Sodexo, MXD Embed Autonomous AI in Sales to Compress Deal Cycles

Sodexo与MXD将自主AI嵌入销售以缩短交易周期 随着超90%的B2B买家采用AI进行前期调研，营收负责人认为销售团队必须将无需判断的任务交由后台自动化处理，让真人销售代表重新专注于建立信任。 营收运营负责人探讨了如何从被动的AI分析走向直接加快交易节奏的主动、自主执行。与会嘉宾分享了具体的落地案例，例如将定制工业设备的报价周期从五个工作日压缩至90秒，以及构建用于客户评分和企业知识检索的自主智能体反馈回路。他们强调，必须将“隐形AI”直接嵌入现有工作流程，并从零开始重新规划买家旅程，而非将不良流程盲目自动化。最后，专家组指出，随着由AI赋能的买家寻求更高层次的专业指引，现代销售招聘必须优先考量与生俱来的好奇心、理解力与同理心，而非单纯的技术熟练度。 超过90%的B2B买家已在采购过程中使用AI，促使营收团队将整个商业GTM体系围绕AI重新设计，而非仅仅叠加单点工具。 (3:41) AI战略正从被动洞察转向与销售人员协同自主执行；MXD Process通过前端“vibe coding”与稳固后端的结合，将定制工业设备的报价周期从5天缩短至90秒。 (7:04) 衡量RevOps的AI投资回报需聚焦管线流速与交付周期等运营杠杆，40%的产能优化配合40%的人效提升可产生约1.6倍的整体生产力复合增长。 (11:46) Nooks通过自主Agent构建动态客户打分模型，并将销售人员的实时执行反馈持续注入评估体系，以持续优化目标模型。 (12:46) 为推动跨年龄层销售团队的采纳率，AI应深度内嵌于日常工作流做到“润物细无声”，同时借鉴IDEO极限用户配对法让技术熟练与排斥技术的员工共同参与设计验证。 (19:35)

Sales operations leaders argue that commercial AI value requires moving beyond passive informational dashboards to active execution systems that directly accelerate pipeline velocity. To succeed, organizations must re-architect core workflows and enterprise taxonomies before deploying tools, embedding automation invisibly into seller routines while elevating human judgment and trust. Moving from Passive Monitoring to Active Sales Execution (3:41) Charlie Wiebe and Eric Seiberling emphasize that insights do not equal outcomes; commercial leverage requires revenue agents to execute tasks alongside sellers rather than generating passive dashboards. At Nooks, this approach shortened the learning loop by deploying autonomous agents that built account scoring models, conducted background research, and logged whitespace directly into HubSpot. By tracking whether reps accept or modify suggested actions, the system continuously refines its evaluation models using live execution feedback. In production, prioritizing execution directly accelerates deal velocity. MXD Process reduced custom industrial tank quoting from five days to roughly 90 seconds by vibe coding a prototype in three weeks with Claude Co-work and integration partner Level 12. Similarly, Nooks deployed "Nooks Navigator," a federated search agent in Slack that indexes internal knowledge bases to answer complex prospect queries instantly during live trials, helping support proof-of-concept win rates of 80% compared to a 35% baseline. Ground-Up Process Redesign and Common Enterprise Taxonomies (33:03) The panel cautions that applying AI as a tactical patch onto unexamined processes produces generic, fragile workflows. Seiberling advocates mapping buyer journeys through dozens of physical whiteboard iterations before implementing software, separating frictionless background automations from structural tools that fundamentally redefine sales roles. At Sodexo, Olcay Bulgun Gardner notes that overhauling proposal generation required restructuring organizational teams and workflows alongside AI content generation, rather than simply deploying an AI writing assistant. Transitioning from manual data retrieval to proactive push notifications also requires disciplined infrastructure. Automated systems should handle repetitive, judgment-free tasks entirely in the background, reserving push alerts strictly for moments requiring human discretion. Gardner explains that in complex enterprises, establishing the unified taxonomy and shared data definitions needed to govern automated data pushes can require 12 to 18 months of foundational organizational alignment. Overcoming Adoption Divides Through Co-Design and Internal NPS (16:30) Enterprise AI adoption faces significant friction across generational divides. Seiberling describes sales teams split between 30- to 40-year field veterans reliant on physical sales catalogs and digital-native recent graduates. To drive adoption across this spectrum, MXD Process embeds AI natively into daily CRM workspaces as "invisible AI," allowing reps to benefit from automation without navigating distinct, unfamiliar applications. To prevent teams from building unused software, Gardner and Wiebe emphasize frontline co-design. While 20% to 25% of enterprise sellers proactively adopt new technology, engaging the remaining 75% to 80% requires involving reps from day one. Wiebe tracks an internal "RevOps NPS" to confirm whether rapid prototypes solve tangible daily bottlenecks, preventing low-value software sprawl. Similarly, MXD Process pairs extreme user profiles—the most tech-proficient seller alongside the most tech-resistant veteran—in development rooms to ensure new tools accommodate all skill levels before broad deployment. Redefining Hiring and the Seller's Role Around Trust (43:47) With over 90% of B2B buyers leveraging AI during procurement research, the sales rep's traditional function as an informational gatekeeper is obsolete. Gardner notes that Sodexo redesigned its commercial competency frameworks around this shift: because buyers independently gather baseline specifications, reps must now focus on delivering executive reassurance, strategic alignment, and confidence to multi-stakeholder buying committees. Consequently, hiring priorities have shifted away from narrow technical backgrounds toward foundational human traits. Seiberling highlights curiosity, intellect, and empathy as the core criteria for modern sales recruitment. While technical workflows and AI tools can be taught, curiosity drives rep adaptability to evolving systems, intellect ensures rapid comprehension of complex industrial products, and empathy preserves the personal rapport essential for closing enterprise deals. While autonomous agents and automated workflows yield measurable velocity gains, long-term commercial impact requires disciplined change management and clean operational foundations. The panelists emphasize that deploying AI without a common data language, rigorous rep co-design, and a primary focus on human trust risks producing fragmented tools that fail to accelerate revenue.

