legacy enterprise resource planning (ERP) suites. Beck cautions technology leaders to temper board expectations that unrealistically anticipate immediate 40% headcount cuts, noting that enterprise data readiness currently hovers around a 50% usability threshold, requiring significant remediation. To retain long-term agility, Beck enforces a zero-tolerance policy against proprietary vendor lock-in, insisting that enterprise architectures remain decoupled and modular so individual components can be swapped as superior models emerge. For vendor procurement, Solomon advocates a "dual bet" strategy that combines established platform providers for core stability with selective, specialized startups for niche innovation. Briggs adds that because even large corporations face internal talent constraints, leveraging vetted external partner ecosystems is essential to scale implementations without building everything in-house. While the panel anticipates widespread restructuring of enterprise IT architectures over the next decade, unresolved practical challenges remain for mid-sized organizations attempting to evaluate wildly divergent vendor quotes, manage uncurated legacy data, and assess advanced agentic frameworks without incurring excessive financial overhead.

企业人工智能转型的关键在于跨越核心技术实验与日常运营之间的断层，将赋能重心转移至一线业务流程。来自Daimler Truck North America、Airbus、Freshworks与AI Collaborator的行业高管指出，成功推进AI不仅需要以严谨的业务发现阶段替代盲目的概念验证，还须在架构上彻底解耦以杜绝供应商锁定，并在审慎治理与全员技能培育中平衡组织预期。 业务发现与一线工作流赋能 (3:03) James Briggs指出，过去五年企业大多专注于将AI嵌入核心产品，但大量非产品业务与运营管理者仍处于“AI荒漠”之中；若无法在日常工作流中提供支持，企业将面临一线核心骨干流失的风险。Payal Patel进一步分析，许多企业AI项目停滞是因为缺乏将具体能力映射到可量化商业成果的机制。企业必须建立战略能力矩阵，放弃失效的传统KPI，转而采用以工时节省和产出提升为核心的迭代指标来衡量实际业务价值。 针对项目夭折问题，专家一致强调前期深入发现阶段对规避高昂“闲置软件”的决定性作用。Briggs与Patel认为，匆忙推进概念验证（PoC）与试点往往会因缺乏跨职能协同而失败。通过早期梳理员工实际痛点、邀请最终用户共同设计应用界面，并以协作式“Beta测试”的形式推行，员工才会将新系统视为真正的减负助手，从而建立对新技术的根本信任。 负责任治理与安全赋能沙盒 (10:41) 在治理体系构建上，Briggs主张将负责任的AI原则作为初始底座，在此基础上建立用例合规验证、模型数据出入控制以及集中式企业监控平台，以便管理层全面追踪所有活跃模型与外部方案。Patel介绍了Freshworks的防护实践：内部通过“Cloudverse”安全沙盒让员工使用Google Gemini与OpenAI ChatGPT等前沿模型而避免专有数据泄露；对外则建立“Freddy AI Trust”透明框架，向企业客户明确数据流转与加密协议以消除安全顾虑。 针对安全事故的应对，Reena Solomon以过往企业因员工违规粘贴源代码和保密邮件导致数据泄露后全面封杀AI的案例说明，简单粗暴的禁令只会扼杀创新并催生不受控的影子AI。Airbus通过在各业务单元派驻专项工作组提供安全合规工具并划定清晰边界；同时依托硅谷的Acubed创新中心开展高风险原型验证，去风险化后再移交给中央评估与生产部门。此外，Daimler Truck North America通过全球与区域AI委员会、首席数据官（CDO）、200人规模的印度技术交付中心，以及覆盖约900名员工的“HACK”内部社群，自下而上推动全员技能进阶与文化认同。 架构解耦与自研核心能力 (33:52) Lutz Beck指出，传统单体企业资源规划（ERP）系统正面临AI自主流程的冲击，DTNA已在供应链与采购业务中利用内部AI工具实现90%的端到端自动化，完全绕开传统ERP系统。然而，企业在拥抱AI时普遍遭遇数据瓶颈：现有企业数据中约50%可以直接使用，其余50%需要进行大量治理与修复。Beck预测2030至2040年间，随着拒绝使用传统手动工具的AI原生代员工进入职场，企业70%的系统运作将由AI驱动，因此技术决策者必须直面现实并校准董事会的过高预期，打破立刻削减40%人力或短期节省数百万美元的虚假承诺。 在核心技术掌控方面，DTNA已在得克萨斯州公共道路实现商用自动驾驶卡车常态化运行。面对商业智能体普遍存在的利用客户输入训练底层模型的隐私条款，DTNA坚持在售后和销售等高价值场景自主研发并编排内部Agent。Beck强调企业必须建立解耦与模块化的现代IT架构，坚决执行对供应商锁定的零容忍准则：无论外部产品性能多么出众，一旦存在生态锁定便立即予以排除，以确保架构能够随着技术演进随时进行“即插即用”式的组件更替。 供应商评估与生态系统协同 (39:51) 面对外部市场从20万美元到逾100万美元不等、充斥各色术语的混乱报价，Solomon建议企业采取“双重下注”采购策略：一方面采购主流平台型厂商的基础套件保障整体稳定性，另一方面甄选在检索增强生成（RAG）或特定智能体领域具有突破性创新的专业初创公司，并根据客户实际采用率、资方背景与落地案例严格筛选。Patel与Noa Davidson指出，企业在决策时应穿透光鲜的演示，通过早期沙盒测试精确计算试点的部署周期、配置门槛与总体维护成本，防止因低估实施难度而造成巨额投资损失。 Briggs总结认为，即使是全球规模最大、资本最雄厚的跨国企业，也受制于内部专业人才与资源瓶颈，无法凭借一己之力构建涵盖所有垂直领域的AI方案。企业转型必须转向外部生态合作，依托经过严格筛选与验证的专业AI服务商网络来协同落地。这种生态化分工与第三方部署支持，能够有效降低企业的试错风险，是组织在快速变化的技术周期中维系竞争力的必然路径。 该讨论表明，企业AI转型的最终成效取决于前期业务发现的深度、底层数据的可用率以及模块化架构对技术锁定的防范。目前企业约50%的数据治理缺口、组织内部对岗位被替代的忧虑，以及董事会对短期降本裁员的不切实际预期，仍是推进转型的核心阻力；技术领导者必须在自主研发核心智能体与融入外部专业生态之间保持战略平衡。

## Background 本内容由 YouTube 频道 Ai4 于 2025 年 8 月 27 日发布，记录了一场题为《引领 AI 转型：指导项目从战略走向落地》的圆桌论坛。视频汇聚了来自 FPV Ventures、Freshworks、戴姆勒卡车北美（Daimler Truck North America）、空中客车（Airbus）以及 AI Collaborator 的高管与专家，重点探讨企业领导者如何跨越技术与组织鸿沟，将 AI 倡议从顶层战略平稳推进至全流程规模化落地。 ## TL;DR 圆桌嘉宾围绕企业级 AI 转型展开了深度剖析，指出打破核心部门外的“AI 荒漠”并夯实前期的“需求发现阶段（Discovery Phase）”是获取业务认同与确保 ROI 的核心前提。针对治理与文化挑战，专家们主张避免“一刀切”封禁，通过设立内部安全沙箱、全员技能重塑（Upskilling）和新型 KPI 体系来化解员工焦虑并规范数据合规。在架构与选型层面，戴姆勒卡车与空客等领军企业强调将核心 Agent 编排能力收归内部，坚决拒绝厂商锁定（Vendor Lock-in），并依托模块化即插即用架构与优质生态伙伴，步步为营地推进智能化重构。 ## Atmosphere & style 整场讨论务实清醒且极具战略高度，彻底褪去了大模型炒作的浮躁外衣，直击大型企业数字化转型的深水区。来自跨国制造业巨头、云服务商与投资机构的嘉宾们交流默契，既有坦诚分享早期数据泄露失败教训时的自省，也有在痛批传统 ERP 僵化与厂商锁定时的一针见血。对话节奏沉稳干练，将高维战略洞察与一线实操避坑指南无缝交织，让听众深刻感受到在强监管与高不确定性地基上“筑造摩天大楼”的真实张力与破局智慧。 ## Key takeaways - [03:03](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=183s) 破除企业内部“AI 荒漠”：企业不能仅把 AI 资源集中在核心产品研发上，必须深入一线业务管理者的日常工作流，将技术能力与顶层商业战略及 ROI 矩阵紧密绑定。 - [05:29](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=329s) 大型组织治理架构：戴姆勒卡车设立全球与区域 AI 治理委员会剔除无商业价值的“玩具项目”，由 CDO 攻坚数据质量清洗，并组建数百人交付团队与跨部门兴趣社群。 - [07:50](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=470s) 空客“双轨推进”创新架构：通过硅谷创新中心 Acubed 开展高风险敏捷前沿实验，再经由中央安全评审分层平滑过渡至业务单元，平衡草根创新活力与航空级合规标准。 - [11:08](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=668s) 构建负责任 AI 与全周期护栏：AI 治理必须从用例立项初期就嵌入核心价值观审查，打造跨业务部门的统一监控与追溯平台，拒绝“设置好就一劳永逸”。 - [13:18](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=798s) 以受控沙箱替代“一刀切”禁令：面对数据安全隐患，粗暴封禁会适得其反；企业应构建受控测试沙箱（如 Cloudverse），并向业务部门派驻专属工作组进行规范宣导。 - [17:42](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1062s) 核心能力内化与警惕数据反哺：企业应将 Agent 开发与数据编排能力沉淀在内部，严格审查第三方服务协议，防止敏感工业数据被外部供应商用于模型训练。 - [21:36](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1296s) 将重心前置到“需求发现”阶段：盲目上线大型试点容易无人问津，企业应优先投资前期业务调研，拉齐所有利益相关者共识，从小步快跑的 PoC 稳步扩展。 - [25:23](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1523s) 破除 AI 焦虑与全员技能重塑：员工抵触往往源于对失业的恐惧，企业须推行全员通识教育与 Upskilling，让员工深刻认知 AI 正在成为像 PC 一样不可或缺的生产力工具。 - [33:52](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2032s) 去 ERP 化趋势与正视数据治理现实：未来 5-10 年 AI 将颠覆传统企业 IT 架构，但当前企业往往仅有约 50% 的数据可用，管理层必须纠正“AI 立即裁员 40%”等不切实际的泡沫预期。 - [37:58](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2278s) 利益相关者映射与自下而上试点：借鉴达美航空经验，在业务部门中发掘先锋员工参与前期定义与 Beta 试点，以“赋能与减轻负担”的沟通姿态化解落地阻力。 - [38:50](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2330s) 中小企业选型聚焦易用与运维：面对繁杂的 AI 概念与高昂报价，企业应重点评估实施难度、配置复杂度和长期维护成本，优选开箱即用、去繁就简的方案。 - [43:16](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2596s) 拥抱模块化架构与严防“厂商锁定”：企业 IT 架构必须告别单体束缚转向即插即用的模块化设计，将是否存在 Vendor Lock-in 作为技术选型的一票否决项。 - [44:26](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2666s) 借力生态伙伴破解资源瓶颈：没有任何单家企业拥有解决所有 AI 问题的全部资源，借助专业协作平台对垂直供应商进行前置审查，是保持敏捷转型的核心策略。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=4s) 开场介绍与嘉宾背景 主持人 Noa Davidson（FPV Ventures 运营合伙人）宣布圆桌论坛正式开始。FPV Ventures 是一家总部位于旧金山、资产管理规模（AUM）达 10 亿美元的风险投资机构，主要聚焦于 Series A 和 Series B 阶段的初创企业。随后，四位深耕 AI 领域的重量级嘉宾依次进行了自我介绍： - **Lutz Beck**（Daimler Truck North America / 戴姆勒卡车北美首席信息官 CIO）：长期在美负责企业信息化与数字化，并在多家风险投资机构担任 CXO 顾问，热衷于挖掘前沿初创团队。他强调戴姆勒卡车在自动驾驶卡车领域已有长达 7 年的深度研发积累，AI 始终是公司业务创新的核心引擎。 - **Payal Patel**（Freshworks 美洲区解决方案工程负责人）：在 Freshworks 工作已有 6 年半，带领团队与各行业客户紧密合作，主要借助客户体验（CX）与员工体验（EX）解决方案帮助企业攻克业务痛点。 - **James Briggs**（AI Collaborator, Inc. 首席执行官 CEO）：专注于帮助大型企业应对席卷 IT 部门的 AI 激增需求。公司打造了行业首创的智能体 AI（Agentic AI）需求管理平台，并依托全球 AI 方案与服务商网络，帮助企业全面构建应对 AI 浪潮的敏捷能力（AI agility）。 - **Reena Solomon**（Airbus / 空中客车 AI 合作伙伴关系负责人）：常驻空客位于硅谷的创新中心 Acubed（$A^3$），主导通过生态合作推进空客的 AI 转型。此前曾在三星（Samsung）及多家知名科技企业任职，长期致力于探索 AI 对产业格局的颠覆与重塑。 --- ### [03:03](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=183s) 业务优先度与破除企业内部的“AI 荒漠” 主持人提到，现阶段很多企业高管每天都会收到大量询问“我们公司在 AI 方面究竟在做什么”的邮件；她在主管可口可乐全球加速器时，办公桌上也堆满了各色机器人和原型。针对企业究竟该如何起步并进行优先级排序的问题，James Briggs 指出，过去五年间多数企业把 AI 资源高度集中在核心产品研发上，导致核心产品线之外的绝大多数业务经理都处于资源匮乏的“AI 荒漠”（AI deserts）之中。企业必须将 AI 落地到实际承担业务交付的一线管理者手中，让他们能在日常工作流中用上 AI，否则企业将面临严重的人才流失风险。 Payal Patel 补充了从业务结果出发的价值评估框架。他指出，AI 项目若想产生切实的投资回报率（ROI），就必须将技术能力映射到底层使用者的具体工作成效上。团队必须与各级利益相关者深入沟通，并促使高管层将 AI 工具能力与公司的顶层战略举措牢牢绑定，这样才能让企业真正看清 AI 方案引入后的商业价值与落地回报。 ```mermaid flowchart LR accTitle: AI 优先级排序与价值落地链路 subgraph S1["突破核心边界"] A["核心产品线专用 AI"] --> B["赋能一线业务经理"] end subgraph S2["业务价值对齐"] B --> C["梳理日常工作流痛点"] C --> D["匹配高管层战略目标"] end subgraph S3["成果转化"] D --> E["明确 ROI 评估矩阵"] E --> F["规模化业务落地"] end ``` --- ### [05:29](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=329s) 戴姆勒卡车的大型组织架构与价值驱动机制 针对大型全球化企业如何搭建 AI 组织架构的问题，Lutz Beck 详细拆解了戴姆勒卡车的全球与区域协同体系： - **全球与区域 AI 治理委员会**：设立全球统一以及各区域并行的 AI 与数据分析委员会（AI & Analytics Board），负责紧扣价值驱动要素（Value Drivers），对项目组合进行严格的战略对齐与高优先级业务筛选，坚决剔除没有商业价值的“玩具式项目”。 - **首席数据官（CDO）与数据质量攻坚**：作为一家历史悠久的制造企业，历史遗留数据的质量往往难以直接支撑现代 AI 模型。各区域配置首席数据官团队，专门负责数据供给（Data Provisioning）与数据质量清洗。 - **全球交付团队与本土人才社区**：在印度设立了约 200 人规模的 AI 交付中心，各区域同步配置交付团队。同时，在北美内部孵化了一个名为“hack”的兴趣社群，汇集了公司内部约 900 名对 AI 和数据充满热情的员工，用于跨部门技能培训与组织技能升级。 --- ### [07:50](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=470s) 空中客车的“边创新、边合规”双轨推进架构 Reena Solomon 表示，空中客车同样搭建了专职 AI Hub、创新中心以及下沉至各业务单元的数据专员（Data Champions）网络。面对 AI 几乎每天都在爆发突破性进展的现状，Reena 形容这种扩张过程就像“在不断发生地质位移的地基上盖摩天大楼，而且建筑设计图纸还在随时被修改”。尤其在航空航天这种将安全、可靠与合规置于绝对首位的强监管行业，空客探索出了一套独特的混合创新架构： ```mermaid flowchart TD accTitle: 空客 Acubed 创新与业务转化双轨架构 subgraph Lab["硅谷创新中心 (Acubed)"] A["吸纳硅谷本地顶尖技术人才"] B["开展高风险、高回报前沿实验"] C["敏捷快速原型制作与概念验证 (PoC)"] A --> B --> C end subgraph Filter["跨越鸿沟与集中评审"] D["验证技术可行性与商业价值"] E["中央集中评估与合规安全审查"] C --> D --> E end subgraph Core["核心业务部门 (Business Units)"] F["分层式平滑过渡 (Layered Approach)"] G["高可靠性、强合规的规模化应用"] E --> F --> G end ``` - **外围敏捷探索与去风险化**：硅谷创新中心 Acubed 负责在核心业务之外吸纳本地人才，开展高风险、高回报的前沿实验与快速概念验证（PoC）。 - **分层过渡与集中评审**：一旦技术可行性被充分验证并确认具有清晰的商业价值，项目会经由中央评估流程进行安全审查，再以分层递进的方式平滑交接给各业务单元，既保护了自下而上的草根创新活力，又确保了严苛的航空安全与合规标准。 --- ### [10:41](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=641s) 负责任 AI 与治理框架的初步构建 当主持人引导话题转向各大企业避无可避的“AI 治理”（Governance）与合规防线时，James Briggs 提出，企业在部署 AI 时的平衡之道必须始于扎实的“负责任 AI”（Responsible AI）实践基石。企业应当在推行创新的同时，建立起贯穿整个生命周期的保障机制，确保所有应用都在安全可控的边界内运行。 ### [11:08](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=668s) 构筑责任底座与全局治理护栏 在讨论企业 AI 落地时，James Briggs 回顾道，几年前的行业峰会上“负责任的 AI（Responsible AI）”是一个无处不在的关键词，而如今这一话题似乎在热烈的技术讨论中有所淡化。然而，面对生成式 AI 的强大能力，企业必须把“负责任”作为最基础的第一步，并在其之上层层构建防护栏（Guardrails）。护栏不仅是管理输入和输出模型的数据内容，更要从全局视角把控：在业务人员提出用例定义的最初阶段，就需要验证其规划是否符合企业的核心价值观与既定准则。AI 治理绝非“设置好就一劳永逸（set it and forget it）”，而是一个需要持续校验和动态追踪的过程。 当首席执行官（CEO）介入并询问“我们在推进执行的同时，如何对所有这些 AI 项目进行全面监管？”时，企业需要一个集中的平台与全局治理（Global Governance）体系来实现跨组织的统一可见性。James 建议，企业应借助市场上现有的工具平台与合作伙伴，从统一的顶层视角跟踪所有 AI 倡议，而不是让各个项目在孤岛中自行其是。 ```mermaid flowchart TD accTitle: 企业级负责任 AI 治理三层架构 subgraph S1["第一层：核心价值观"] A["负责任 AI 基础准则<br/>(Responsible AI Foundation)"] end subgraph S2["第二层：全周期护栏"] B["用例定义与业务对齐"] --> C["数据出入模型管控与合规校验"] end subgraph S3["第三层：全局治理平台"] D["统一管理平台与可见性面板<br/>(Centralized Oversight & Governance)"] end A --> S2 S2 --> S3 ``` - **持续验证机制**：护栏机制必须深度嵌入用例的生命周期，从立项初期即评估业务契合度与伦理风险。 - **全局管理视角**：应对高层监管需求，必须打破单一项目壁垒，建立跨业务部门的集中式监控与追溯能力。 --- ### [13:18](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=798s) 内部沙箱隔离与客户信任协议 Payal Patel 针对安全落地分享了 Freshworks 的内部实践。为了防止员工不慎将敏感商业数据输入公共模型，他们在内部构建了名为“Cloudverse”的安全沙箱环境。员工可以在该受控环境中自由调用 Gemini、ChatGPT 等前沿大模型进行探索与测试，同时杜绝敏感数据外流的风险。除技术环境外，Freshworks 的安全团队还制定了严格的规章协议，并持续向全员推送最佳实践培训。 作为云解决方案提供商，Freshworks 同样将安全透明度延伸到客户侧。Payal 介绍了他们推出的“Freddy AI Trust”安全框架，该框架向企业客户清晰展示了数据从 Freshworks 平台流向 OpenAI 等底层模型供应商时所遵循的全部安全防护与脱敏协议，以此打消高管在采购评估时最关切的数据隐私顾虑。 - **Cloudverse 安全沙箱**：提供主流前沿模型的内部安全测试入口，物理级隔离敏感数据外泄通道。 - **Freddy AI Trust**：面向外部客户解构数据链路与加密协议，建立企业级 AI 产品的信任基础。 --- ### [14:42](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=882s) 规避“一刀切”禁令：从数据泄露事故中汲取的治理经验 当被问及成功案例时，Reena Solomon 选择分享其多年前在另一家公司经历的真实失败案例，因为“从失败中学到的往往比成功更多”。在生成式 AI 刚爆发的早期阶段，各业务部门（包括营销、工程研发及软件开发等）对 ChatGPT 提升生产力的潜力充满热情，但当时企业尚未建立完备的护栏机制。很快，警报被拉响——员工在不知不觉中将专有源代码和机密邮件内容直接粘贴进了公共工具中。 面对重大数据泄露事件，该公司的第一反应是全面禁用所有生成式 AI 工具。Reena 指出，这种“一刀切”的封杀是错误的应对方式，极大地阻碍了企业的技术演进。员工对提升生产力的工具具有刚性需求，粗暴封禁只会适得其反。在加入 Airbus（空中客车）后，团队采取了截然不同的治理思路：在每个业务部门派驻专门的 AI 工作组（Task Force），动态跟踪技术前沿与工具生态，为主流工具配置合规安全的外壳，同时对员工开展持续宣教，明确指出哪些操作可行、哪些行为坚决禁止，从而在确保安全合规的前提下释放生产力。 - **非恶意违规**：早期数据泄露往往源于员工缺乏正规操作指引与安全意识，而非主观恶意。 - **嵌入式工作组模式**：在各业务单元派驻专属任务组，兼顾前沿技术采纳与日常操作规范宣导。 --- ### [17:42](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1062s) 核心能力内化与初创生态协作 Lutz Beck 谈到了 Daimler Truck（戴姆勒卡车）在实际落地中的阶段性成果与战略定力。除了让自动驾驶卡车在德克萨斯州公路上安全行驶等硬核业务突破外，Lutz 认为更深层次的成功在于“将 AI 核心知识与编排能力收归企业内部”。在大模型热潮下，市场上涌现出大量宣称“无所不能”的外部服务商，但如果企业完全依赖外部供应商，很快就会迷失方向。 Lutz 特别强调了数据隐私条款中的隐形陷阱：许多现成的 AI Agent 在服务协议中用隐蔽的细则注明“有权使用客户数据进行模型训练”，这是大型制造企业绝对无法接受的。因此，Daimler Truck 选择自主开发定制化的内部 Agent 来处理核心敏感数据。在外部合作方面，他们更倾向于同具有高度创新性的初创公司合作，而非传统的大型外包机构。与此同时，推进售后（Aftermarket）与销售等高价值用例时，管理层必须正视员工对于“AI 是否会取代工作”的普遍恐惧，通过合理的组织引导消除顾虑，避免员工因防备心理而拒绝沉淀和共享业务知识。 - **能力自研与编排**：拒绝被大型外部集成商绑架，将数据流转控制权与 Agent 开发能力沉淀在内部团队。 - **警惕数据反哺条款**：严格审查第三方 Agent 的服务条款（T&C），防止核心工业数据被用于外部模型迭代。 - **拥抱初创团队**：相比传统大型厂商，更侧重与灵活敏捷的创新型初创企业展开针对性合作。 --- ### [20:36](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1236s) 成功因人而异：把重心前置到需求发现阶段 针对企业高管在 AI 项目中的认知盲区，James Briggs 提出了一个核心观点：“AI 的成功在很大程度上取决于观察者的视角（AI success is in the eye of the beholder）”。在给一个项目判定成功或失败之前，必须结合该企业在 AI 成熟度曲线上所处的具体发展阶段。 当前许多企业陷入了一种急于求成的误区，团队上下都在高喊“我们必须赶紧上线试点（Pilot）”，试图跳过基础步骤直奔成果。James 强调，与其盲目冲刺试点，不如将精力集中在早期的“需求发现（Discovery）”流程中。一个扎实的发现阶段能够把所有核心利益相关者（Stakeholders）召集到同一张谈判桌前，充分达成共识并取得各方的真正认同（Buy-in），这才是决定后续项目能否平稳推进的根本前提。 ### [21:36](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1296s) 需求发现阶段与概念验证（PoC）的关键作用 在规划任何 AI 举措时，明确无论对内还是对外的真实需求与预期结果是决定项目成败的核心前提。很多组织反复犯下的一个典型错误是：跳过深入的业务调研与利益相关者沟通，直接盲目斥巨资采购或构建一套庞大的大型系统，结果因为没有真正切中一线人员的需求，最终落得无人问津、彻底闲置的尴尬下场。 要从根本上提高成功概率并规避失败，关键在于扎实投资于前期的“需求发现阶段”（Discovery Phase）。将所有关键利益相关者召集到一张桌子前，逐一梳理他们的日常工作流，深入了解他们对业务价值的认知、目前觉得哪些工具有用或难用。基于这些真实洞察先落地一个小型的概念验证（PoC），在 PoC 阶段赢取大家的信任与支持（Buy-in），随后再将其顺理成章地扩展为更庞大的业务倡议。