本次研讨的核心论点指出，随着超过90%的B2B买家在采购流程中采用AI，营收运营（RevOps）必须从被动的报表洞察与聊天界面，转向深度嵌入日常工作流的闭环执行与组织重构；企业的真正优势在于通过压缩交易周期、实施人机决策分流，以及强化销售人员不可替代的信任构建能力来驱动业务增长。 从信息洞察转向自动化闭环执行 (3:41) Nooks的Charlie Wiebe提出“认知不等于行动，洞察不等于结果”，强调AI系统必须超越被动监控，转由智能体与人类协同执行以缩短学习闭环。以Nooks重塑的客户评分体系为例，RevOps团队不再花费数月手动处理电子表格，而是开发AI技能驱动智能体自主生成评分模型与评估框架，在HubSpot中完成客户分层并挖掘空白线索；在销售跟进过程中，智能体会追踪其建议任务是否被采纳或修改，将执行信号实时回流至模型进行自我校正。Eric Seiberling同样主张将洞察转化为CRM中可转化的自动化动作。MXD Process联合集成商Level 12，借助Claude Co-work协同开发前端并搭建稳健后端，耗时3周完成原型开发、5周完成数据与工程模型验证，上线了面向工业装备的Tank Designer应用，将原本需要工程团队耗费5个工作日的定制报价流程压缩至约90秒，显著加速了重资本支出（CapEx）的销售节奏。 压缩交易周期与释放组织产能倍增效应 (11:46) Sodexo的Olcay Bulgun Gardner指出，在企业级服务领域“时间是交易的头号杀手”，在销售管道中停滞过久的交易最终往往以失败告终。因此，RevOps评估AI投资回报率（ROI）必须紧扣业务痛点，重点追踪管道周转速度、提案交付周期与转化效率，而非孤立的技术指标。Sodexo通过将AI内容生成与组织架构重塑深度结合，重组了原本极其繁重、耗费大量人力的合同提案流程。运营数据显示，40%的人力容量缩减结合人均生产率40%的提升，为整个组织带来了约1.6倍的综合生产率倍增效应。 在应对高负荷商机时，AI对交易周期的压缩效果尤为显著。MXD Process依靠7名销售代表应对每年12,000个进线商机，系统通过在通话前自动生成包含设备需求预测与定制发现问题的简报，使传统报价周期从6天缩短至24小时以内，结合配置工具甚至降至数分钟，最终推动成单率提升10%。Nooks则针对其关键的2周试用期（试用阶段成单率达80%，而整体成单率为35%）推出Nooks Navigator，利用向量化语义检索汇总散落于Notion、Google Drive和表格中的零散知识，在Slack中实时辅助销售解答复杂监管与产品疑问，消除了繁琐跨系统查找对成单速度的拖累。 隐形AI与极端用户共创破除采纳壁垒 (16:30) 在跨代际团队与大型企业中推行AI面临显著的采纳阻力。Seiberling指出，MXD Process的销售团队横跨两极：既有依赖纸质画册和面谈、拥有30至40年行业经验的老销售，也有习惯数字化工作流的年轻毕业生。为避免系统沦为摆设，MXD Process采取“隐形AI”（Invisible AI）策略——将技术像墙内的电线一样无形嵌入日常流程，让销售代表无需学习独立软件即可享受自动化便利，全流程仅由1名商业技术负责人与1名网络IT管理员极简维系。Gardner亦补充，大型企业中通常仅有20%至25%的早期拥抱者会主动学习智能体，其余75%至80%的群体则需要绝对无缝的工作流体验，因此必须从第一天起就让一线销售参与流程与工具的共创设计。 为防止RevOps陷入“几小时内快速构建出无法解决实际问题的无效工具”陷阱，Wiebe将内部销售净推荐值（RevOps NPS）作为核心衡量标准，定期向SDR和AE验证工具是否真正解决了日常销售瓶颈。Seiberling则引入IDEO的“极端用户设计框架”，在开发会议中同时把最懂技术的代表与最抗拒技术的资深销售安排在同一个房间，直接对原型展开极限压力测试与实体白板推演，确保方案在全面推广前兼顾不同技能水平的用户需求。 决策分流架构与B2B销售核心胜任力演进 (33:03) 面对从“销售手动检索仪表板”（Pull）向“系统主动推送关键数据”（Push）的架构转型，专家们强调必须建立严格的人机分流机制。若迫使销售在企业内四处搜寻线索，交易速度势必停滞；但盲目推送同样会造成干扰。合理的架构是将纯数据与确定性任务完全交由后台自主执行，仅在需要人类运用战略判断的关键节点，才主动向销售推送提醒和上下文。然而Gardner提醒，大型传统企业实施主动推送的前提是耗费12至18个月打通数据孤岛、建立跨部门统一的业务词汇表和标准化分类法，否则缺乏上下文的推送只会放大混乱。 与流程演进相伴的是人才评价标准的根本转变。由于买家已普遍利用AI自主获取基础产品信息，销售代表的价值不再是信息传递，而是交付信任与确定性。Seiberling与Gardner一致认为，现代AI销售招聘不再以特定技术专长为唯一门槛，而是聚焦于三大底层特质：驱动自驱适应与工具学习的好奇心（Curiosity）、理解复杂工业与商业逻辑的学习力（Intellect/Smarts），以及在多方决策委员会面前建立人际信赖的同理心（Empathy）。 该研讨表明，AI重构销售运营的核心不在于盲目部署算法，而在于将前沿工具与底层业务流程、组织共识及以人为本的信任关系进行深度编排。然而，对于数字化基础薄弱、流程未标准化的传统大型企业而言，跨系统数据治理与建立通用词汇表的前置周期依然漫长，如何在耗时数季度的底层架构改造期间平稳平衡短期业务产出与长期自动化构想，仍是销售管理者面临的现实挑战。

## Background 本内容由频道 Ai4 于 2026 年 8 月 21 日发布在 YouTube 平台上。视频记录了一场关于现代销售运营（RevOps）中 AI 落地的专题圆桌讨论，汇聚了来自 Sodexo、Nooks 和 MXD Process 的行业高管与投资人，共同探讨企业如何借助 AI 识别核心商机、实现管理任务自动化、实时赋能销售代表以及重构端到端的销售漏斗。 ## TL;DR 圆桌嘉宾围绕 AI 在收入运营中的实际业务闭环展开了深度剖析，强调“洞察不等于结果”，AI 的终极价值在于直接执行并缩短学习闭环。MXD Process 分享了利用 AI 将工业储罐报价周期从 6 天压缩至数分钟、推动赢单率提升 10% 的实践，Sodexo 与 Nooks 则分别阐述了提案生成重塑、智能客户评分以及企业级知识 Agent 的落地经验。嘉宾们一致认为，企业必须破除“AI 与非 AI”的技术站队思维，将 AI 像“电力”一样隐形嵌入既有工作流，并重构以好奇心、聪颖度与同理心为核心的新时代销售人才画像。 ## Atmosphere & style 整场圆桌气氛活跃、务实且极接地气，开场以轻松的摇滚演唱会破冰拉近了嘉宾与观众的距离。主持人 Chris Yeh 节奏把控极佳，配合幽默风趣的穿插互动，与三位来自跨国服务、SaaS 与重工业制造领域的实战派高管碰撞出密集火花。对话摒弃了空洞的技术炒作，充满对一线销售真实痛点的同理心，展现出一种既紧跟前沿技术又脚踏实地注重商业转化的务实风格。 ## Key takeaways - [03:41](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=221s) 超过 90% 的企业买家已在采购决策链各环节使用 AI，驱动收入运营（RevOps）必须从单纯添加单点工具升级为重塑整体商业运转与 GTM 架构。 - [03:41](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=221s) AI 落地逻辑已从单纯提供洞察转向闭环执行，通过缩短学习闭环让智能体在实践中快速迭代并修正下一次行动。 - [07:04](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=424s) MXD Process 结合 ChatGPT 梳理工程规则与自然语言编程（Vibe Coding），配合专业系统集成商开发出储罐设计器，将原本需 5 天的工业报价缩短至 1 分半钟。 - [09:43](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=583s) 评估 AI 投资回报率不能只看单点功能，而应聚焦于业务流速与转化率等中间过程指标，索迪斯借此实现了提案团队 1.6 倍的复合生产力提升。 - [11:46](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=706s) Nooks 利用核心 Prompt 衍生出一组 AI 智能体群，完成客户评分与空白市场挖掘，并与一线销售协同形成“执行-反馈-反哺框架”的自适应闭环。 - [16:12](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=972s) 面对传统老将与年轻数字原住民之间的技术代差，AI 应像电力一样无感且隐形地嵌入现有 CRM 和工作流，而非强推独立新软件。 - [19:35](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1175s) 收入运营应引入内部净推荐值（RevOps NPS）作为衡量标准，并在设计初期同时引入“极客用户”与“传统销售”两种极端画像以降低全员推广摩擦。 - [22:13](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1333s) 时间是交易的最大杀手，工业制造场景中借助 AI 提前预研线索并在数分钟内交付报价，直接带动成单转化率提升 10%。 - [24:45](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1485s) 针对重度试用期知识碎片化的痛点，Nooks 建立向量化与联邦搜索智能助手 Nooks Navigator，支持销售秒级检索冷门规则并由人工确认回复。 - [27:00](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1620s) 避免陷入“AI 还是非 AI”的技术二元论与新事物综合征，AI 的本质是作为传统软件与数据之上的控制层，以解决业务问题和满足客户为最终目标。 - [32:00](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1920s) 销售信息分发正从“销售主动拉取数据”转向“系统智能主动推送”，凡不需要人类判断的环节全自动化，需商业判断的才精准推送到工作台或 Slack。 - [35:25](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2125s) 冷营销中的礼品机制不应作为孤立噱头，而必须嵌入销售序列并紧扣买家旅程，结合 AI 深度个性化调研以建立信任光环。 - [38:49](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2329s) 迈向 AI-First 组织最关键的工具是“白板思考”，需先厘清买家旅程并重塑流程，明确 AI 是在后台隐形提效还是彻底颠覆工作性质。 - [43:47](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2627s) AI 时代销售的核心胜任力画像聚焦于好奇心、聪颖度与同理心，销售角色正从信息传递者升维为向采购决策委员会交付信任与笃定感的关键载体。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=4s) 开场与嘉宾破冰：从摇滚演唱会到 AI 销售运营 主持人 Chris Yeh（《闪电式扩张》合著者、Blitzscaling Ventures 创始合伙人）热情开场，点明本次圆桌的核心议题：如何利用 AI 真正推动交易达成与业务变现。为了让现场气氛迅速升温，Chris 提议各位嘉宾在介绍自己专业背景的同时，分享各自人生中第一次出于自愿（而非被父母强拉着去）参加的音乐会。Chris 首先自告奋勇，透露自己人生第一场自发前往的演唱会是在奥克兰体育馆（Oakland Coliseum）观看的 Pink Floyd 演出，现场标志性的巨型充气飞猪道具令他至今记忆犹新。 随后各位嘉宾依次亮相并分享了各自充满趣味的个人经历： - **Olcay Bulgun Gardner**：Sodexo（索迪斯，全球最大的餐饮和设施管理服务提供商之一）收入运营（RevOps）高级副总裁。她加入 Sodexo 已有两年，回忆起自己的第一场演唱会是在伊斯坦布尔举办的 Metallica 演唱会——这也是伊斯坦布尔历史上第一场体育场级别的巨星演出。当时年少轻狂的她买了内场站票，结果切身体验到了真实的“人踩人狂潮”，引得现场一片笑声。 - **Charlie Wiebe**：AI 收入 Agent 平台 Nooks 的 RevOps 副总裁。Nooks 致力于打造能够与人类协同工作、不仅能提取洞察更能直接执行任务的 AI Agent。Charlie 于去年初加入公司，他自愿参加的第一场演出是 2002 年在伯克利希腊剧院（Greek Theatre）举办的鲍勃·迪伦（Bob Dylan）与其乐队的现场演出。 - **Eric Seiberling**：MXD Process 销售与市场副总裁，同时担任一家三级包装公司的首席收入官。MXD Process 是一家年营收约 3000 万美元的工业设备制造商，专注于生产用于各类生产线的不锈钢储罐与工业搅拌器，客户应用场景横跨从烧烤酱生产到航空航天涂料等数十个行业。Eric 的第一场演唱会经历发生在他就读于美国空军学院（Air Force Academy）担任学员期间，当时摇滚女将琼·杰特与黑心乐队（Joan Jett and the Blackhearts）到校演出，他作为现场工作人员在执勤中看完了整场演出。 --- ### [03:41](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=221s) 核心议题：当前案头最紧迫的 AI 优先事项 进入正式讨论后，主持人抛出了第一个问题：各位嘉宾案头目前优先级最高的 AI 任务究竟是什么？ Sodexo 的 Olcay Bulgun Gardner 率先发言，她指出当前最高优先级的战略是围绕 AI 全面重塑企业的商业运转系统。她强调市场的买方行为正在发生深刻剧变：“超过 90% 的企业买家已经在其采购决策链的各个环节中使用 AI 工具。”因此，收入运营的核心职责已不再是单纯增添单点工具，而是全面升级 Go-to-Market（GTM）组织架构，在买方旅程的每一个触点赋能销售人员。 ```mermaid flowchart LR accTitle: AI Agent Execution Loop accDescription: Moving from raw insights to executed actions with closed feedback loops subgraph Input["感知层"] A["商业洞察与数据<br/>Insights"] end subgraph Agent["Nooks Agent 执行层"] B["协同人类执行任务<br/>Action & Execution"] C["结果反馈与校准<br/>Short Learning Loop"] end subgraph Output["业务产出"] D["实际商业结果<br/>Revenue Outcomes"] end A --> B B --> C C --> B B --> D ``` Nooks 的 Charlie Wiebe 随后深入拆解了 AI 落地中的底层逻辑转移。他提出了两个核心原则：**“知（Knowing）与行（Doing）截然不同，洞察（Insights）并不等同于结果（Outcomes）”**。当前 AI 赋能的终极目标，是将各个业务触点的 AI 从单纯提供分析建议推进到直接落地执行。Charlie 目前最紧迫的任务，是通过实际执行来缩短 AI Agent 的“学习闭环”（Learning Loop）。当 Agent 与人类销售并肩工作、亲自执行外联和转化动作时，它们能够直接在实践中学习何为最佳实践，并在极短的周期内基于上一次操作的结果迅速修正下一次行动。 MXD Process 的 Eric Seiberling 紧接着呼应了前两者的观点。MXD Process 涉及 20 至 30 个高度复杂的工业制造领域，前期他们基于化工与工程专业知识构建了定制化 GPT（Custom GPTs）作为底层洞察引擎。如今团队的首要任务，是将这些模型从孤立的聊天框中“解救出来”，直接嵌入到 HubSpot CRM 等核心销售系统里，将工程洞察自动转化为销售人员可立即执行的转化动作，以极快的速度推动大型资本支出（CapEx）项目的商务进程。 --- ### [07:04](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=424s) 现场互动与实战案例：MXD Process 打造“储罐设计器” 为了摸清行业现状，主持人向现场观众发起了两轮即时举手表决：首先询问有多少人已将 AI 作为标准组件融入 GTM 和 RevOps 工作流，随后追问这批先行者中又有多少人已经看到了确凿的收入增长。在看到台下不少观众举手示意已获得早期回报后，讨论转向了具体的落地案例解析。 Eric Seiberling 详细拆解了 MXD Process 刚刚上线的旗舰 AI 项目——“储罐设计器”（Tank Designer）： - **业务痛点**：工业储罐的报价极度复杂，既要考虑食品级卫生标准，又要兼顾强腐蚀性化学品所需的防爆要求，传统模式下技术销售团队为客户出具一份不锈钢储罐报价单需要整整 5 天。 - **技术落地与“Vibe Coding”**：团队首先提炼出工程设计的确定性规则框架，利用 ChatGPT 整理工程逻辑，并借助 Claude 协作进行敏捷的前端“Vibe Coding”（氛围编码/自然语言编程）。 - **专业系统集成（SI）保障**：由于纯 Vibe Coding 极为脆弱且容易崩溃，团队联合了名为 Level 12 的专业软件集成商。在业务团队进行前端自然语言构建的同一天，工程师团队同步开发稳固的后端架构。 - **项目周期与突破性成效**：整个项目仅用 3 周便完成了可运行的原型开发，随后历时 5 周进行工程模型严谨验证与生产数据导入。在圆桌召开的 9 天前，该工具正式在 `designer.mxdprocess.app` 上线投产。 > “过去通常需要 5 天才能完成的不锈钢储罐报价，现在客户在网页端大约只需 1 分半钟就能搞定。” Chris 随即幽默调侃道：“只要是 Vibe Coding 崩溃而不是真实的储罐在工厂里炸裂，那就没问题。”Eric 笑着回应称现在的生产环境早已脱离脆弱的氛围编码阶段，系统非常稳固可靠。 --- ### [09:43](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=583s) 衡量 AI 投资回报与 Sodexo 的方案生成变革 针对主持人提出关于“如何向高管层证明 AI 投资回报率（ROI）”的疑问，Sodexo 的 Olcay Bulgun Gardner 给出了系统性的思考路径。她强调，衡量 AI 的成效绝不能仅仅停留在“部署了某项具体功能”的浅层视角，而是必须审视 AI 融入其中的全新工作模式与组织流程，看它究竟解决了何种核心业务痛点。 Olcay 指出，尽管销售组织的终极衡量指标永远是销售额与营收，但更具诊断价值的过程指标是**业务流速（Velocity）**与**转化率（Conversion Rate）**： - **提案生成（Proposal Development）重构**：这是 Sodexo 在 RevOps 领域落地的首批重点 AI 举措。 - **行业特殊性挑战**：与传统科技行业精简的 SaaS 提案截然不同，餐饮与综合设施管理属于重度依赖服务体验的款待业（Hospitality），其竞标提案内容篇幅庞大、文案极为繁重、高度依赖人工编写。 - **流程与组织协同重塑**：Sodexo 在引入 AI 加速文本创作和排版提效的同时，同步重构了内部的提案团队组织架构与协同审批流程，从而在全生态系统维度上实现了提案交付周期的压缩与竞标胜率的提升。 ### [11:46](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=706s) 销售效能的度量与 AI 驱动的客户评分机制 Olcay Bulgun Gardner 分享了通过优化流程所取得的显著成效：团队不仅成功削减了 40% 的冗余产能需求，同时让每个人的生产力提升了 40%。这意味着整体生产力实现了约 1.6 倍的复合增长。他强调这并不仅靠 AI 技术本身，而是关注那些直接影响最终营收的中间指标，例如商机入库量（velocity）以及从发起提案流程到最终交付给客户所耗费的天数。 随后 Charlie Wiebe 分享了 Nooks 近期在客户评分（Account Scoring）项目上的革新实践。传统上，收入运营（RevOps）团队要耗费数月甚至数个季度，在混乱的电子表格、重复数据和低劣的补全信息中苦苦挣扎，评分本身反而成了终极目标。进入 2026 年，他们将核心关注点转移到每位 SDR 带来的销售管线（Pipeline per SDR）以及每位 AE 最终关闭的营收（Revenue close per AE），核心命题转变为“销售人员是否在一开始就跟进了对的客户，而不是在低匹配度客户身上浪费时间”。 ```mermaid flowchart TD accTitle: AI 客户评分与人机协作闭环 subgraph ModelGen["AI 智能体群构建阶段"] A["单一 Skill 与 Prompt 精炼"] --> B["生成 AI 智能体群"] B --> C["构建各梯队评分模型 & 评估框架"] C --> D["HubSpot 全库打分 & 挖掘空白客户"] end subgraph Execution["日常人机协作阶段"] D --> E["周一早晨分配销售目标<br/>(例如：向 Sodexo 销售)"] E --> F["AI 智能体完成深度背景调研"] F --> G["人机协作执行任务<br/>(人类负责判断与建立信任)"] end subgraph Feedback["自适应闭环"] G --> H["收集人类反馈与任务采纳结果"] H -->|反哺数据与记忆| C end ``` Nooks 没有继续耗费大量人力去手工处理表格，而是通过精炼 Prompt 构建了一套核心 Skill，从而派生出一组能够自我构建客户评分模型和应用界面的 AI Agent 群。这批智能体制定了分层模型与“黄金客户”评估框架，对 HubSpot 中的所有客户进行打分，并主动挖掘出数据库外的空白市场（whitespace accounts）。 更关键的是形成了持续迭代的闭环：在评分过程中积累了“记忆”的这批智能体，直接作为一线销售日常工作的辅助工具。当销售人员周一早上决定去开拓目标客户（如 Sodexo）时，智能体能够协同进行背景调研并寻找关键联系人。人类发挥自身所长——建立信任、进行价值判断与承担责任；而智能体根据销售人员的采纳与修改反馈，持续将这些行为数据反哺至评估框架中，实现了从评分生成、执行到动态迭代的高效运转。 --- ### [16:12](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=972s) 跨越技术代差：将 AI 隐形化嵌入业务工作流 Chris Yeh 提出了一个行业普遍痛点：即便搭建了先进的基础设施，销售团队依然面临使用意愿不高的问题。他向 Eric Seiberling 提问，面对背景差异极大的销售群体（从经验丰富的老牌销售到科技素养不高的工业制造销售），如何推动工具的采纳。 Eric 首先幽默地纠正了 Chris 的玩笑，指出 MXD Process 并非直接销售热酱（hot sauce）或润滑油，而是制造用于加工这些产品的大型不锈钢混合设备。他指出团队内部存在巨大的代际鸿沟：一部分老销售已经在这个领域深耕了三四十年，职业生涯早期甚至只靠一本纸质销售手册走访客户，连电脑都极少接触；而另一端则是新入职的大学毕业生，对新技术适应极快。 为了应对不均衡的工具采纳率，Eric 认为绝不能抱有“只要建好了工具，他们自然会来用”（Field of Dreams）的天真想法。AI 的落地重点已不再是引入更多独立的新软件，而是将 AI 深度“烘焙”到现有的日常工作流中，使技术像电一样无形化——用户无需了解电线是如何布置的，只需打开开关获得照明。对于中小型企业，RevOps 的核心职责正是建立这种无缝集成的底层架构，让销售人员在不知不觉中享受 AI 带来的赋能。 - **代际差异**：销售人员跨度极大，涵盖从业 30-40 年、从无纸化时代走来的老将，到刚毕业的技术原生代年轻人。 - **隐形化设计**：AI 落地不应强加新工具负担，而应像“电力基础设施”一样无感嵌入现有系统。 - **精简技术人员**：以极小规模的技术支持力量（如一人负责商业全盘兼顾创新，配合一名网络 IT），倒逼工具集成方案必须极度稳健、简便且易于维护。 --- ### [19:35](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1175s) 业务协同设计与 RevOps 内部净推荐值（NPS） Olcay Bulgun Gardner 补充指出，工具采纳难题在大型复杂企业中更为严峻。通常只有前 20% 到 25% 的高敏锐度用户会主动拥抱并尝试 AI 智能体，要带动其余绝大多数人，唯一的出路就是在系统层面实现无缝的工作流嵌入。他强调，必须让销售团队从最开始就深度参与到新流程与工具的设计中；缺乏这种前期共创，后续的推动和赋能必将遭遇巨大阻力。 Charlie Wiebe 随后分享了评估收入运营成功与否的核心度量标准——**RevOps 内部净推荐值（Internal RevOps NPS）**。在 AI 辅助编程（如“氛围编程” vibe coding）可以在几个小时内快速开发出一个原型的时代，团队很容易陷入制造销售人员根本用不上的功能陷阱。因此，他们持续用最朴素的问题向 SDR 和 AE 调研：“RevOps 搭建的工具是否切实解决了你日常工作中的真实问题？