当真正到了全员上线推广的节点，一线员工早已深度参与其中，整体落地的成功率自然会呈指数级提升。 ```mermaid flowchart LR accTitle: AI 项目从需求发现到规模化落地的实施路径 subgraph S1["前期对齐"] A["召集核心利益相关者"] --> B["需求发现与痛点梳理"] end subgraph S2["验证闭环"] B --> C["构建小步快跑 PoC"] C --> D["获取各方深度认同"] end subgraph S3["全员扩展"] D --> E["拓展为大型业务举措"] E --> F["全面部署与高效采纳"] end ``` --- ### [23:21](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1401s) 内部认同、小步快跑与 AI 时代的新 KPI 指标 对于组织在规模化应用 AI 过程中面临的领导力与文化挑战，Freshworks 北美解决方案工程负责人 Payal Patel 强调，无论是企业对内赋能还是面向外部客户，最大的关卡始终是获取利益相关者的“认同感”（Buy-in）。如果在评估和引入新工具的最初阶段，实际使用者没有参与决策，看不到工具如何切实减轻日常负担，实际采纳率就会趋近于零。 面对许多客户“不知该从何下手”的普遍迷茫，盲目求大往往会适得其反。推荐的策略是“以小博大”——从一个能带来显著影响的具体微小场景切入。在 Freshworks 内部，团队推出了专门的“AI 咨询计划”（AI Advisory Program），帮助客户深入剖析现有业务流程，精准定位 AI 能够接管并缩减工时的具体环节。同时，传统的业务 KPI 已无法有效衡量 AI 的真实价值，企业必须设计与 AI 举措深度绑定的新型指标，并在持续迭代中监控这些核心指标的变化趋势，以此验证 AI 是否真正解决了业务痛点。 - **由小切入（Start Small, Big Impact）**：优先挑选能快速产生显著收益且容易凝聚人心的轻量级场景，避免一开始就铺开庞大架构。 - **AI 咨询计划（AI Advisory Program）**：协助团队与客户梳理现有流程，明确 AI 切入点，以大幅缩减传统操作所需的时间。 - **重构衡量指标（Redefining KPIs）**：传统业务 KPI 难以适配 AI 创新，企业需要制定针对 AI 特性的迭代型衡量标准，追踪效率提升与价值转化。 --- ### [25:23](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1523s) 空中客车（Airbus）应对员工 AI 焦虑与全面技能赋能 空中客车（Airbus）AI 合作伙伴负责人 Reena Solomon 坦言，文化转变与技术推广中最现实的障碍之一是员工对“AI 会取代我工作”的恐惧与抵触心理。如果无法从根本上转变全员的思维认知，即使管理层投入天价预算采购最顶尖的工具，最终也只会沦为巨大的资源浪费。 为了破除这种心理壁垒，空中客车采取了“全员通识与技能重塑”（Upskilling）的推进策略。这种培训不仅仅是讲解市面上有哪些 AI 工具或技术参数，而是聚焦于解释“为什么 AI 如此重要”、“为什么每个人都必须关注它”。从高管层到一线工程师再到职能部门，全面的通识教育让员工深刻意识到，AI 工具正在变得像当年的个人电脑（PC）一样不可或缺。领导层应当亲自“牵着员工的手”完成这趟转型之旅，通过赋能与技能提升化解恐惧，让员工转变为拥抱工具的受益者。 > “AI 工具正在变得如同过去彻底改变我们工作方式的计算机一样重要且不可或缺……请投资于你的员工，为了组织的长远发展对他们进行技能升级。” —— Reena Solomon --- ### [27:10](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1630s) 领导者个人生产力利器大盘点：从 NotebookLM 到企业级 AI Agent 主持人 Noa Davidson 随后向各位嘉宾抛出了一个颇具个人色彩的互动问题：作为大型组织的领导者，大家在个人工作或日常生活中都在使用哪些具体的 AI 生产力工具？ 各位嘉宾分享了各自的核心工具栈与使用场景： - **Reena Solomon（空中客车）**：笑称自己“注意力跨度堪比金鱼”，已经完全告别了枯燥翻阅传统论文期刊的习惯。她目前极为推崇 Google 的 **NotebookLM**，将技术文档或资料转化为互动式播客进行听力消化，并能直接针对特定内容进行精准提问。 - **James Briggs（AI Collaborator）**：得益于专属数据科学团队搭建的安全护栏（Guardrails），他在日常工作中广泛尝试与测试各种前沿的多模态大语言模型（LLMs），根据特定任务灵活切换。 - **Payal Patel（Freshworks）**：公司全面采用 Google 生态，她高度依赖基于大模型的工具辅助理解长篇资料、梳理会议纪要，并将零散的笔记一键生成结构化的演讲演示文稿构思，极大缓解了整天开会挤占独立工作时间的压力。 - **Lutz Beck（戴姆勒卡车北美）**：日常工作深度基于微软生态，全面使用 **Microsoft Copilot** 处理会议摘要与大文件精读；出于数据安全合规原因内部屏蔽了公开版 ChatGPT，转而采用 **LibreChat** 开源方案；同时正在推进自研的自主智能体（Agent）概念验证，目标是在年底前将所有的日常非增值事务（如繁琐的流程审批）全部移交给 Agent 自动化触发与执行。 - **Noa Davidson（FPV Ventures）**：分享了自己近期深度试用的自动化侦察与监控工具（如 Scouts 测试版），用于自动追踪被投企业的招聘动态、业务增长、人事变动，甚至自动监控并锁定热门餐厅的预订。 --- ### [31:29](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1889s) 现场观众提问：当 AI 项目未能立即兑现预期价值时如何重建业务信任 在小组讨论进入现场问答环节后，一位观众提出了一个极具共鸣的现实挑战：当业务部门（尤其是销售、营销等直接背负业绩压力的团队）开始满怀期待地推进 AI 用例，但 AI 工具上线后却并未能“立竿见影”地兑现最初承诺的商业价值时，业务团队对 AI 的信任往往会迅速崩塌甚至开始质疑技术路线。观众迫切希望了解，技术与创新领导者此时应当采取怎样的应对策略，既能稳住各方情绪，又能带着修正后的新方法重新向业务端证明 AI 的真正商业价值。 ### [32:10](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1930s) 企业落地 AI 的核心：用户参与、工作流重构与 ROI 在 AI 探索阶段，一旦企业在初期遭遇挫折，往往很难获得第二次机会重新站稳脚跟向前推进。因此，在项目初期建立完善的“发现阶段”（Discovery Phase）至关重要，只有这样才能提早争取到利益相关者的支持（Stakeholder Buy-in）。当前企业在 AI 落地上面临着显著的应用与采纳阻力；员工对这项技术的潜在危害和不确定性心存疑虑，但不可否认的是，AI 带来的实际效益是巨大的。 要真正让业务团队采纳 AI，必须邀请他们参与到前期的定义过程中。企业不仅要关注底层的模型本身，更要关注模型所嵌入的业务应用，以及这些应用如何深度重构员工的实际工作流（Workflows）。如果工具不能真正优化流程、为员工节省宝贵的时间或资金成本，他们就很难买账。没有前期的深度协同，后续的所有设计都只是盲目猜测；即便研发出的应用在功能上完全达到了预期，若无法真正融入业务流程并被员工接纳，企业就绝不可能获得预期的投资回报率（ROI）。 ```mermaid flowchart TD accTitle: AI 企业落地与工作流重构闭环 subgraph Phase1 ["前期探索与对齐 (Discovery Phase)"] A["业务与技术利益相关者入席"] --> B["识别工作流痛点与真实价值"] end subgraph Phase2 ["应用与流程深度融合"] B --> C["设计嵌入式 AI 应用而非单纯模型"] C --> D["优化业务流程与返还时间/成本"] end subgraph Phase3 ["业务采纳与价值实现"] D --> E["业务团队主动采纳与推进"] E --> F["实现预期 ROI 与业务变革"] end ``` --- ### [33:52](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2032s) 行业全面重构与去 ERP 化：客观应对泡沫与董事会预期 无论身处哪一个行业，未来 5 到 10 年内都必须全面拥抱 AI——甚至可以说“10 年都已经太长了，不拥抱就会被淘汰出局（out of business）”。正确实施 AI 将彻底颠覆现有的 IT 架构与技术格局。今天普遍运行的传统 ERP 系统在某种意义上已经沦为历史；虽然 ERP 厂商也在努力叠加 AI，但在许多业务流程中，企业已不再需要它们。例如，在戴姆勒北美卡车（Daimler Truck North America）的供应链和采购业务中，团队完全依靠内部自研的 AI 工具实现了 90% 的流程自动化，全程没有依赖任何 ERP 系统。 然而，企业在拥抱 AI 的过程中必然会遭遇失败，这往往会引发对 AI 是否名副其实的质疑。这种质疑很大程度上源自当前市场上普遍存在的过度炒作（Hype）——外界盲目认为 AI 可以包治百病，但现实并非如此。企业必须正视残酷的数据治理现实：行业中能直接适配 AI 解决方案的高质量数据，乐观估计也只有 50% 左右，剩下 50% 根本无法使用。 尽管数据治理任重道远，但时间不等人。预计到 2030 年至 2040 年，企业架构中多达 70% 的工作将由 AI 工具承载；届时进入职场的“AI 原住民”（AI Natives）将完全依赖 AI 工具工作，绝不会再使用老旧系统。科技行业在此过程中存在过度承诺（Over-promising）的通病，这拉高了董事会的盲目预期，甚至让董事会误以为 AI 能轻易裁员 40% 或立刻带来数百万美元的节省。技术领导者必须通过严谨的发现阶段去跟踪实际业务价值；如果某项技术的收益在短期内无法兑现，应果断搁置，并与董事会展开坦诚透彻的沟通。同时，企业应致力于将核心技能重新收归内部（In-house），摆脱外部供应商画大饼的困境，步步为营地推进智能化转型。 - **去 ERP 自动化实例**：供应链/采购关键环节实现 90% 纯 AI 工具自动化。 - **数据质量现状**：企业内部仅有约 50% 的数据可用，另外 50% 无法直接用于 AI。 - **未来架构趋势（2030–2040）**：预计 70% 的企业 IT 架构与工作负载将由 AI 驱动。 - **劳动力代际更替**：未来的 AI 原住民员工将完全依赖 AI 原生工具开展工作。 - **破除裁员迷思**：纠正董事会关于“AI 立即裁员 40% 并大幅缩减开支”的不切实际预期。 --- ### [37:58](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2278s) 利益相关者映射与自下而上的 Beta 试点策略 为了确保 AI 项目的顺利推进，企业必须建立强有力的利益相关者映射机制（Stakeholder Mapping）。例如借鉴达美航空（Delta Air Lines）创新中心“The Hangar”的经验，主动发掘各业务条线中那些充满热情、渴望学习并乐于参与其中的员工。如果推行的是销售相关的流程与工具，就必须将销售团队的核心成员提前拉入项目。 这些先锋员工能够在早期阶段提供宝贵的第一手反馈，并作为内部代表向整个团队引荐 Beta 试点版本，从而避免生硬的自上而下式强推。没有人喜欢纯粹由上至下的行政命令，AI 落地必须兼顾“自下而上”（Bottom-up）的路径。在推行过程中，运用正确的沟通话术至关重要：强调“我们都在同一条船上”、“AI 是为了让大家的工作变得更轻松”。这种赋能的姿态能促使员工积极提出建设性的改进意见，而不是产生抵触情绪或断言“工具根本不起作用”。 --- ### [38:50](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2330s) 选型困局：中小企业如何辨析厂商能力与实施成本 在问答环节，一位来自阿肯色州（Arkansas）的小企业主抛出了一个普遍的选型困境：当前 AI 市场充斥着形形色色的供应商，报价从 20 万美元、50 万美元到上百万美元不等，均宣称能彻底重塑运输、售后维修、零配件管理或销售部门。面对铺天盖地的 Agent 策略、RAG 以及生成式 AI（GenAI）营销概念，小企业该如何辨识谁才是真正的一流厂商？ 针对这一痛点，嘉宾们给出了多维度的评估建议： 一方面，企业需要持续监控 AI 生态格局的发展。当前市场的演化规律是：一旦某家初创企业开发出卓越的单点功能，大型科技巨头往往会迅速跟进复制该功能并将其整合进自家的产品套件中，导致初创公司的功能优势迅速被稀释。因此，许多大企业会采取“双重押注”（Dual Bet）策略：一方面采购大型平台以获取能够兜底的端到端安全方案，另一方面试用有潜力的高成长初创企业。在面对 CB Insights 等机构描绘的上百家初创公司时，企业应重点聚焦那些在投资势头、客户增长及技术突破上具备实质动能的厂商，并深入开展沟通以验证其落地案例。 另一方面，来自 Freshworks 的代表指出，在评估供应商时，切勿被“AI 蓝图”所迷惑，真正决定成败的是实施与运维难度： 1. **实施难度**（How hard is it to implement） 2. **配置复杂度**（How hard is it to configure） 3. **长期维护成本**（How hard is it to maintain） 中小企业在对抗 ServiceNow 或 Salesforce 等行业巨头时，往往更看重 AI 的“去繁就简”（Uncomplicated AI）——即无需漫长复杂的定制开发，只需一键开启开关（Toggle it on），系统便能自动从业务历史数据中持续学习并适应特定的垂直业务场景（如运输调度等）。企业在做决策时，应以此类易用性和可维护性作为核心衡量指标。 ### [42:18](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2538s) 定制化 Copilot、沙箱评估与供应商 POC 落地成本 当下大语言模型层出不穷，单纯部署一个现成的 Copilot 已经不再具备技术壁垒，“人人都能做 Copilot”，但真正拉开差距的关键，在于能否为企业自身业务量身打造定制化方案，并切实改善员工的日常工作流程。针对企业高管在这一转型阶段可能面临的困惑与支持缺口，拥有风险投资背景的嘉宾风趣地表示“你身旁友好的风险投资机构随时愿意提供帮助”。同时，回顾在亚马逊（Amazon）与可口可乐（Coca-Cola）的实战经历，搭建高效完备的“沙箱”（Sandbox）测试环境是验证技术可行性的生命线。企业必须清晰掌握供应商进行一次演示（Demo）或概念验证（Pilot）究竟需要投入多少时间、资源与预算，进而精确核算其投资回报率与业务增值。 对于在场的 AI 方案服务商，嘉宾给出了极具实操性的建议：务必在商业计划书或路演 Deck 中单独加入一页，明确说明项目启动需要做哪些准备、部署一套演示环境或 Pilot 的确切成本是多少，以及后续若要扩展至大规模生产环境需要哪些前置条件。没有哪家企业愿意在看不清回报前景的情况下，贸然投入数年时间与数百万美元的资金。 - **定制化落地核心**：通用型大模型泛滥的背景下，评估价值的标准是工具是否“bespoke”（量身定制）并真正融入员工日常业务场景。 - **沙箱机制与 ROI 测算**：通过受控的沙箱环境摸清服务商的技术门槛与部署成本，算清价值账后再决定是否进行规模化采购。 - **对供应商的演示建议**：在方案推介中主动披露部署流程、时间周期及概念验证预算，减少甲方的决策不确定性。 ### [43:16](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2596s) 告别单体束缚：模块化架构与“反厂商锁定”底线 从大型企业 CIO 的视角来看，如今买方市场正处于一个绝佳的风口期：市场上每天都在涌现更优秀的 AI 工具与新方案，技术迭代日新月异。然而，要真正享受这种技术红利，企业 IT 架构必须彻底告别传统的单体式（Monolithic）架构——例如僵化固化的老一代 ERP 系统——转向高度集成与模块化（Modular）的架构设计。这种架构的核心优势在于“即插即用”：当某一项技术过时或出现更好的替代品时，企业可以随时拔掉现存组件，迅速接入全新模块。 在选型与集成评估中，评判标准不仅是“接入有多容易”，更关键的是“把它从系统里剥离替换出来有多容易”。嘉宾斩钉截铁地强调，任何带有厂商锁定（Vendor Lock-in）倾向的技术方案都是绝对的禁区：“一旦我看到有锁定绑定（Lock-in），那就是一票否决（No-go）。”即便某家供应商的产品在当下看起来非常惊艳，只要存在架构绑架风险，企业就绝不会采购。企业必须保持全局的自主可控，坚持“Best-of-Breed”（博采众长）的集成策略，随时替换更优解。 ```mermaid flowchart TD accTitle: 模块化企业AI集成决策流程 subgraph S1["架构设计与选型原则"] A["评估新技术方案"] --> B{"是否存在厂商锁定 (Vendor Lock-in)?"} B -- "是 (绑定私有生态)" --> C["一票否决 (No-go) 终止合作"] B -- "否 (开放接口/标准协议)" --> D["进入沙箱环境测试"] end subgraph S2["即插即用实施"] D --> E["轻量化集成至核心系统"] E --> F["模块化插拔: 随技术演进自由替换"] end ``` ### [44:26](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2666s) 破解资源瓶颈：借力生态伙伴打造协同落地网络 探讨的终局落脚于“协同合作”（Collaboration）。正如 AI Collaborator 公司的创立初衷所揭示的客观现实：没有任何一家企业能够凭一己之力拥有足够的资源去解决 AI 领域的所有问题。即便是全球顶尖规模的巨头跨国企业，在日常的 AI 实践中也反复印证着这一真理。面对汹涌的技术浪潮，单打独斗注定走不通，企业必须找到能够协助统筹、把关与管理交付的顶级外部伙伴。 建立长期的生态合作伙伴网络至关重要。以深耕 AI 解决方案与服务商网络八年以上的行业经验为例，专业协作平台所扮演的角色，就是替大型企业提前完成严格的背景审查、技术评估与信任背书，并将最匹配的优质服务商带到谈判桌前，协同企业高效完成场景部署。在这个瞬息万变的 AI 时代，善于借助生态合力、选对合作伙伴，是企业不仅实现成功转型、更是立足长远得以生存下去的核心生存法则。 - **资源与能力现实**：没有企业能拥有无限预算与自研团队解决所有 AI 场景，外借杠杆是大型企业的必修课。 - **合作伙伴筛选机制**：借助专业三方网络对初创和垂直服务商进行前置 vetting（审查筛选），有效降低企业自研与采购试错成本。 - **生存法则**：在技术极速裂变的周期里，“协同与生态伙伴关系”（Collaboration and partners）是企业保持敏捷与持续竞争力的关键支柱。 ## Notable quotes > [07:50](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=470s) 在不断发生地质位移的地基上盖摩天大楼，而且建筑设计图纸还在随时被修改 — Reena Solomon > [11:08](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=668s) AI 治理绝非‘设置好就一劳永逸（set it and forget it）’，而是一个需要持续校验和动态追踪的过程 — James Briggs > [14:42](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=882s) 从失败中学到的往往比成功更多 — Reena Solomon > [17:42](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1062s) 将 AI 核心知识与编排能力收归企业内部 — Lutz Beck > [20:36](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1236s) AI 的成功在很大程度上取决于观察者的视角（AI success is in the eye of the beholder） — James Briggs > [25:23](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=1523s) AI 工具正在变得如同过去彻底改变我们工作方式的计算机一样重要且不可或缺……请投资于你的员工，为了组织的长远发展对他们进行技能升级。 — Reena Solomon > [33:52](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2032s) 10 年都已经太长了，不拥抱就会被淘汰出局（out of business） — Lutz Beck > [37:58](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2278s) 我们都在同一条船上，AI 是为了让大家的工作变得更轻松 — James Briggs > [42:18](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2538s) 人人都能做 Copilot — Noa Davidson > [43:16](https://www.youtube.com/watch?v=DlWoJt3TcBI&t=2596s) 一旦我看到有锁定绑定（Lock-in），那就是一票否决（No-go） — Lutz Beck ## Mentioned Noa Davidson, Payal Patel, Lutz Beck, Reena Solomon, James Briggs, FPV Ventures, Freshworks, Daimler Truck North America, Airbus, AI Collaborator, Inc., Acubed, Coca-Cola, Samsung, Delta Air Lines, Amazon, ServiceNow, Salesforce, OpenAI, Google, Microsoft, ChatGPT, Gemini, Microsoft Copilot, NotebookLM, LibreChat, Freddy AI Trust, Cloudverse, CB Insights ## Video description Noa Davidson, Operating Partner, FPV Ventures Payal Patel, Head of North America Solution Engineering, Freshworks Lutz Beck, CIO, Daimler Truck North America Reena Solomon, Head of AI Partnerships, Airbus James Briggs, CEO, AI Collaborator, Inc. Abstracts Successfully leading an AI initiative requires a unique blend of strategic vision, technical understanding, and organizational alignment. This session will explore what it takes for AI leaders—CAIOs, CIOs, and IT leaders—to guide AI projects from inception to full implementation. Panelists will discuss the key leadership skills required to oversee the entire AI lifecycle, from setting clear business objectives to ensuring seamless integration and scaling. Learn how to navigate challenges such as cross-functional collaboration, managing resources, and fostering a culture of innovation while driving impactful AI transformation. [Source](https://www.youtube.com/watch?v=DlWoJt3TcBI)

[7:50] Reena Solomon: 在不断发生地质位移的地基上盖摩天大楼，而且建筑设计图纸还在随时被修改 [11:08] James Briggs: AI 治理绝非‘设置好就一劳永逸（set it and forget it）’，而是一个需要持续校验和动态追踪的过程 [14:42] Reena Solomon: 从失败中学到的往往比成功更多 [17:42] Lutz Beck: 将 AI 核心知识与编排能力收归企业内部 [20:36] James Briggs: AI 的成功在很大程度上取决于观察者的视角（AI success is in the eye of the beholder） [25:23] Reena Solomon: AI 工具正在变得如同过去彻底改变我们工作方式的计算机一样重要且不可或缺……请投资于你的员工，为了组织的长远发展对他们进行技能升级。 [33:52] Lutz Beck: 10 年都已经太长了，不拥抱就会被淘汰出局（out of business） [37:58] James Briggs: 我们都在同一条船上，AI 是为了让大家的工作变得更轻松 [42:18] Noa Davidson: 人人都能做 Copilot [43:16] Lutz Beck: 一旦我看到有锁定绑定（Lock-in），那就是一票否决（No-go）

[3:03] James Briggs: are not going to be with you for long. [7:50] Reena Solomon: trying to build a skyscraper when the foundation is shifting on an ongoing basis and the blueprints for your building are changing as well. [14:42] Reena Solomon: Employees are going to try and find ways to be productive and use the tools that they need. So the best thing for an organization to do is to make sure you have those safe guardrails in place and provide these tools in a compliant and a safe and secure managed manner. [20:36] James Briggs: AI success is in the eye of the beholder [20:36] James Briggs: we got to get to pilot, we got to get to pilot [27:10] Reena Solomon: the attention span of a goldfish [37:58] make your life easier [38:50] Audience Member: From a small business standpoint, pricing, right? We're getting quotes of a million dollars: 'Hey, this will transform your transportation department, this will transform your service department, your parts department, your sales organization.' We're getting a price of 500,000, 200,000. How do you navigate to know which companies are the best at this? Which companies are the best at agent strategy... which companies are the best at RAG, which companies are the best at GenAI? [43:16] Lutz Beck: As soon as I see a lock-in, it's a no-go. [44:28] James Briggs: no business is ever going to have enough resources to do everything that they want to do to solve for AI.