你真的在用它吗？” Eric Seiberling 总结了他们借鉴 IDEO 设计思维所采取的实践策略：在进行工具原型设计或编码时，特意邀请两位“极端用户”同时在场——一位是对新技术极度敏感的技术极客，另一位则是对计算机极不感冒的传统销售人员。只要设计出的解决方案能够同时满足这两类极端人群的实际需求，工具在全员推广时就能大幅降低摩擦，显著提高上线成功率。 ### [22:13](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1333s) 工业制造与长周期交易中的 AI 赋能：从数天到数分钟的报价跨越 在探讨销售团队如何具体借助 AI 提高成交速度与胜率时，MXD Process 销售与市场副总裁 Eric Seiberling 分享了工业制造领域的实际挑战。MXD Process 服务的行业极为广泛，产品复杂度高，销售代表可能上一秒还在沟通用于流体混合的“高剪切乳化机”（high shear mixing），下一秒就要对接油田防沉积用的大型侧入式搅拌器（big side entry mixers）。面对仅有 7 名销售代表却要承接年均 12,000 条线索（leads）的庞大业务量，团队利用 AI 提前完成客户背景研究。在销售代表拨通客户电话前，系统已自动生成简报，明确标出该工艺大概率适用的设备配置、潜在竞争对手清单以及推荐的探寻问题（discovery questions）。 更深远的突破体现在报价效率与转化率上。过去销售人员需要花费数天收集参数、计算成本，平均报价周期长达 6 天；引入前期 AI 辅助后缩短至 1 天以内，而在推出“储罐设计器”（Tank Designer）这一自动化工具后，出具复杂方案的时间直接被压缩到数分钟级。报价速度的大幅提升和对客户需求的深刻理解，直接拉动赢单率（close rate）上升了 10%。 Sodexo 收入运营高级副总裁 Olcay Bulgun Gardner 对此深表赞同，直言“在我们的行业经验中，时间绝对是交易的头号杀手”（time is the absolute killer of deals）。销售管线中那些最终流失的交易，往往都是拖延时间最长、进展最慢的。在缺乏 AI 工具支持时，销售人员耗费大量时间从各处抓取基准指标、定价与成本数据，严重拖慢了决策进度。通过 AI 极大缩短“从获取信息到采取行动”（distance between learning and action）的路径，与最终的营收增长及转化率提升存在直接的正相关关系。 - **MXD Process 团队负荷**：7 名销售代表应对年均 12,000 个销售线索。 - **报价响应周期变革**：行业基准平均 6 天 $\rightarrow$ AI 辅助流程降至 1 天以内 $\rightarrow$ 结合 Tank Designer 缩短至数分钟。 - **业务收益**：整体成单转化率直接提高 10%。 ```mermaid flowchart LR accTitle: AI-driven Sales Enablement Workflow subgraph DataGathering["线索流入与预研"] L["12,000 工业线索/年"] --> AI["AI 自动化预研与简报生成"] AI --> BR["设备配置 / 竞对分析 / 探寻话术"] end subgraph RepAction["销售快速响应"] BR --> REP["7人销售团队精准沟通"] REP --> TD["Tank Designer 快速选型报价"] end subgraph Outcome["业务成效"] TD --> RES["报价周期：6天 ➔ 数分钟<br/>赢单率提升：+10%"] end ``` --- ### [24:45](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1485s) Nooks 的试用攻坚与企业知识检索助手“Nooks Navigator” Nooks 收入运营副总裁 Charlie Wiebe 随后分享了 SaaS 销售场景下的 AI 落地案例。Nooks 的全流程平均赢单率为 35%，但只要潜在客户进入为期两周的产品试用（2-week trial），赢单率就会飙升至 80% 左右。这意味着交易的成败几乎完全取决于能否引导客户开启试用——销售往往会直接向客户承诺：“让你们的 SDR 试用两周 Nooks，外呼到对话转化率（call to conversation rate）和对话到会议转化率（conversation to meeting rate）将直接翻倍；如果达成这个效果，你们买不买？” 然而，两周试用期的运营管理极其重度。客户经理（AE）必须常驻在与客户共享的 Slack 频道中，每周陪同 SDR 参加 3 到 4 场集中拨号会（call blocks）。在此期间，潜在客户会提出大量高度复杂且细致的技术及合规问题，而相关答案散落在 Notion、Google Drive、电子表格及演示文档等各个角落。 为解决知识碎片化难题，Nooks 在内部构建了一款名为“Nooks Navigator”的 Slack 智能 Agent： 1. **向量化与联邦搜索**：系统将公司所有数据库进行向量化处理，建立支持语义搜索的企业级联邦搜索（federated enterprise search）层。 2. **自然语言问答交互**：当客户提出诸如“如果我在巴西，能否用澳大利亚号码打往德国？”这类冷门规则问题时，销售代表可以在 Slack 侧栏直接向 Nooks Navigator 提问，或直接将客户消息转发给该 Agent。 3. **人工确认机制**：AI 检索出最优信源并生成回答草稿，销售代表审核、调整后即可秒级回复客户。这一工具彻底化解了销售在多套文档间反复翻找的困境。 --- ### [27:00](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1620s) 破除“AI 与非 AI”的技术二元论：业务目标、碎片化与执行瓶颈 针对主持人 Chris Yeh 关于“AI 如何与现有非 AI 软件系统及一线员工协同工作”的提问，专家们展开了深入探讨： Olcay Bulgun Gardner 首先从非技术业务视角的顶层设计切入。她强调，谈论 AI 时本质上是在谈论“人、流程与技术的有机结合”（people, processes, and technology）。企业不需要盲目追逐哪款技术参数更高，而是要明确自身的增长飞轮（growth mechanism）。Sodexo 内部推行了业务端与技术端的深度一体化协作模式：业务负责人负责梳理真实痛点并提出商业愿景，IT 及技术团队以此为基础协同设计落地架构，并将最终使用者（一线销售代表）全面纳入解决方案的共创过程中。 Charlie Wiebe 指出，这并非一个全新的技术难题，其根本痛点依然是“系统与数据的碎片化”（fragmentation）。当前 AI Agent 的核心逻辑是模拟人类行为（如 Claude 操作 Chrome 浏览器），因此 Agent 的失败场景和人类如出一辙：如果某项信息在底层知识库中本身就是缺失、混乱或人类员工也查不到的，那么 Agent 同样会撞墙，甚至直接产生幻觉（hallucinate）。Charlie 强调，AI 固然擅长提炼总结和生成建议，但商业转化真正的瓶颈永远在“执行”：与其用 AI 再生成一份毫无实质意义的“会议纪要总结”，不如将 AI 作为非 AI 工具之上的控制层，在人类经验的判断把关下切实执行业务操作。 Eric Seiberling 更是尖锐地指出，陷入“AI 还是非 AI”的技术站队是非常愚蠢的，工具不过是达成目标的手段。销售的根本终局非常纯粹：“一是赚钱，二是让客户满意”（make money and make customers happy）。他告诫企业切勿患上“AI 炫酷新事物综合征”（shiny object syndrome），成天将“Agent、Agent、Agent”挂在嘴边只会让团队无所适从。正确的路径应是以业务为中心：明确客户需求与决策所需数据，进而编排（orchestrate）AI 与传统软件工具，以最简洁高效的方式让机器提取数据，最终辅助人类做出影响业务的正确决策。 --- ### [32:00](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1920s) 现场互动：从“主动拉取”到“智能推送”的销售信息分发转型 进入现场问答环节后，来自 Adobe 的参会者 BV 提出了一个核心转变问题：过去销售人员通常需要主动登录控制面板（dashboard）或特定界面去“拉取”（pull）数据，企业如今应如何完成范式转移，转而将关键信息与行动指引主动“推送”（push）给一线销售团队？ Eric Seiberling 率先解答了 MXD Process 的实践方案。他们在 CRM 体系中专门构建了一套“销售工作台”（sales workspace），其核心设计逻辑就是摒弃让员工在系统里到处捞取线索的旧模式，转由系统主动将当天必须打的电话、必须跟进的商机、需要即刻处理的待办事项以及配套的上下文数据集中推送到销售眼前。 ### [03:30:03](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=12603s) 从洞察到行动：主动推送与跨组织通用语言的重要性 销售运营的核心在于解决销售人员的实际痛点。当交易陷入停滞时，团队需要设身处地思考销售代表究竟需要什么资源才能做好工作，并主动将这些信息推送给他们。如果销售代表必须在企业内部四处“搜寻”数据，这种流程注定会失败。构建以销售代表体验为核心的机制虽然需要时间与理解，但通过每天循序渐进的调整，能从根本上改善工作流。 ```mermaid flowchart TD accTitle: AI信息分发与决策工作流 subgraph S1["数据识别与处理"] A["业务数据 / 交易动态"] --> B{"是否需要人类判断？"} end subgraph S2["执行路径"] B -- "否 (纯自动化)" --> C["AI / 机器自动执行<br/>无需人工介入"] B -- "是 (需判断力)" --> D["主动推送给销售代表<br/>(Slack 提示 / 决策卡片)"] D --> E["销售做出判断并由系统跟进"] end ``` 在讨论数据与行动的关系时，嘉宾们强调了“认知不等于执行，洞察不等于结果”（"knowing is not equal to doing and insights are not equal to outcomes"）。如果某项数据完全不需要人类参与判断，就应交给机器人自动化执行；只有在需要行使人类商业判断时，才应采取“主动推送”（Push）的机制，例如在 Slack 中直接 @销售代表 提出需决策的关键问题。此外，对于大型复杂企业而言，若没有统一的业务语言和数据标准，AI 推送将无从谈起。组织往往需要先花费 12 到 18 个月构建统一的业务框架与语言体系，随后才能顺利推进 AI 工具与能力的落地。 --- ### [03:35:25](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=12925s) 破解冷接触僵局：深度个性化与礼品策略的结合 现场观众 Shrar（来自 Cloud Paths，开场幽默表示自己最喜欢的乐队是 Maroon 5）提出了关于冷邮件泛滥及陌生销售送礼箱（如饼干盒）造成困扰的问题，询问 AI 如何改善冷触达与冷营销。 ```mermaid flowchart LR accTitle: 销售序列中的礼品递进模型 A["领英好友请求通过"] --> B["产生双向沟通"] B --> C["赠送小额礼品 (如咖啡券/午餐)"] C --> D["预约演示 Demo (如赠送 AirPods)"] ``` 针对这一现象，嘉宾们回顾了营销历史并分享了破局思路： - **从物理冲击到 AI 深度个性化**：Chris Yeh 回忆自己早年担任营销副总裁时，曾寄送装有幸运饼干的箱子，主张“要在客户桌上放一个能发出沉重声响（thunk）的东西”来吸引眼球。如今有了 AI，销售团队能够实现更深度的个性化准备。例如在重大投标前，利用 AI 调研参会人员的社交动态与项目背景，将原本枯燥的标准化演示数据替换为客户团队的真实信息与头像，形成极强的信任光环（Halo Effect）。 - **将礼品嵌入销售序列（Sequence）**：Charlie Wiebe 介绍 Nooks 采用了礼品获客机制（甚至向首次预约 Demo 的客户提供免费 AirPods），但核心逻辑在于计算“单次会议获取成本”（Cost to Book a Meeting）是否被礼品带来的效率提升所抵消。物理或虚拟礼品绝不能是孤立的噱头，而必须作为销售序列中的合理步骤（如建立双向联系后送一杯咖啡或 Uber Eats 代金券）。 - **拒绝噱头，紧扣买家旅程**：礼品与互动必须具有相关性、时效性并契合买家旅程。如果盲目给不饮酒的客户送红酒便毫无意义，精准贴合客户需求才是转化关键。 --- ### [03:38:49](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=13129s) 白板思考与重塑流程：从底层业务逻辑出发迈向 AI-First 在 Chris Yeh 幽默展示并分发他亲笔用 Sharpie 签名的《闪电式扩张》（*Blitzscaling*）书签后，现场观众 Erin 提问：除了用 AI 自动化缩短流程耗时外，企业应如何从零开始彻底重塑业务流程？ ```mermaid flowchart TD accTitle: 从白板思考到AI赋能的演进阶梯 subgraph Layer1["底层战略与流程重构"] W["白板思考 (Whiteboard)"] --> J["锚定买家旅程 (Buyer's Journey)"] J --> R["重塑业务步骤 / 去除冗余"] end subgraph Layer2["技术与AI定位"] R --> S{"明确 AI 变革光谱"} S -->|隐形辅助| A1["无感知提效 (路由/自动化分发)"] S -->|范式重塑| A2["重塑工作本质 (如 Cursor: 从写代码到审代码)"] end ``` Eric Seiberling 强调，在他所有的 AI 工具箱中，最强有力的工具其实是**白板**。在加入 MXD Process 的三年里，面对原本只把 HubSpot 当成简陋数据库的现状，他经历了大约 30 次流程迭代。他总结道： - **先思考，后工具**：成为 "AI-First" 组织的关键在于先花时间深入思考买家旅程，理清如何跳过冗余步骤、减轻客户负担并加速成单。如果直接依赖 AI 帮你做战略思考，只会得到与所有人雷同的标准答案。 - **定位 AI 改变工作的光谱**：企业必须在一开始就明确 AI 处于光谱的哪一端。一端是**完全隐形在后台**，让流程自然变顺畅（如网站线索自动智能分配给对应的领地销售）；另一端则是**从根本上重塑工作性质**（如同 Cursor 让开发者的角色从“编写代码”转变为“审查代码”，销售代表的角色也会发生类似重构）。 - **引入新能力而非单纯替代流程**：AI 的真正价值在于为组织带来全新的业务能力，而不仅仅是对既有流程的简单机械替换。 随后，现场主持人抓紧最后时间引入了下一位观众关于“招聘新销售代表时应关注哪些 AI 或技术素养”的提问。 ### [43:47](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2627s) 销售人才画像的三大核心特质：好奇心、聪颖与同理心 在被问及在 AI 时代招聘销售人才看重哪些能力时，嘉宾强调核心考量甚至无关技术本身，而在于底层的人性与认知特质。招聘的首要标准是“好奇心”（Curiosity）与“聪颖度”（Smarts）。具备好奇心意味着销售人员愿意主动学习、拥抱并适应全新的工作方式与技术工具；而由于业务本身极为复杂（例如涉及化学工程等专业背景），必须具备足够高的智商和理解力才能迅速掌握业务精髓与行业知识。 除了好奇心与智商，第三个不可或缺的特质则是“人性的同理心”（Human Empathy）。“因为人们总是愿意向自己喜欢的人买单”，同理心是销售达成的终极基石。好奇心、聪颖与同理心构成了选拔人才的三维画像，企业在招聘流程中甚至会专门针对这三项能力进行测试与评估。 - **好奇心（Curiosity）**：驱动销售持续探索、主动拥抱新技术与新工作流。 - **聪颖度（Smarts）**：在复杂的业务场景（如化学工程与制造流程）中快速理解并构建专业能力。 - **同理心（Human Empathy）**：建立深度人际连接与好感度，奠定交易的信任基础。 ```mermaid flowchart TD accTitle: Modern Sales Talent Profile subgraph CoreTraits["AI 时代销售胜任力核心三要素"] A["好奇心 (Curiosity)<br>主动学习、拥抱新工作流与工具"] B["聪颖度 (Smarts)<br>快速理解复杂业务与工程背景"] C["同理心 (Empathy)<br>建立人际好感与情感连接"] end CoreTraits --> D["构建买家信任与交易闭环"] ``` --- ### [44:34](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2674s) 采购旅程变革下的胜任力重塑：从传递信息到建立信任与信心 索迪斯（Sodexo）在销售人才画像与胜任力模型（Competency Profiles）上进行了深度投入，并严格依据这些模型开展精准招聘。随着采购旅程的深刻变革，超过 90% 的企业买家已经在采购流程中利用 AI 进行前期调研与比价。然而，大量研究依然表明，“人们终究是向人买单”（People buy from people）。正如多年前互联网兴起时，外界曾预言销售这一职业将彻底消亡，但事实证明销售岗位并没有消失，改变的只是买家对销售人员的诉求与期望。 当销售代表面对客户时，其核心价值已经不再是“传递基础信息”，而是向买家输出“信任感与信心”（Trust and Confidence）——“买家买的实际上是信心”。因此，当下的胜任力评估重点在于销售人员是否具备足够的气场与综合素养，能否从容地站在复杂的买方采购决策委员会面前，调动整个集团的完整实力与资源，为买家彻底建立战略合作的笃定信心。 在全场热烈的掌声中，主持人向 Olcay Bulgun Gardner、Charlie Wiebe 和 Eric Seiberling 三位圆桌嘉宾表达了致谢，圆满结束了关于 AI 赋能现代销售运营的深度探讨。 - **买家行为变迁**：超 90% 采购方已在决策链条中采用 AI，信息差红利逐渐消失。 - **销售角色跃迁**：从单纯的“信息单向传递者”升维为“信任与信心交付者”。 - **大客户公关能力**：要求销售能代表庞大的组织资源，面对复杂的跨职能采购委员会建立深度笃定感。 ## Notable quotes > [03:41](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=221s) 超过 90% 的企业买家已经在其采购决策链的各个环节中使用 AI 工具。 — Olcay Bulgun Gardner > [03:41](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=221s) 知（Knowing）与行（Doing）截然不同，洞察（Insights）并不等同于结果（Outcomes）。 — Charlie Wiebe > [07:04](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=424s) 过去通常需要 5 天才能完成的不锈钢储罐报价，现在客户在网页端大约只需 1 分半钟就能搞定。 — Eric Seiberling > [07:04](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=424s) 只要是 Vibe Coding 崩溃而不是真实的储罐在工厂里炸裂，那就没问题。 — Chris Yeh > [22:13](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1333s) 在我们的行业经验中，时间绝对是交易的头号杀手。 — Olcay Bulgun Gardner > [27:00](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=1620s) 一是赚钱，二是让客户满意。 — Eric Seiberling > [03:35:25](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=12925s) 要在客户桌上放一个能发出沉重声响（thunk）的东西。 — Chris Yeh > [44:34](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2674s) 人们终究是向人买单。 — Olcay Bulgun Gardner > [44:34](https://www.youtube.com/watch?v=TNZ2_hy4JdU&t=2674s) 买家买的实际上是信心。 — Olcay Bulgun Gardner ## Mentioned Chris Yeh, Olcay Bulgun Gardner, Charlie Wiebe, Eric Seiberling, Sodexo, Blitzscaling Ventures, Nooks, MXD Process, Pink Floyd, Metallica, Bob Dylan, Joan Jett and the Blackhearts, Level 12, HubSpot, ChatGPT, Claude, Nooks Navigator, Notion, Google Drive, Slack, Cursor, IDEO, AirPods, Uber Eats, Adobe, Cloud Paths, Maroon 5, 《闪电式扩张》（Blitzscaling） ## Video description Olcay Bulgun Gardner - SVP, Revenue Operations, Sodexo Chris Yeh - Founding Partner, Blitzscaling Ventures Charlie Wiebe - Vice President of Revenue Operations, Nooks Eric Seiberling - VP of Sales and Marketing, MXD Process Today’s high-performing sales teams are powered by AI. This panel explores how leading companies are using AI to identify top opportunities, automate admin tasks, and coach reps in real time. From forecasting accuracy to proposal generation, learn how AI is improving speed, consistency, and effectiveness across the sales funnel. Panelists will also address integration challenges and how to ensure reps trust the tools that guide their deals. [Source](https://www.youtube.com/watch?v=TNZ2_hy4JdU)

[3:41] Olcay Bulgun Gardner: 超过 90% 的企业买家已经在其采购决策链的各个环节中使用 AI 工具。 [3:41] Charlie Wiebe: 知（Knowing）与行（Doing）截然不同，洞察（Insights）并不等同于结果（Outcomes）。 [7:04] Eric Seiberling: 过去通常需要 5 天才能完成的不锈钢储罐报价，现在客户在网页端大约只需 1 分半钟就能搞定。 [7:04] Chris Yeh: 只要是 Vibe Coding 崩溃而不是真实的储罐在工厂里炸裂，那就没问题。 [22:13] Olcay Bulgun Gardner: 在我们的行业经验中，时间绝对是交易的头号杀手。 [27:00] Eric Seiberling: 一是赚钱，二是让客户满意。 [215:25] Chris Yeh: 要在客户桌上放一个能发出沉重声响（thunk）的东西。 [44:34] Olcay Bulgun Gardner: 人们终究是向人买单。 [44:34] Olcay Bulgun Gardner: 买家买的实际上是信心。

[3:41] Charlie Wiebe: knowing is different than doing and insights are not the same as outcomes. [19:35] Charlie Wiebe: Does this solve a real daily problem? [22:13] Olcay Bulgun Gardner: time in our experience in our industry is the absolute killer of deals. [33:03] scavenging expedition across the enterprise [35:25] Chris Yeh: If you want to get people's attention, put something on their desk that lands with a thunk. [38:45] Olcay Bulgun Gardner: what processes can we replace with AI [38:45] Olcay Bulgun Gardner: what new capabilities can we bring to the organization. [43:47] Eric Seiberling: people still buy from people they like.