[0:04] Um, we are so happy to have you all for [0:06] this panel. It's always wonderful to be [0:08] with leaders at the forefront of AI [0:11] innovation and we have an incredible [0:13] group of panelists this evening who are [0:15] here to talk about their experiences and [0:17] help guide you through your journeys. [0:19] Um, so I'm Noah Davidson. I'm an [0:22] operating partner at FPV Ventures, which [0:24] is a billion-dollar AUM venture capital [0:27] firm based in San Francisco. We focus [0:29] predominantly on series A and B [0:31] companies. So if you guys have any great [0:34] companies, feel free to find me at the [0:35] end and or send me a note on LinkedIn. [0:38] Um I want to introduce our panelists and [0:40] if you guys want to go down the line and [0:42] say a little bit about yourselves. [0:43] >> Uh Lutzbeck. I'm the uh CIO of Daimler [0:47] Truck North America. [0:49] um since quite some time here in the in [0:52] the US and working similarly I'm uh [0:56] having CXO seats at venture capital [0:59] companies so I'm looking also always for [1:02] new and exciting startups um and of [1:05] course also for for our work which we're [1:08] doing at DNA truck you can imagine AI is [1:11] for us we're working on autonomous [1:13] trucking since 7 years and this is [1:16] something which we are having in in our [1:19] heart and we we push a lot in in that [1:22] area. [1:24] >> Hi everybody. Uh my name is Pile Patel. [1:28] If you're ever wondering how to say it, [1:29] it's like a pile of cash. It's really [1:31] easy for everybody to remember. Um I I'm [1:34] at Freshworks. I've been there for about [1:36] 6 and a half years. I lead the solution [1:38] engineering org across Americas. Uh my [1:42] team basically works with customers to [1:43] help them identify pain points or [1:45] challenges that we can solve with our [1:47] customer experience and employee [1:49] experience solutions. I will uh looking [1:52] forward to the conversation. I'll pass [1:53] it on. [1:56] >> Hi, I'm James Briggs, CEO of AI [1:59] Collaborator. Uh we help enterprise [2:01] firms manage the surge in AI demand that [2:04] they're that's currently overwhelming [2:06] their IT organizations. Uh we do that [2:09] through purpose-built innovation such as [2:11] an Agentic AI really agentic AI first of [2:14] its kind demand management platform and [2:17] then a global network of AI solutions [2:19] and services providers. Uh what we like [2:22] to say is that we help companies build [2:24] AI agility. Looking forward to the [2:26] conversation. [2:30] >> Hello everyone. Good morning. Uh I am [2:32] Reena Solomon. Uh I'm from Airbus. I [2:36] lead AI transformation uh at Airbus [2:39] through partnerships. I am based in [2:41] their uh Silicon Valley innovation [2:44] center uh called Aube. Uh I've been in [2:47] this industry for a while. Before Airbus [2:49] uh I was at Samsung and few other [2:53] technology companies focusing on um how [2:56] AI is transforming uh the sector. I'm [2:59] looking forward to the conversation [3:01] today. [3:03] >> Amazing. Thank you everyone. So many of [3:05] you in this room are probably the victim [3:07] of receiving a lot of emails from [3:09] co-workers, other seauite leaders, your [3:11] family. What are you doing with AI in [3:13] your organization? Um I led Coca-Cola's [3:16] global accelerator for many years and [3:18] landed with robots at my desk, all kinds [3:20] of things. And so it's really a [3:23] interesting part of the conversation to [3:25] have of where do you actually start? Um [3:27] and so James, I just want you to kick it [3:29] off. Um where where do you start with [3:31] prioritizing AI initiatives? um and how [3:33] should leaders start thinking about it? [3:37] >> Really what we've seen over the really [3:39] the last five years is that uh where AI [3:42] sits in the organization uh inside of an [3:45] enterprise is majority of companies at [3:47] this point have AI dedicated and focused [3:50] towards their products and do a great [3:52] job of getting AI to those core pro [3:55] products but when you get outside uh of [3:58] those core most other managers are [4:00] sitting in AI deserts. What we recommend [4:03] is now starting with getting AI into the [4:07] hands of the managers that deliver [4:10] inside of the organization. So that's [4:12] really the starting point. And so from [4:14] there then you can really start to talk [4:16] about how do you prioritize within that [4:18] section but uh it's going beyond the [4:23] core products getting it to the managers [4:26] cuz understanding that if you're not [4:28] giving those managers an opportunity to [4:31] put this stuff in their workflows [4:33] they're not going to be with you for [4:34] long. That's our that's my [4:36] recommendation. [4:37] >> Amazing. Kyle, do you have anything to [4:38] add? [4:39] Yeah, I think one thing that we've [4:41] noticed as we work with customers is [4:43] kind of partnering with them to [4:44] understand the strategic initiatives [4:46] inside the organization that will lead [4:49] to some form of business outcome. A lot [4:51] of times AI strategies don't really get [4:54] the ROI or the realization unless you [4:56] are creating a matrix of what that [4:59] capability is going to do to the [5:00] individual that's going to leverage that [5:02] capability. So the best way to do that [5:04] and I think it's some of the things that [5:06] were just said here is work with [5:07] stakeholders on what should be [5:10] prioritized and make sure executives [5:12] above also are matching those [5:15] capabilities to strategic initiatives [5:17] within the organization and that can [5:19] kind of help you kind of deliver the ROI [5:21] of AI. It can help you explain the [5:23] benefit and the value as you bring in AI [5:26] solutions into your organization. [5:29] >> Thank you. Let's say as the CIO of a [5:31] large global organization, how is your [5:33] organization structured to support AI? [5:36] I mean we we are having a global [5:39] structure in place um where we have [5:41] globally let's say um I don't like the [5:44] word but some kind of a governance uh uh [5:47] in structure where we say we have we [5:49] have a and analytics board which is [5:53] globally and then which is in the [5:54] regions as well uh where we have that uh [5:57] and where we basically go exactly what P [6:00] said we we basically look at the value [6:03] where are the value drivers what do we [6:05] need to do and how we need to do it. Uh [6:07] and then of course we have um chief data [6:10] officers in place which are doing the [6:12] data provisioning. Uh they are basically [6:15] they're they're responsible really to [6:17] look at the data looking at the data [6:19] quality because I mean um as you can [6:21] imagine as a [6:24] old company in a certain sense right our [6:27] data is not necessarily the one which [6:29] you should use to do AI in in some [6:31] senses and we need to work on that on [6:33] the data quality. So this is the chief [6:35] data offices. We have that regionally as [6:38] well. And then we have basically our [6:41] delivery teams on that. Um we have a big [6:44] delivery center in India with uh roughly [6:46] 200 people working on AI. And then uh we [6:50] have in the regions as well the whole [6:51] topic in and then what is also [6:54] additionally in a certain sense we have [6:56] a group here in the US which is called [6:58] hack. This is all people in within the [7:01] company which are interested in AI and [7:04] data topics. Uh and uh it's around 900 [7:08] people uh and we are using this group as [7:11] well to to educate and skill up people [7:14] in the organization to work on it. Um [7:17] and the biggest topic is of course the [7:19] board our board globally but also the [7:21] board here sits on everything what we're [7:24] doing in in the AI space. So everything [7:26] what we have in terms of prioritization [7:28] the portfolio which we are doing is [7:31] agreed with the board uh because at the [7:34] end of the day we need to look in I mean [7:35] you can play around with AI and [7:37] everybody knows that uh there is a lot [7:39] of things you can do and they add no [7:41] value at all uh so we really looking at [7:44] the high value cases and and prioritize [7:48] them as well on our side. [7:50] >> Amazing. and Reena as head of AI for [7:52] Airbus, could you tell us a little bit [7:53] about how you're organized? [7:55] >> Absolutely. Um, I resonate very much [7:58] with what Litz mentioned. Uh, Airbus is [8:00] also a global company. So, we have [8:03] dedicated AI hubs. We have AI innovation [8:06] centers. We have uh data officers and [8:09] champions uh embedded in different [8:11] business units to accelerate adoption. [8:14] So, all of these are really critical. [8:16] But um I'd like to touch a point upon [8:19] one point which is important for the [8:21] structure as we start scaling AI uh in [8:25] an organization. [8:27] Uh I'm sure you all have this pain point [8:30] right AI is moving at a really fast [8:32] pace. You pretty much hear about [8:34] breakthroughs almost every day. Um and [8:37] you hear about new applications and [8:40] changes. So how do you uh scale an [8:43] organization AI in an organization when [8:45] things are moving so fast? Uh it's like [8:48] trying to build a skyscraper when the [8:50] foundation is shifting on an ongoing [8:52] basis and the blueprints for your um [8:55] building are changing as well. So it's a [8:58] huge challenge especially for Ebas where [9:00] uh safety and reliability are a key [9:03] concern and that's where um innovation [9:06] centers play a huge role. Uh for [9:09] example, Aub that's based in Silicon [9:12] Valley uh recruits local talent and the [9:15] idea of a cubed is to make sure we [9:18] conduct high risk and high reward [9:21] experiments outside the core business [9:23] for Airbus. So we do rapid prototype pip [9:27] ro prototyping proof of concepts [9:30] experiment with new technologies and [9:32] tools and once we have droed the [9:35] technology we understand that there's a [9:37] clear business value that's when we hand [9:40] it over and transition transition that [9:43] to the business units through a layered [9:45] approach. This actually helps Airbus [9:49] have an agile pace, be able to [9:52] experiment new technologies at the same [9:54] time be true to its core value for of [9:57] being safe and compliant in such a [10:01] highly regulated industry and we also [10:05] form have like a hybrid approach when it [10:08] comes to innovation. So through these [10:10] innovation centers we are able to do a [10:13] bottom-up innovation grassroot level uh [10:16] recommendations of new tools and [10:18] experiments [10:20] but when before it scales and is [10:22] actually adopted by a business unit it [10:25] goes through a central evaluation [10:26] process as well. And this has been [10:29] really critical for us to be able to [10:32] have this um experimentation mindset [10:36] while maintaining the reliability that's [10:38] needed for the industry. [10:41] >> Thank you. And let's we're about to use [10:43] your least favorite word, governance, [10:44] but it has to come up in this [10:45] environment. So um James, can you talk [10:48] about what governance or safeguards are [10:51] essential when deploying AI and how are [10:53] you balancing innovation and compliance? [10:58] What I suggest there is start with a [11:00] good foundation of responsible AI [11:02] practices within within the [11:04] organization. Uh that word was uh if you [11:08] go back a couple conference ago, a [11:09] couple years ago, you couldn't escape [11:12] responsible, responsible, responsible. [11:14] It seems to get a little bit lost now in [11:16] the conversation. But I think uh given [11:20] uh what these technologies are capable [11:22] of, you have to start with a solid [11:24] foundation around responsible as the [11:26] first thing that you do. And then on top [11:29] of that, you can then start to layer in [11:31] guardrails. Uh and and from a guardrail [11:34] perspective, that guardrails could be [11:36] things that uh help you to manage what [11:39] goes in and out of any models and things [11:42] like that. But you also have to think [11:43] guardrails, think about guardrails in a [11:45] broader perspective. uh what does that [11:47] mean when someone is at the beginning of [11:49] that use case definition? Uh how can you [11:52] validate that the things that they're [11:55] looking to do uh with the tool that they [11:58] ultimately are going to deploy meets [12:00] your business values and aligns with [12:02] those uh foundations that you've set [12:04] inside of the organization. So you have [12:06] to constantly be checking those [12:07] processes. Uh and you really can't uh [12:11] set it and forget it. [12:13] uh you have to constantly be looking at [12:16] the really the third pillar of all of [12:18] that is how do I have something help me [12:20] manage this stuff and give me global [12:23] governance across everything that we're [12:25] doing. When the CEO comes in and asks [12:28] you, yes, it's great that we're now [12:30] implementing and executing, but what are [12:33] we doing to have oversight over all this [12:36] stuff? You want to be able to have some [12:37] type of centralized platform, [12:39] centralized way of accessing that [12:41] information and have visibility and [12:44] really be able to get visibility across [12:45] the organizations. So those are the [12:48] steps that I would recommend. Uh and [12:50] there's definitely there's platforms, [12:52] there's tools that are coming to market [12:54] now that can help you do all of those [12:55] things. But start with that solid [12:58] foundation of responsible building guard [13:01] rails all along the way and then get [13:04] tools, get partners, uh get solutions in [13:08] place to help you track that stuff all [13:09] across everything that's going on in the [13:12] organization and not just across one [13:15] initiative or another. [13:18] >> Kyle, do you have anything that you want [13:19] to add to that? [13:20] >> Yeah, I think you've mentioned a lot of [13:22] really good things. One unique thing [13:24] that we did internally within our [13:26] organization is we created our own I'll [13:29] call it sandbox environment to be able [13:31] to do testing with um various [13:34] capabilities of AI. So for example we [13:37] call it cloud verse you can access [13:38] Gemini you can ask access chat GPT but [13:41] it's all secured within an environment [13:43] so that your employees are not putting [13:45] sensitive information where it shouldn't [13:47] belong. So that's one great way to think [13:49] about how do you secure the use of AI [13:51] within employees in your organization. [13:54] And then just like James said, we've [13:55] created a lot of protocols. We're [13:57] training our employees on best [13:59] practices. We send out a lot of [14:00] training. Our security team is big on [14:02] that, which a lot of organizations are [14:04] looking for. And because we're a a cloud [14:07] solution provider, we also do a lot of [14:09] educating to our customers of best [14:11] practices that we're taking in terms of [14:14] our product portfolio and making sure [14:16] that's secure as well. We created um a [14:19] couple of couple of things that allow [14:21] for security. We call it Freddy AI trust [14:24] and we explain all the protocols that [14:26] are there as data is going from our [14:28] platform to uh open AAI or which any any [14:31] other platform that we're using so that [14:33] we're educating customers as they're [14:35] evaluating our solution because that [14:36] becomes the top priority or the top [14:38] concern for many executives within an [14:41] organization. [14:42] >> Thank you. And Reena, you dove into the [14:45] incredible structure that your [14:47] organization has put into place. Could [14:49] you tell us about a success story and [14:50] some of the learnings? [14:52] >> Absolutely. If you don't mind, I would [14:55] talk like to talk about a failed project [14:57] instead because I feel like uh we learn [15:00] a lot more from failures than successes. [15:03] So just to be clear, um this failure or [15:06] this incident did not happen at Airbus. [15:08] It was another company that I worked [15:10] couple of years ago. [15:13] And uh um it was kind of the early times [15:17] of uh Genai and chat GPT. People were [15:20] excited about it and uh trying to [15:23] understand uh how they can use it in [15:26] improving productivity in their roles [15:27] everything from marketing to engineering [15:30] to software development etc. And uh it [15:33] was since it was early stages there [15:35] weren't many guardrails and they didn't [15:37] have the smart things that pile and uh [15:39] James spoke about and um [15:44] the alarm started going when we started [15:46] realizing that people were putting in [15:48] proprietary information such as co [15:50] source code confidential email data into [15:54] the emails. This is actually public [15:55] information. So if you're curious you [15:57] can go look at it. But the point is it [16:00] was not um something malicious. uh these [16:03] employees uh did not have the real [16:07] guidance to do the right thing and in [16:09] terms of handling these tools [16:11] appropriately and because of that there [16:14] was a huge data leak and the way the [16:17] company responded to this was like [16:20] saying hey we're going to take away all [16:21] these tools right and that was not the [16:23] right thing to do because it sets you [16:25] back so much and how do you actually [16:30] deal with it right and one of the key [16:32] things that we have been doing with in [16:34] Airbus is actually educating people on [16:38] what are the right ways to handle these [16:41] tools. We have dedicated task force [16:44] embedded in each business unit where [16:47] which puts in which understands what's [16:50] happening in the landscape. What are the [16:52] key tools that are available? How do we [16:55] make these safe for these employees and [16:57] puts the guard rails in place? But at [17:00] the same time, we also communicate to [17:02] the employees on an ongoing basis [17:04] saying, "Hey, this is the way you work [17:06] with these tools and these are the [17:08] things that you cannot do." And that has [17:10] actually helped us. So, in terms of [17:12] takeaway for us from that incident was [17:15] like employees are going to try and uh [17:18] find ways to be productive and use the [17:21] tools that they need. So the best thing [17:24] for an organization to do is to make [17:26] sure you have those safe guardrails in [17:28] place and provide these tools in a [17:30] compliant and a safe and secure manager [17:33] managed so uh manner so that they can uh [17:36] improve their productivity and can be [17:38] deployed in a much more responsible [17:40] manner. [17:42] >> Thank you. Let's do you have a success [17:44] story or a failure for AI project or [17:46] learning that you want to share? I mean [17:47] we we have of course a lot of uh [17:50] successes right I mean uh just uh having [17:54] having our trucks running in Texas on [17:56] public roads autonomously is a is a big [17:59] success right so I think there is there [18:02] is a lot of things but I agree what was [18:04] said at the moment right I mean uh [18:06] Jennai was somehow [18:09] making it easy to access everything [18:12] we're doing right and everybody feels [18:14] like now I I know what it is and I I can [18:17] do it and I'm I'm able to do all of that [18:20] and there was a keynote this morning [18:21] which I uh was attending and and it was [18:24] right you know you need to be basically [18:26] coming back um with all the successes [18:28] and everything because everybody is [18:30] coming and offering you new topics and [18:33] you can do that and this is the best [18:35] tool ever and whatever. Um so what we [18:37] what we did also and this is a success [18:39] from my point of view as well. We [18:42] brought the knowledge back inhouse [18:44] because you need to orchestrate the [18:46] whole thing by yourself. I mean if [18:49] you're just listening to to outside [18:50] companies then you are lost in the whole [18:52] space right so we brought a lot of the [18:54] knowhow back in in house over the time [18:57] now and and made sure that we have the [19:00] the way of handling everything making [19:03] sure for instance that not the wrong [19:05] data is going out especially when you [19:07] look at AI right there is so many agents [19:10] but uh most of them have a very small [19:13] term in the terms and condition that [19:15] they use their data or the data for [19:17] learning and this you don't want. So we [19:20] do a lot of the topics now in C also we [19:22] create our own agents as well making [19:24] sure that for very sensible data we have [19:28] our own stuff and I think this is this [19:31] is the topic of course you can do use [19:32] cases in in the aftermarket space you [19:34] can do use cases in the sales space and [19:37] this is all adding value we just look of [19:39] course for the value if everything what [19:41] is is adding a lot of value we use that [19:44] the the employees they look at it they [19:46] want to use it but there is also a fear [19:49] here that it's taking the job away. So [19:51] you don't want to share everything on [19:53] your knowledge. So there is a lot of [19:55] topics which we still have to do. But I [19:58] think the success beside basically the [20:01] projects is more that we that we learn [20:03] to take that back and that we [20:06] orchestrate the whole topic uh with the [20:08] internal team and not more relying on on [20:11] external parties that much because this [20:13] gives us an opportunity uh because all [20:16] the known [20:18] externals right they run around and they [20:20] say we can do everything but at the end [20:22] of the day we work more with startups [20:24] because it's for us more uh innovative [20:27] than everything else. So I think this [20:30] was a good success internally as well to [20:32] make sure that we get this and build up [20:34] also the knowledge inhouse. [20:36] >> Amazing. And we're going to take it from [20:38] a different lens cuz James has been [20:39] working in this industry with seuite [20:41] executives from public companies for a [20:43] long time working with AI and I'd love [20:45] to hear about a success story that [20:47] you've had or a learning or failure. [20:52] So I always try to what I what I say to [20:55] people is that AI success is in the eye [20:57] of the beholder. And so you really have [21:00] to take into account where the [21:01] organization is in its journey uh before [21:04] you start to define things as successes [21:07] or failure, right? It takes a while to [21:09] to learn. uh and what we do is so rather [21:12] than focus uh I mean and and so many [21:16] people focus on we got to get to pilot [21:18] we got to get to pilot and then from [21:20] pilot we can go from there what we try [21:23] to do is to get companies to focus on [21:26] discovery in the upfront if you go [21:29] through a great discovery process where [21:31] you're getting the stakeholders to the [21:33] table and stakeholders buying into the [21:36] value and sharing with whoever it is [21:39] internal or external [21:41] what they really need and what they [21:42] really expect is from the outcome. [21:45] That's the way you're going to get [21:46] success. And what we see time and time [21:49] again is the folks that invest in that [21:52] um in that discovery process, you can [21:55] get that discovery process, you can turn [21:57] it into a PC, you get buy in at that PC, [22:00] and that PC turns into something that's [22:02] a much bigger initiative. Uh when we [22:05] don't see that, we see people go and [22:06] they invest in a big system. that big [22:09] system usually ends up being something [22:11] that is not no one's using because [22:14] people are just not buying in. So if [22:17] anything I would suggest if you want to [22:19] increase your chances of having success [22:21] and reduce your chances of having [22:22] failures, invest in that discovery [22:24] phase, get the stakeholders to the [22:27] table, go through the processes with [22:29] those stakeholders to really understand [22:31] their perceptions, their value, what [22:34] tools they use that are useful for them, [22:36] what's not useful for them, and then do [22:38] those things. And then when it's time [22:40] for them to use that tool that you're [22:43] building, they buy in. They're there [22:45] with you. And then there's a much [22:47] greater success rate on the other side [22:49] of the entire deployment. [22:52] >> Thank you. And I think in previous years [22:54] we've seen a lot of people have been [22:56] pushing their organizations to be [22:58] innovative, but it didn't fall for [23:00] everyone in the organization to have a [23:02] tech background or to understand tools [23:04] and technology. And so we're seeing a [23:06] large cultural shift in these [23:08] organizations. And so I would love to [23:10] hear what the biggest leadership or [23:12] cultural challenges in scaling AI and [23:14] how you are encouraging your [23:16] organization to leverage some of these [23:18] tools and explore. Kyle, do you want to [23:19] start? [23:21] >> Sure. Um, from