[0:04] So, we're very excited to be here today talking  about the process of using AI for actually closing [0:10] deals. Who would think making money, right?  Something that you're all pretty pretty keen [0:15] on. So, what we're going to do today is we have  a fantastic panel discussion and I'm going to be [0:19] asking a bunch of questions, but you guys are  going to have a chance to ask your questions [0:24] as well. So, when we get to the end, we'll move  into a Q&A session, and you'll get a chance to ask [0:29] your questions. Get them ready now. Experience  tells me that being early is a lot better than [0:35] being a late. Usually, there's a lot of questions  that go unanswered. So, just be ready. Ceue it up [0:40] in advance. And with that, I'm just going to go  ahead and get underway. I'm going to introduce [0:45] the fine panelists that are going to actually  introduce themselves. Uh, so very quickly, [0:50] I'm going to model it. So, my name is Chris Yay.  I'm best known as the co-author of Blitzcaling, [0:55] my friend Reed Hoffman, the founder of LinkedIn. I  also run a venture capital firm called Blitzcaling [1:00] Ventures. And in addition to telling you about  ourselves professionally, we're going to say [1:04] something personal, which is the first concert  that we attended voluntarily ourselves as opposed [1:10] to being dragged to by our parents. And I saw Pink  Floyd at the Oakland Coliseum. They had the giant [1:18] inflatable pigs as part of their show. All right,  Oljai, over to you. Hi everyone. I'm Oli Bulgan [1:25] Gardner, uh, senior vice president of revenue  operations at Sodexo. Been with Sodexo for about [1:31] two years. We are one of the world's largest uh,  food and facilities management services providers [1:37] globally. My first concert was Metallica in  Istanbul, the first ever stadium concert hosted in [1:46] Istanbul. And I was a silly teenager that thought  I could survive being on the floor where there are [1:55] no assigned seats. And I experienced what a human  stampede is like. So back then you were not 7 feet [2:02] tall, 300 lb. You were trampled underfoot. I was  awesome. Charlie, not nearly as dramatic as that, [2:10] but um great to see you all. Charlie Weebi, uh  vice president of revops at Nooks. Uh Nooks is [2:17] a revenue agent platform where uh agents work  together not just to surface information but also [2:24] to execute on it alongside humans. Uh I joined  early last year. My first concert from the Bay [2:31] Area was I think 2002 Bob Dylan and the band or  at least some iteration of the band however old [2:38] they were at the time at uh the Greek theater in  Berkeley. technically dragged there by my dad but [2:46] was a willing participant so I think it still  counts right when it comes to Bob Dylan we make [2:51] allowances and the Eric thanks # worth it anyway  all right my name is Eric Cyberling I'm the VP [3:00] of sales and marketing at MXD process and so it's  an industrial equipment company about $30 million [3:06] so we make stainless steel tanks and mixers that  go into manufacturing lines in order to be able [3:10] to make the things that you love like barbecue  sauce all the way up to aerospace coding. So, [3:14] we do a lot of different things and that's why we  use AI. Um, I'm also the chief revenue officer for [3:20] a tertiary package company to help set up their  business operating system and that's the role I [3:24] play there. So, the first concert I ever went to  is when I was a cadet at the Air Force Academy [3:29] and Joan Jet and the Blackheart showed up and I  got to work the concert. So, that was my first [3:34] concert. I do love rock and roll. I'm so glad  to see this incredible experience on the stage. [3:41] So let's kick it off with a question for all the  panelists and you can choose what order you want [3:45] to answer it but the question is what is  the absolute highest AI priority on your [3:51] desk right now? Go ahead. I'll go first. I know  Brian's in the audience looking at me. What is [3:59] she going to add to my list now? [laughter] Um,  but really our biggest priority is redesigning [4:08] our entire commercial system around AI. Our  buyers are changing, our market is changing, [4:15] more than 90% of buyers are using AI to make their  decisions in every step of their buying journey. [4:22] So as rev ops professionals, our job is to figure  out with all of these buyer trends evolving, [4:30] how do we evolve our goto market organizations and  how do we enable our sellers with AI in all the [4:38] steps of the buying journey. All right, Charlie,  you're helping people do this to some extent, [4:45] right? Talk about that hopefully. Um yeah so I  said uh Nooks is a revenue agent platform and the [4:53] idea is uh agents shouldn't just observe things  they should also execute and I guess there's like [4:59] sort of two fundamental principles that underly  that that Nooks spends a lot of time thinking [5:03] about but so do I and my team is that one is that  knowing is different than doing and uh insights [5:11] are not the same as outcomes right and the goal is  to move AI I on all surfaces from knowing to doing [5:19] and from insights to outcomes. So I think the  highest priority on my desk is to close or shorten [5:28] the learning loop uh for the agents that we use  um by way of execution. So the idea being that [5:37] if you have agents working alongside humans to  actually execute rather than just observe tasks, [5:45] uh they learn what good looks like by doing it  themselves rather than just being told before [5:51] or after the fact. And then you get that  learning loop of oh I understand if this [5:57] was correct or incorrect. And so my job is to  like shorten the distance between uh the initial [6:04] outcome learning and then the next action. And I  guess that's sort of Nook's job at large. Yeah. [6:12] And you know for us I think both what old Jay said  and what Charlie said are kind of the priorities [6:16] for almost all of us as sales professionals  which is you know number one how do I change [6:20] how I operate and how I sell and is in this  new world how do I do that better right and [6:25] then second is how do I go from like oh that's  interesting like an insight into how do I turn [6:29] this into convertible actions that get my sales  people to act right away and so for us you know we [6:35] developed custom GPTs around chemical engineering  and we do very complex sales and we handle you [6:42] 20 30 different industries. And so, you know,  we built the insight engine and now, you know, [6:48] my whole goal is how do I break that out of  the chatbot and get it into ways that drive [6:53] action and for us in HubSpot in order to be  able to move forward and and move much more [6:58] faster around facilitating the sale around these  big capex projects, right? All right. Quick poll [7:04] for the audience. Who in the audience has already  incorporated AI as a standard part of your go-to [7:12] market workflows, your RevOps workflows? Who in  the audience has already done that? All right, [7:17] great. For those who have already done that, how  many people have seen visible evidence of revenue [7:23] uplift from the from the implementation you've  done? Excellent. All right. See? All right. So [7:31] what we're seeing here is you know people are  in a number of different states of adoption but [7:36] we're seeing some people who are driving some  early results which is the important thought. [7:40] So let's talk about specifics. Maybe we can walk  through a specific AI project that you've deployed [7:49] or are working on and what it actually solves  for your revops and what measurable impact it [7:56] has on your revenue engine. So maybe uh Eric, why  don't you start with this one? Sounds great. So um [8:04] we ended up and again I said we had custom GPTs  and so we have that around um tank design. Okay. [8:09] So it's been how we've moved that to an actual  from an insight to an action is we were able to [8:14] go through and build deterministic principles that  we can end up so we ended up really coming through [8:19] and like what are the characteristics of a tank  for something that needs to be sanitary for food [8:23] versus something that is needs to be explosion  proof because I'm dealing with costic chemicals [8:28] and so we we we used we built that using our  chat GPT then we actually use claude co-work [8:33] in order to be able to vibe code a new way to  quote stainless steel tanks now the nice thing [8:38] about that is we know vibe coding breaks and it's  very fragile. What we ended up doing is we had a [8:44] um we had our SI, it was a very small software  company called Level 12 with us. The day we were [8:50] vibe coding and so he was building the back end  as we were building the front. And so in 3 weeks [8:55] we had a good working prototype. And then the next  five weeks was all around validation of the models [9:00] and getting all the data uploaded. And so this  is now 9 days ago we launched something which [9:05] is called tank designer. So, where it normally  takes 5 days to quote a stainless steel tank, [9:10] you can now do it in about a minute and a half.  And so, if you go to designer.mxdprocess.app, [9:16] give it a whirl while you're sitting here.  And I'd love to hear your feedback. So long, [9:20] it's the vibe coding that breaks and not the tank,  right? Yeah. Well, it's not vibecoded anymore. So, [9:25] you know, we're much better now. But yeah, no,  you're absolutely right. Well, I'm curious out [9:29] in the audience, how many people have some sort of  vibecoded RevOps tool in production as we speak? [9:36] Look at that. Pretty impressive. I think a lot  of you could be on this stage as well. Uh so [9:43] when we talk about these tools, obviously one of  the things we're really trying to figure out is [9:48] measurement. How do we know that we're getting  the benefit? How are you measuring that benefit? [9:53] How are you proving it to other senior managers  you might be reporting to? Uh talk through how [10:00] you think about that measurement of AI impact.  Sure. I think in connection with measurement I [10:09] think it's also really important to think about  not just the what is the AI functionality I'm [10:14] implementing here but what is the way of working  what is that system I am driving where AI is part [10:22] of it and what business problem am I solving with  that once we define what that problem is then the [10:28] metrics we want to look at to assess how quickly  we're addressing that problem and the results that [10:34] is impacting become a little bit clearer to  articulate. Um, in terms of ultimate metrics, [10:41] obviously it's about, you know, in a sales  organization, it's about sales and revenue, [10:46] but the ways in which we get there are probably  more important. So, we're looking at velocity, [10:53] we're looking at conversion, we're looking at  um the entire ecosystem that's about getting [10:59] that work done. one of the key areas for our  business where it was actually one of our first [11:05] AI initiatives um within Sodexo RevOps and it's  in our um proposal development motions very human [11:13] inensive time consuming very content heavy copy  heavy I mean we're in the hospitality business so [11:20] our proposals are look a little bit different than  a tech proposal I I I saw that my eyes opened when [11:26] I when I switched careers from the tech industry  to uh to Sedexo. Um, so we needed to do something [11:33] different. At the same time as we leaned into AI  to speed things up and create efficiencies for us, [11:40] we also redesigned the org. We redesigned our  processes and the entire combination of all of [11:46] those four factors actually allowed us to reduce  capacity and inclivity at the same time. So we [11:55] ended up reducing our capacity by 40%. And  improved productivity per individual by 40%. [12:01] So for those who are better at math than me, it's  probably a 1.6x in u overall productivity um that [12:07] we generated. It wasn't just with AI, you know.  So we looked at what are the what are the interim [12:15] measures that ultimately impact sales and revenue  and that was the number of deals coming in through [12:20] the door the velocity the number of days from the  initiation of the proposal motion to the delivery [12:27] of that to to the client and we measured those.  How about the you how about you guys? What is [12:33] the measurement that you're really focused on? Uh  yeah, I mean I can Do you want me to go into like [12:42] a more detailed example? Whatever you think will  help convey the point you'd love to make. Okay, [12:46] then yeah. Um maybe I'll talk about a project  then that we're working on or we're working [12:52] on recently. Um and that is uh account scoring.  So for folks who are unfamiliar the at its most [12:59] basic you want to score your database of accounts  so you can sort of orient your sales force around [13:05] this is what a high fit account looks like for  Nooks this is what a low fit account looks like [13:09] then your marketing team can provide air cover  you're all speaking the same language right [13:13] um and I would say for years and years in revops  account scoring itself like getting a score was [13:19] sort of the destination right it took months if  not quarters of work a lot of spreadsheets a lot [13:25] of probably duplicates, a lot of bad enrichment  data, and you were just putting a lot of work [13:29] into like the actual fundamentals of building the  score. Um, this year, or I guess going into 2026 [13:36] when we were rebuilding our account score, uh, the  like measurement we cared was really like pipeline [13:42] per SDR and revenue close per AE. Uh, and a lot  of that comes down to like are we just working [13:48] the right accounts to begin with or are we wasting  sellers time working poor fit accounts? Um, and [13:54] rather than spending all of our time as a RevOps  team working in spreadsheets and just trying to [13:58] get to the account score, we uh invested a whole  lot of capacity in uh actually just building a [14:04] single skill and doing a lot of prompt refinement.  