an internal perspective, [23:24] I'll take it from two lenses. I'll take [23:26] it first from the internal perspective. [23:27] Um, I think the way the biggest [23:30] challenge is getting I think stakeholder [23:33] buyin and I think what James said [23:34] resonated a lot with discovery because [23:37] if you don't get everybody bought in on [23:40] the solution that you are trying to [23:42] bring into the organization and that [23:44] that person can see it affecting their [23:46] day-to-day and adding value, I don't [23:49] think you're going to get adoption at [23:50] all. So that's number one from an [23:52] internal perspective. um we have a lot [23:54] of tools that help us internally to do [23:57] our work get done work done faster but [23:59] if we weren't a part of the evaluation [24:02] process I don't think we would be bought [24:03] in as much um when I speak to customers [24:06] I think the biggest challenge that [24:08] they're facing is very similar is [24:10] getting internal stakeholder buyin um a [24:12] lot of times people don't know where to [24:14] start with AI that's the biggest [24:16] challenge do you start small do you [24:17] start big and what we advise is start [24:19] with something small that's going to [24:21] create a significant impact and get [24:24] people surrounding that concept of it. [24:27] We have a program, we call it the AI [24:28] advisory program. We walk customers [24:31] through doing discovery of their current [24:33] processes and saying, "Okay, here's [24:36] where our solution is going to step in [24:37] and help you be maybe more productive. [24:40] Focus on KPIs that are going to maybe [24:42] take the time that it took them [24:43] previously to do something and it's [24:45] going to reduce it by a significant [24:47] amount. Also, traditional KPIs don't [24:50] work well anymore. with AI, you have to [24:52] start thinking about KPIs that are going [24:54] to be really aligned to AI um type uh in [24:59] initiatives, right? So, think through [25:01] what is the KPI that you're measuring [25:03] and how are you going to iterative [25:05] iteratively continue to look at that [25:07] metric? Is it decreasing? Is it [25:09] increasing with the help of AI to really [25:11] see if that success is happening within [25:14] to really tackle some of the challenges [25:16] that people are finding with AI? [25:18] >> Thank you. And Reena, what about you? [25:23] Yeah. Um, cultural shift is definitely a [25:26] challenge. Lots mentioned, right? I [25:29] mean, the there is fear of AI taking [25:31] over the your jobs and that is real in [25:33] every organization. So, how do you [25:35] change this mindset? Um, if you can [25:38] invest a ton of money in new tools, [25:41] etc., but if people are not adopting and [25:43] using it, it's pretty much for waste, [25:45] right? So, it's this is a challenge that [25:47] every organization is thinking about. So [25:50] one of the ways that Airbus we're [25:51] addressing is educating the employees [25:54] right on not just what AI is or uh what [25:59] are the tools that are out there but on [26:02] the on the concept of why is why is it [26:04] important why is it important for you to [26:07] focus on it and this education is [26:09] actually across the board uh everybody [26:12] from executives to engineers to other [26:15] departments they educating them on the [26:18] state of AI and why they need to do this [26:20] has helped change the mindset and has [26:24] opened them to be more um receptive to [26:27] using AI. So if you think about it uh AI [26:33] tools are becoming as important and as [26:35] indispensable as the computers that [26:38] changed our the way we work in the past. [26:41] So it is going to be really important [26:43] for [26:45] employees across industries to be able [26:48] to learn these tools really well to work [26:50] with them and equipping our employees to [26:54] overcome their fear and educating them [26:56] and upskilling them and holding their [26:59] hand through the process of making this [27:01] transition I think is really important [27:04] and I would say invest in your employees [27:06] and upscale them for the best of your [27:09] organization. [27:10] Thank you. Um, so I get to spend a lot [27:13] of time with seuite executives at large [27:15] companies and one question I always ask [27:18] more for a lot of personal gain is what [27:21] AI tools are you using yourself for [27:23] individual productivity whether that's [27:25] at work or at home. Um, and if you guys [27:27] want to share a use case or a tool or [27:30] something that you're using to help [27:32] yourself as a leader in a large [27:33] organization. [27:36] Does anyone want to start? I threw this [27:37] a curveball at you. [27:41] I can start. Um, I have the attention [27:44] span of a goldfish. So, anything that [27:47] helps me to understand new stuff in in a [27:50] way that's interactive is super helpful. [27:53] I moved away from reading research [27:55] papers or like uh journals to actually [27:58] listening to you YouTube videos to [28:00] become intelligent on stuff. But one of [28:03] the tools that but I do lose out on a [28:05] lot of technical content, right? And one [28:07] of the tools I love love love is [28:09] Notebook LLM. So the ability to actually [28:14] um use content in any form and make it [28:16] really interactive in terms of a [28:18] podcast. Being able to listen to that is [28:21] a much much better way to assimilate [28:23] information than pouring through papers [28:26] or listening to YouTube videos. And you [28:28] can actually even ask questions and ask [28:31] uh and get like really pointed answers, [28:33] right? And that's been super helpful for [28:35] me. And if anybody hasn't tried it, I [28:37] highly recommend it. [28:39] >> Notebook LLM by Google. [28:42] >> Yeah. [28:43] [Music] [28:43] [Laughter] [28:47] >> I'm pretty simple when it comes to I I [28:49] use a lot of LLMs. And the great thing [28:51] about having a lot of data science [28:53] resources dedicated to you and your team [28:56] is that they can put a lot of guard [28:58] rails and give you an opportunity to [28:59] test a lot of different models. So uh so [29:03] you you name the model uh typically [29:06] using it just lots of them do different [29:08] things right so if it's multimmodal or [29:11] that just depending on what the need is [29:13] I'll I'll turn to one and our team have [29:15] g guard rails set up around those models [29:18] that that I can deploy safely into for [29:21] my own personal use [29:24] >> I was going to say the the Google LLM [29:26] because I use it so much these days we [29:28] have we're a Google suite uh company so [29:31] um it really helps me digest content or [29:33] if I want to come up with like [29:34] presentations, I can like put in my [29:36] notes and it gives me ideas of how to [29:38] even think about the presentation. It [29:40] just gives me more efficiency. Daytoday [29:42] I'm in like meetings helping my team [29:44] out. So when I need to get work done, I [29:46] have very limited time and it really [29:48] really helps you. So I'll just double [29:49] emphasize what was shared on that tool. [29:52] It's really great. [29:54] >> Oops. [29:55] >> I'm using Copilot right where you're [29:57] sitting on on a Microsoft platform. So [30:00] everything what I do is I'm using [30:02] copilot. I'm educating copilot in a [30:05] certain sense to do everything meeting [30:07] summaries doing summaries of uh large [30:10] files and so on and so on. So everything [30:12] what we have there we're using Libra [30:14] chat as well for for certain elements [30:16] right uh because we blocked jet GBT for [30:20] certain reasons u and uh then of course [30:24] you use that and we are currently in a [30:26] in a trial a p phase uh with an agent [30:30] which we built on our by ourselves who [30:32] can trigger stuff in co-pilot and in [30:35] other agents. So we we try to to look [30:38] into how we can do that because at the [30:40] end of the day all the nonvalue adding [30:42] topics I want to get rid of on my job [30:44] and I have a lot of nonvalue adding [30:46] topics like approvals and everything. So [30:48] this I'm the target is by end of the [30:51] year everything is done uh via agents [30:53] and not more from myself. [30:55] >> Amazing. Um and this is probably a good [30:58] opportunity for me to plug a portfolio [30:59] company but I will not. Um, I have been [31:02] using the beta for scouts by uttory. I [31:05] don't know if anyone has played with [31:06] that yet, but I have been using it to [31:08] find restaurant reservations, to monitor [31:10] portfolio companies, growth, hiring, job [31:13] changes, anything that you can think of. [31:15] It sends out scouts for you, and you can [31:17] set email notifications and alerts. Um, [31:20] and I have had a lot of fun playing with [31:21] it over the last month. Um, does anyone [31:24] I don't know how much time we have left [31:25] actually, but does anyone have any [31:27] questions for this group? [31:29] >> Hey, hi. Uh thanks for the all the [31:31] insight which you provided. My question [31:33] is on um once the business value starts [31:37] pursuing AI use cases and other thing [31:40] and if business do not get what is being [31:43] promised immediately the trust started [31:46] diluting on AI and especially when you [31:49] walk with the sales marketing and all [31:51] these business folks. So in this [31:54] situation [31:56] what is the next arm you have to are [31:59] going with the right approach to the [32:02] business and saying that oh no I have a [32:05] another method to bring the more [32:08] business value you know what I mean [32:10] because it's a very subtle area once you [32:13] start stumbling you hardly get the [32:16] second chance to stand up and move [32:17] forward [32:21] >> and and the point that I will make there [32:24] is that's why I highly recommend that [32:27] discovery phase process getting that [32:29] stakeholder buy in at the beginning of [32:32] the process because uh AI is having [32:36] adoption problems I mean great we're [32:38] well there's adoption in lots of [32:39] different ways right but really adoption [32:42] of use inside of the enterprise uh and [32:46] because people are concerned about [32:47] there's just lots of reasons to be [32:49] concerned right about different uh [32:52] potential little detrimental things [32:54] about the technology, but the fact of [32:55] the matter is there's significant [32:57] benefits, right? But in order to uh to [33:01] really get people to adopt, you got to [33:03] get them to the table and let them be [33:04] part of the process and helping to [33:06] define and that process is not only [33:09] about the model, it's about the [33:10] application that that model sits in and [33:13] these sorts of things are really going [33:15] to impact their workflows. Uh, and so if [33:18] it's not going to impact their rec [33:20] workflows and give them back time, give [33:23] them back money, uh, then you're going [33:25] to have a difficult time getting them to [33:27] buy into it. But you have to get them in [33:29] uh, at the beginning uh, if you really [33:32] want them to ensure that you're going to [33:33] be successful. The rest of it is just [33:35] guessing if you're not right. So the the [33:38] the application can do everything you [33:40] ever intended to do, but if people don't [33:43] buy into it impacting their workflows, [33:46] you're not going to get the return on [33:47] the investment that you want. [33:52] >> Look, I mean, we need to make one thing [33:54] clear, right? Uh from my point of view, [33:57] all the industries, it doesn't matter [33:59] which industries it is, need to adopt AI [34:03] over the next 5 to 10 years. 10 years is [34:07] already too long, right? If you're not [34:09] doing it, you are out of business. This [34:11] is my this is my perception of [34:13] everything what is going on. You have a [34:15] huge element of opportunities when [34:18] you're doing it correctly and that means [34:21] you are changing the landscapes the [34:23] architecture of everything what we know [34:25] today you're changing that significantly [34:27] because you will work completely [34:29] different right the ERP systems which [34:31] are running today are history in a [34:34] certain sense right they're trying to [34:36] adopt of course with AI as well but uh [34:39] at the end of the day you don't need it [34:40] anymore for a lot of processes which you [34:42] do um you don't need did we we did uh [34:46] one process in the supply chain or [34:48] procurement area we automated it 90% [34:51] without any ERP and whatever just based [34:54] on AI tools which we have in house so [34:56] you can do that and you need to adopt [34:58] that you know and of course you have [35:00] failure on the way to go there and uh it [35:03] immediately raises the question is AI [35:06] really uh bringing what it's promising [35:09] and whatever but this is also because [35:11] it's a unfortunate hype around that [35:14] topic still and everybody thinks AI is [35:16] the solution for the world's problems [35:18] and this is not the case right we also [35:21] need to go into a place where we say yes [35:24] if you adopt it correctly then it's [35:26] working but you have a lot of work to do [35:28] on the industry on your data your data [35:31] quality is the most important and we do [35:33] not have we maybe have a 50% if we are [35:36] good we have 50% of data which works and [35:39] 50% of data which doesn't work with AI [35:41] solutions so you need to make that clear [35:43] and you need to look at it. But on the [35:45] other side we don't have the time to to [35:47] do a lot of things because from my point [35:49] of view 2030 if you talk about [35:51] architectures 20 4035 or 2040 uh you [35:56] have elements where you say you have 70% [35:59] of it is done by AI tools and taking [36:02] also into consideration that you will [36:04] have AI natives coming basically into [36:07] the workforce at this point of time they [36:09] will not work with any other tool [36:10] anymore they need to have that. So we [36:12] have to do it and we need to go through. [36:16] Uh yes there is things you know uh which [36:20] are going wrong and it's difficult to [36:22] explain but the value is not always [36:24] coming as it's indicated and this is [36:26] what I think we do all on an technology [36:30] space a bit wrong because we are [36:31] promising too much on it. we need to be [36:33] a bit more careful what we're promised [36:35] with these tools and then I think it's [36:37] working also correctly and that's [36:38] exactly how we should do it you know you [36:41] stepwise take the the steps and do it [36:44] and then uh I think it's working not [36:46] everyone is expecting now millions of [36:48] dollars in savings but the problem what [36:51] we have is with the automation and [36:53] everything which is out there and with [36:54] the hype which is out there is an [36:56] expectation of uh the boards that AI [37:00] will solve basically you can cut down [37:02] your your employees by 40% and so on and [37:05] so on and this is not the case right [37:07] there is other areas where you need to [37:09] look into that's the reason I think the [37:11] discovery phase which you mentioned is a [37:13] very good thing to look really what is [37:15] the value and tracking it also and if [37:17] the value is not coming immediately then [37:19] skip it for now right say okay it's not [37:22] it's not yet the time to do it might be [37:25] a bit a later time that this is coming [37:27] but I think this is the discussion you [37:29] need to have very openly with the boards [37:31] as as well to say what are we doing and [37:34] how we are approaching it and that's the [37:36] reason why for a lot of us is very [37:38] important to get the skill set back [37:40] inhouse because the suppliers which we [37:43] had in the past they are promising you [37:44] the world with it and it's not always [37:47] coming so and this is the biggest topic [37:48] what we have but there is no doubt that [37:51] if you are not jumping on this train now [37:55] you are out right this is this is a [37:57] clearcut [37:58] >> yeah and I think just to add to that um [38:01] what we used to do is really strong [38:03] stakeholder mapping and I used to use a [38:05] template from Delta Airlines innovation [38:07] hub the hanger and they would bring in [38:10] individuals that really wanted to be [38:12] involved. They wanted to learn more. [38:13] They wanted to be a part of it. So if [38:14] it's your sales team or a sales process, [38:16] you would have someone there that can [38:18] help give some of that feedback early on [38:19] and then they help you present it to the [38:21] team as a beta and kind of get some of [38:23] that feedback before you just pushing it [38:25] down on the organization. Nobody loves a [38:27] top- down only approach. It has to be [38:29] bottom up as well. And I really think [38:32] rolling things out as a beta and using [38:33] the right terminology and verbiage and [38:35] saying we're all here for the ride and [38:36] we're here to make your life easier [38:38] really does help people get on board and [38:40] give constructive feedback rather than [38:42] saying we're not going to do this or it [38:44] doesn't work. [38:46] I think we have time for maybe one more [38:48] question. Okay, perfect. [38:50] >> Oh, so quick question on pricing, you [38:52] know, because there's we're seeing tons [38:55] of companies that are in the AI game [38:57] now. Um, and coming from a business [39:00] owner in Arkansas, [39:02] most of y'all probably don't know where [39:04] Arkansas is, but anyways, um, [39:08] you know, from a small business [39:10] standpoint, pricing, right? We're [39:12] getting we're getting quotes of a [39:13] million dollars. Hey, this will [39:14] transform your transportation [39:16] department. This will transform your [39:17] service department, your parts [39:18] department, your sales organization. [39:20] We're getting a price of 500,000, [39:22] 200,000. [39:24] How do you navigate to know which [39:26] companies are the best at this? which [39:28] companies are the best at agent [39:29] strategy, etc. Which companies are the [39:31] best at RAW, RA, which companies are the [39:35] best at Genai? [39:36] Like the it just seems very difficult to [39:39] navigate right now. Um because there's [39:41] so much going on, so much buzz around [39:43] it. Would love some feedback and also [39:46] just some um ideas around that if you [39:49] don't mind. [39:51] >> Use AI for it. That's a quick answer, [39:53] right? But not really. [39:56] Um I think uh the important thing is to [40:00] make sure you're continuously monitoring [40:02] this landscape. Whoever it is, whether [40:04] you're an executive or you're a partner [40:07] or whatever it is, uh when you're making [40:10] decision, um there are new startups, [40:13] there are big companies. The way things [40:15] are evolving is when a startup does [40:17] something really well, these big [40:18] companies come and take over that [40:21] particular feature and then add it to [40:22] the set of suites. and then then this [40:25] feature just vanishes from the startup [40:27] right so some of the ways big companies [40:30] are handling it is by making a dual bet [40:32] so they bet in a big organization which [40:34] can provide the end-to-end tool in terms [40:37] of like hey I this is a safe bet for me [40:40] so if there's something new that's [40:42] coming up this company will provide it [40:44] for me on the other end they also bet in [40:46] small startups which have the potential [40:48] to grow but in order to understand who [40:51] is the best of class at some at [40:54] something. The biggest thing is to [40:56] continue to monitor the landscape, [40:58] understand in terms of how people are [41:00] investing in these particular companies, [41:03] talk to them, have conversations on the [41:06] short-listed ones and then um continue [41:09] that engagement uh and understand how [41:12] they showcase success. I think those are [41:15] the things that are really important as [41:17] we move forward. I know it's a [41:18] challenging environment. you have this [41:20] big uh CB insights map and there are [41:23] like hundreds of startups in that space, [41:26] right? Who do you choose? But you want [41:28] to look at the ones that are having [41:29] momentum both from an investment [41:32] standpoint, from a customer standpoint [41:34] and breakthrough standpoint. And [41:35] understanding that I think would be [41:37] helpful as you navigate this landscape. [41:40] say quickly like I think the biggest [41:42] impact is everybody's giving you this [41:44] dream of how AI can work really well but [41:47] how hard is it to implement how hard is [41:50] it to configure how hard is it to [41:52] maintain that's what you should really [41:54] be looking at when you compare vendors [41:56] because I I I we're we're a smaller [41:59] brand but we compete with some of the [42:01] biggest names out there like Service Now [42:03] and Salesforce and all these companies [42:05] but what we've done that's really made [42:07] us unique is that we've uncomplicated AI [42:09] and What we mean by that is you toggle [42:11] it on. It learns from your system. It [42:13] learns your specific business. You said [42:16] I think transportation or something or [42:18] Yeah. So we would learn based on your [42:21] insights, your things which is unique. A [42:23] lot of people are playing around with [42:24] large language models. Everybody can do [42:27] a co-pilot. Anybody can do co-pilot. But [42:29] what about co-pilot bespoke to you and [42:31] your business and how does that impact [42:33] the day-to-day of your people? That's [42:35] key. I can also add for any executives [42:38] if you need support with that. Your [42:40] friendly venture capital firm is always [42:42] there to help you. Um, but something [42:43] that we did uh in both at Amazon and at [42:46] Coca-Cola was setting up really good [42:47] sandbox environments and understanding [42:49] what it took for the provider to [42:52] actually do a demo and how much time and [42:54] how much money it would cost and then [42:55] you can do the calculations on value ad [42:57] there. Um, and so if you are a service [42:59] provider in the room, I highly recommend [43:01] adding a slide in your deck or your [43:03] pitch about what setup it's going to [43:05] take. Um, and what a cost of a demo [43:07] could be or a pilot and then thinking [43:09] about what that larger scale product is. [43:11] No one wants to invest years and [43:13] millions of dollars in something if it's [43:15] not going to give you the returns. [43:16] >> Yeah. And you need to look at it right. [43:19] You have a it's a very good situation [43:21] for us, right? because there is a lot of [43:23] people coming and they offer you the [43:25] solutions and you have every day a [43:27] better solution on the market because [43:29] every day this is evolving. So what you [43:31] need to make sure is that you build your [43:33] architecture very modular that you are [43:35] able to integrate this kind of solutions [43:37] very easily because you come with a very [43:39] monolithic way of architecture nowadays [43:42] right because you have this whatever ERP [43:45] models or whatever whatever and now you [43:47] come to a point where you say no I need [43:49] to have a very integrative and modular [43:51] architecture design so that you can take [43:54] one piece out and you can just block in [43:56] another piece and this is what I would [43:58] look at how easy it is to take that into [44:01] your architecture. And how easy is this [44:03] to take it out again from your [44:04] architecture? Everything where you have [44:06] a lock in, it's a no-go, right? As soon [44:10] as I see a login, it's a no-go. Um, [44:12] these vendors are out, even the solution [44:14] might be a very good one because this is [44:16] not what you want to have. You want to [44:18] have a very modular thing where you can [44:20] really make the use of best best of in a [44:24] sense, right? And this is what you need [44:25] to look into. The final thing that I [44:28] would say to that is collaboration. It's [44:30] in my it's in our name, AI collaborator. [44:34] I founded the company based on the [44:35] premise that no business is ever going [44:37] to have enough resources to do [44:39] everything that they want to do to solve [44:41] for AI. We work with some of the largest [44:43] companies in the world and that's been [44:45] proving true every single day. Find [44:48] great partners that can help you oversee [44:51] manage the you're not going to be able [44:53] to do it by yourself all by yourself. uh [44:55] find great partners that invest in doing [44:58] that work with those partners get those [45:00] recommendations [45:01] and because that's a big part of what we [45:03] do uh we built a network of AI solutions [45:06] and services provider going back eight [45:08] years now and we bring those companies [45:10] to the table we vet them we trust them [45:14] we help the enterprise deploy with them [45:16] so we're out doing that work for them [45:18] find some way to do something like that [45:21] collaboration and partners is key for [45:23] surviving in this industry Okay.