Uh, and the whole goal of the scale was to stand [14:10] up a cohort of agents that would build its own  account scoring model and its own account scoring [14:16] like interface application essentially. And so  we built the scale. Uh it created agents. The [14:24] agents created the model per tier. They created  the uh like evaluation harness of like this is [14:32] what a gold customer looks like. Uh they tiered  the entire database so that operationalized it as [14:38] well. Scored every account in HubSpot, identified  whitespace accounts not in HubSpot, added them to [14:42] the database, so on and so forth. uh and uh again  like iteratively tested and the I would say the [14:50] like interesting part that's interesting I guess  in itself but uh the more interesting part is gets [14:56] back to something I was talking about earlier of  like closing that learning loop um is that those [15:02] same agents that were like generated from this  skill that spent all of this time and gathered [15:06] all of these memories quote unquote about like how  the score was built are then the same ones that [15:12] uh reps are using to uh when they start their day  on Monday morning say like, "Oh, great. I'm going [15:18] to try and go sell to Sedexo today." Like, okay,  those same agents then like help do the research, [15:23] help uh find the right people to try and sell  to. Uh and they then execute those tasks at scale [15:30] alongside humans. Humans do what humans do best,  which is uh hold themselves accountable, create [15:37] relationships, exercise judgment, and then those  same agents learn like oh this task I proposed [15:44] like this human let me do it without any feedback  or I didn't do this task at all because it was [15:49] not a good task to do. And then those same agents  put all of that back into like the same evaluation [15:55] harness that they created the model with. And you  get this like fast learning loop of both creating [16:01] the score actually executing on it rather than  just observing how it's executed on and then [16:08] this like basically live iterative process. So, I  [snorts] don't know if that answered the question, [16:12] but No, it definitely did because I think one of  the key points here is it's not just about, oh, [16:17] what is the number, what is the velocity, what is  the sale, but it's about the improvement process, [16:24] the iteration and being able to actually compound  those advantages. But of course, you mentioned AI [16:30] and humans working together. That's actually one  of these things that I think we all face, which [16:35] is you can build out all this infrastructure.  you still got to get your reps to use it. So, [16:42] Eric, maybe this is a question for you since you  deal with reps who are like selling somehow both [16:48] industrial lubricants and hot sauce, I guess,  which I hope they don't get mixed up. Um, but [16:55] how do you get reps like that with such different  backgrounds to adopt these tools? Because they [17:00] don't seem like they'd be super duper like hot  sauce salesmen don't strike me as very computer [17:04] literate. Well, they don't sell hot sauce. They  make the equipment that makes the hot sauce. So, [17:09] and it's, you know, when you talk about big  stainless steel tanks, but we have people who [17:12] have been selling this stuff for 30, 40 years,  okay? And so, when you start to think about that, [17:18] I mean, when they started their career, they  didn't even use, you know, computers. They [17:23] walked around with a sales book and just went  and talked to customers and and that was it. [17:28] And so we've got everybody from there into a  brand new college graduate that comes in who [17:33] is much more familiar and comfortable. And  adoption's been uneven. And let's face it, [17:38] I think that's probably one of the longest polls  in the tent when it comes to AI is it's we don't [17:44] live in a field of dreams world. You build it and  they will come. It really comes into the fact I [17:48] think number one is, you know, do you have the  right model? Do you have the right framework in [17:52] thinking? Do you have the right business process?  Do you have all those things? But then where we're [17:57] moving now is really because we've seen this be an  issue because we've had uneven adoption is how do [18:01] we start to bake it inside the tool that people  don't even know they're using AI anymore that [18:06] it's just the way that work gets done. And I think  particularly when you think about um our industry, [18:12] you know, we're about 20 years behind everybody  else, which means now we're only 10 years [18:15] behind. But I I think it's one of those things of  anybody who, you know, as you're running a team, [18:22] it's it's how can you bake it into the process  so they don't even know, but it ends up being [18:26] very valuable for them. And so that takes constant  revision and work and rework, but I think that's [18:32] the only way that you really get it adopted inside  particularly small and mid-size companies. And I [18:37] think that's sort of the one of the fundamental  jobs of RevOps now, right? Is to like actually [18:42] build that infrastructure where you can work side  by side in an effective way. Is that fair? Yeah. [18:47] Oh, abs. Absolutely. It's it's the now I'm I'm  a oneman show when it comes to technology. So, [18:52] I'm running the business and all the commercial  functions and I have one IT guy who ends up [18:56] running our network. So when it comes to AI  innovation, it's like I teach myself and then [19:00] I'm like, okay, let me how do I hardwire this in  the organization where a dumb person like me can [19:05] do that and maintain it because I'm now the I'm  now the the fragile factor. But it's it's I think [19:11] that's going to be the way of working is not going  to be more and more and more tools. It's going to [19:16] be more things that can integrate together so  it becomes invisible to the user. Kind of like [19:20] electricity. Um for this we're not trying to  determine how to how to wire our buildings. [19:25] All we're trying to do is figure out what kind of  light do we need and then how do we use it to surv [19:28] to to survive. Same thing as I think is in this  case. I'll add that the same thing is very valid [19:35] for large complex enterprises as well probably  even more so to be honest because the number [19:40] of people you're trying to drive adoption with  and the amount of intricacies you have and the [19:45] dependencies with other adjacent teams the level  of complexity just increases. Yes, you're always [19:52] going to have the first, you know, 20 25% of your  users that are going to be out there and saying, [19:57] "Yes, teach me these agents. I I want it. I I  will use it." And then everybody else, you have [20:02] to find a way to get everybody else on board. And  and and we have to achieve that seamlessness into [20:08] the way they work. If I can also add one more one  more thing in in my experience having worked with [20:15] sellers and sales organizations for a good chunk  of my career. They have to be part of the process [20:20] design. Absolutely. Right. If you don't bring them  into a new tool or a new process, a new system [20:26] you're you're looking to implement, um adoption  is going to be way more challenging afterwards. [20:31] just creating that intimacy with your sales  organization, making sure that they're part of [20:36] that process, making sure that you have eyes into  their workflow and and the enablement factor is um [20:44] embedded from the very first step will inevitably  generate um better results for you because you [20:51] have champions along the way. Can I just add  one thing on the point of measurement? Um, [20:56] a great point and actually one of the like most  important metrics or outcomes for my team is our [21:02] like basically our internal RevOps NPS. So the  simplest form of that question is like hey STR [21:09] hey AE does what RevOps build help you do your  job? Like does that does this feel useful? Like [21:17] are these things you actually use? And of course  the answer we're looking for is yes. And there are [21:21] more complex ways to ask that question, but it's  like it's when you can vibe code something in like [21:27] 6 hours, it becomes really easy to just end up  building things that like don't actually matter or [21:31] aren't going to get used. Um, so that's become a  much more important like success criteria is like, [21:35] oh, like do you feel like the RevOps team like  actually builds things that solve a real problem [21:40] in your day? Yeah. And I think what Jai said and  and then we'll wrap this piece up, which is I I [21:46] try to put two extreme users in the room. So those  of you are familiar with IDO's design methodology, [21:50] you look for the extreme users, not the average.  And so we had one person who's very techsavvy and [21:54] I have one person who's very not. And we have  them both in the room when we're vibe coding or [21:58] trying to build a new tool because if I can end up  meeting the needs, whether it's one tool or two, [22:02] depending on what we're doing, you can end  up being much more successful up front. Now, [22:07] one of the things is the title of this session is  closing smarter. So, what I'd love to hear from [22:13] you guys are some concrete examples of how your  teams or the teams that you work with have been [22:20] able to leverage AI to close deals smarter,  faster, better, close deals that maybe they [22:26] wouldn't have closed before. Yeah, I think  for us it really has been we cover a myriad [22:32] of industries and so and we have a sales rep that  can be handling all sorts of different complicated [22:37] deals whether it's high shear mixing or whether  it's um an oil field where we're putting in mixer [22:42] it's what they're called big side entry mixers  that go to make sure the oil doesn't settle and [22:47] those things and so you know where we've seen the  most help is we've been able to enable them with [22:52] the knowledge so we do the research so they don't  have to so before they get on the phone with a rep [22:57] with a with a customer. It's to give them it gives  them the brief. It says based on what they're [23:02] trying to make, here's the list of equipment  that they're probably going to use. Here may [23:06] be the competitors that they use and then we have  a list of discovery questions and just doing that [23:10] because we deal with I've got seven reps and we  have 12,000 leads a year. Okay? And so this huge [23:16] volume. So the more work that I can take off their  plate so they can really focus on on on two things [23:21] which is number one how are they the absolute most  helpful they can be to improve their process and [23:26] that requires a lot of knowledge and that's how we  educate them so that they're enabled upfront and [23:31] then number two is how fast can we quote because  quoting is really important for us. So just by [23:35] doing the work we've done so far, we've gone  from about average six days to quote to under a [23:40] day. And now with Tank Designer, now we're in the  minutes. And so we we kind of first in line with [23:45] that. And because that speed to deal and the depth  of understanding, we're you know, our close rates [23:50] went up 10%. Couldn't agree with you more there.  You know, time in our experience in our industry [23:57] is the absolute killer of deals. Honestly, we look  at in our pipeline, there are a number of deals, [24:03] unfortunately, that uh don't end up going  anywhere, right? They are always the ones that [24:09] take the longest, right? So, we we're leveraging  AI to get there faster. And I think Charlie, you [24:15] made the point earlier about the distance between  learning and action, right? There is so much time [24:23] spent um without leveraging AI and compiling  the information, compiling the benchmarks, [24:29] the pricing data, the costing data and and and  you're spending time instead of getting to the [24:35] what's the next action I should be taking here.  The more we can shorten that distance is there's a [24:41] direct correlation to revenue there and conversion  rates. Want to add something? I mean, I can answer [24:47] the question if Go. Yeah. Um, I'll be uh quick.  Um, there's an in So, uh, we run twoe trials [24:55] for basically every customer we end up closing.  We have a like 35% overall win rate. Um, if for [25:03] customers who get to trial though, we like an 80%  win rate. So, we lose almost every deal that we're [25:08] going to lose. It's because we don't get somebody  to trial because the idea is basically like, oh, [25:12] like what's your like call to conversation rate  and conversation to meeting rate right now? It's [25:17] like it's this like okay liter like use Nooks for  2 weeks put your STRs on it and it's going to be [25:22] double and if that happens will you buy it? The  answer is like yeah okay great and then it happens [25:26] and great. Um and trials are very like intensive  to manage though. So we have AES like sitting in [25:32] shared Slack channels joining call blocks with  SDRs. They're running three, four of these at [25:38] a time in a week and it like takes a lot of time  and customers or prospects I guess at that point [25:42] have [snorts] complex nuance questions about the  product and that information lives across notion [25:47] and Google Drive and spreadsheets and PowerPoints  and a whole bunch of different things. Uh so we [25:53] built something uh that's called Nooks Navigator.  Open to new ideas. If anyone has any better names, [26:00] come find me after. But basically it lives in  Slack. uh agent for those who are more technical. [26:06] It basically uh takes all of our databases, it  like vectorizes them so it can be semantically [26:10] searched and then it's like federated enterprise  search on top of that. So think about like you [26:15] just ask a question in natural language, it  finds uh the best data source for the answer and [26:22] it gives you the answer. And what reps can now do  is if a customer asks them a question of like can [26:27] I [snorts] make calls from an Australian number  if I'm in Brazil into Germany? I don't know, [26:32] but like it lives somewhere in some presentation  probably. You can just either ask the question [26:38] to Nooks Navigator in a side chat in Slack or  you can just forward the Slack message to it. [26:43] It provides the rep the answer. They can approve  or modify it and then send it to the prospect. Um, [26:50] and that's something that's like I would say is  has helped the answer to the question of like does [26:55] Robots actually help you at all? Be yes. Yeah. All  right. I'm going to ask one more question and then [27:00] I open it up to the audience. So, we'll get our uh  our audience microphone ready. So, the question is [27:07] this. We've heard about AI. We've heard about the  SDRs, but there's also nonAI software that's still [27:14] out there. So, what is AI's role in these systems  of work? How is it going to interact with existing [27:21] software or new software that's not AI and the  humans involved? How do these all work in concert? [27:30] I'll start with that one. As the  least technical person in this panel, [27:35] um I think it's really about the most effective  way to bring together your people, your processes, [27:44] and your technology. Right? We talk about  AI. Often times when we talk about AI, [27:49] we're actually not even talking about AI. Um it's  it's it's about how do we weave those together to [27:56] create better systems of working. So whether  it's uh you know different type of coding or [28:03] conversational intelligence etc. It's really all  about what is the growth mechanism you're driving [28:08] and how are you going to do it and me being on the  business side here there's a very important part [28:14] of this equation and that's our partners in tech  right we can't do it without each other. We've [28:21] created a very very integrated working model  internally within our company where tech and [28:29] the business are in it together. Right? I'm the  one with the business problems and the vision for [28:37] how I think we can achieve there. Then my team  and our tech teams and our leaders in tech are [28:42] working together to solution it better. Um, like  I said, I'm not the one to say which tech or which [28:48] AI is the better solution, but I think it's about  finding the way to open those doors to figure out [28:54] what's the best way to solve that business problem  and then making sure you have the experts on your [28:59] team and you're pulling them in and your users,  right, your sellers and everybody else that's part [29:04] of your commercial ecosystem that are part of it  um to come up with that solution together. Yeah, [29:10] I was just going to say I don't uh I wouldn't say  this is like a a new problem, right? It's just a [29:16] different framing and there's like more potential  tools in the toolbox to address it. I think the [29:20] fundamental issue is still fragmentation. And  the thing about like AI and agents specifically [29:26] is the whole idea is you're trying to like build  a robot to think like a human, right? Like that's [29:33] the whole idea of like an agent. It's like oh like  Claude can use Chrome now. that's like mimicking [29:38] what a human does. So then you have to think  about like okay where do hum what's the failure [29:43] state for like a human not knowing where a piece  of information lives like your agents are going to [29:48] run into those same walls and this is somewhat  the answer or what I was just talking about of [29:54] like how do AI and non-AI tools interact. I think  the like underlying question there is where does [30:01] your information live? And you we prefer to solve  it from like a could a human figure out the answer [30:07] to this question and if the answer is no then  like an agent's either going to have a hard [30:12] time or it's going to hallucinate its way into  thinking it has an answer. Um but at the end of [30:18] the day like I think yeah AI is very good at like  summarizing and creating recommendations in these [30:25] things but like execution is still the bottleneck  and if you can put an AI layer over your nonAI [30:32] tools I think the like biggest impact it can have  is in actually executing with like human judgment [30:39] versus like I don't need another summary of  how the meeting went you know and Sorry. No, [30:46] for you. So anyway, yeah, I think it's I think we  get in this AI versus not AI debate all the time. [30:52] And I think it's a silly one. And the reason why I  say that is, and this is going to be honest, it's [30:56] a means to an end. And so, don't forget the end,  right? What are we here to do as sellers? We're [31:02] here to make money. That's our job. And we're here  to make customers happy. And so, how do I do that? [31:08] What decisions do I need to make? And then what's  kind of my flow in order to be able to get there, [31:13] right? And so that's where it's like, you know,  can I use my AI tool to get my non-AII tools to [31:18] cough up the data I need and put it in a way that  a human can then make a real decision that impacts [31:23] the business. And I think for everybody, don't  get caught in the shiny object syndrome of AI, [31:29] right? And don't get into agent, agent, agent,  agent, agent. No, it's it's then you're just [31:34] going to have a bunch of people screwing around  that don't know what to do. You really do have to [31:38] think about this is the business that I'm in.  This is what my customers care about. This is [31:41] the data that I need to make decisions. this is  how I'm going to go about it and then say, okay, [31:46] both AI and nonAI tools, how do I orchestrate them  in order to do that the easiest way that frankly [31:53] makes humans lives easier? And I think that's  really the goal. All right, who is ready with a [31:59] question? I want to see a good volunteer. I see  the first one got his hand up back there. Let's [32:04] get that man over there. Yes, you. You could see  me. Yes, the microphone is coming your way. I'm [32:10] just stalling until the microphone gets there and  I see a couple of other hands up. Don't worry, [32:14] we'll try to get to as many of you as possible.  When you ask the question, please say who you are, [32:18] who you work for as well. If you want to say your  concert, you can. My name is BV. I work for Adobe. [32:25] Um question on pull versus push. Uh how did you  make the transition where even now when previously [32:33] people would come to a dashboard or a surface to  get the data? uh what are you transforming within [32:40] your organizations in terms of push information  to the sellers, push information to the [32:46] um sales teams? I can go first. Yeah. So for us, I  mean in in in the our CRM, we've got what's called [32:54] a sales workspace. And so our goal is to push all  of the data that they need and kind of like here [32:59] are the calls that you need to make. Here are the  follow-ups that you need to make. Here are where [33:03] deals are starting to become stale and really  think through what does that seller need to be [33:07] able to do their job and do their job well and  push everything that they need to them because [33:12] for for my team at least if if they have to go  out and go on a a scavenging you know expedition [33:18] across the enterprise it's not going to work. And  so it's really understanding starting with the rep [33:24] and like what's the desired experience for them  to be able to be the most successful and pushing [33:30] everything I can to them. Now it just takes time,  right? And it takes understanding, but all we're [33:35] trying to do is like kind of like step our way  there every single day. But that's what you have [33:39] to do because um I just think that's the best way  for us to solve it in our industry. Yeah. As I [33:46] said earlier like the two things uh like knowing  is not equal to doing and insights are not equal [33:52] to outcomes. If there is a thing to be known so  like a piece of data and a robot can know it and [34:00] then take some action on it without a human ever  being in the loop like great it's not push or pull [34:05] it should live somewhere else. uh if there uh is  some human judgment to be exercised then I think [34:12] it should largely be a push motion to go from  insight to action. So make it easily available, [34:19] right? Be able to atment mention something in  Slack to like answer a question if you have it. [34:23] But the only things that should be coming up uh  proactively are where like a judgment call needs [34:29] to be made. But again, you should have the robot  executed if you can. I think and to add to that, [34:36] I think a key foundation that you want to have in  place to be able to start pushing information to [34:42] your sellers is you need a common language across  the board. So might be a little less relevant for [34:47] those in um SMB or younger companies. Uh but if  you're in a large complex enterprise with a lot [34:54] of you know process history and and tech history  um the the not having a common language becomes a [35:04] becomes an inhibitor to being able to push that  information because then you don't know what [35:07] you're going to push. In our organization, the  the first 12 to 18 months of our transformation [35:13] was really spent on creating that common language  and common framework and that's allowing us to [35:19] speed up uh some of these enhancements with AI and  some other tools that we're pushing now into the [35:25] organization. Awesome. We've got a question over  there. Hi, my name is Shrar. My favorite concert [35:32] is Maroon 5. Maroon 5. Ah, I'm a seller at Cloud  Paths. I get lot of cold mails every day. In fact, [35:42] I got cookie box from strange seller, unknown  seller. While it is annoying, I ignore. How do [35:49] you guys see a really addressing this whole calls  and coal marketing? Yeah. So I can also I can [35:59] also talk about that a little bit just because  in my earlier life I was a VP of marketing and [36:04] I do remember actually sending boxes of fortune  cookies to people. I always said if you want to [36:09] get people's attention put something on their desk  that lands with a thunk. And so the thing is that [36:16] kind of approach happened long before AI. Now what  we really should be seeing is the use of AI to do [36:23] the sort of deeper personalization. Uh, one of the  examples I use is when I was a VP of marketing, [36:29] we would pitch to people. Uh, we were pitching  an internet project to people. And what I did is [36:35] I looked up all the people who were going to be  in the meeting, I looked at their social media. [36:41] I looked up all the information about the projects  they've been working on. And when we showed them [36:44] the demo, instead of standard demo data, it was  all of them, their pictures, their knowledge, [36:49] their information. Now, I did that because that  was the biggest deal we're ever trying to win. [36:53] But you normally don't have time with AI, I think  you do. And so the personalization is the part [36:58] where they demonstrate they really understand  you and your needs. And again, demonstrating [37:03] they understand your needs comes in part from the  halo effect of demonstrating they understand you. [37:07] Other thoughts? I mean, we use gifting. We have  a gifting platform that we use. Uh I mean, if you [37:14] go to our website right now and request a demo,  we will give you a free pair of AirPods just for [37:18] coming. I'm serious. Any Does that apply for me,  too? That applies for you. So long as you haven't [37:22] taken a meeting with us before, um, we should, the  way we think about it is like what's the average [37:27] cost to book a meeting today? Can you incorporate  gifting into a sequence that and then is like the [37:33] cost of the gift going to save enough other steps  in trying to book a meeting such that it offsets? [37:39] And if the answer is yes, then great. Like give  somebody a Starbucks gift card for coming to the [37:44] meeting, have lunch with them virtually, give them  like an Uber Eats gift card. But um I really the [37:50] way we allow or like strongly recommend reps to  use uh gifting and like physical gifts or virtual [37:56] gifts today is to have it be a step in a sequence  rather than just a standalone thing. Uh and that [38:04] step should have a rationale behind it. It should  be like LinkedIn connection request accepted. We [38:10] have one birectional communication. Now I'm going  to like send you a cup of coffee rather or again [38:16] book a demo. go get some AirPods. Yeah. I mean,  I think the key thing is don't make it a gimmick, [38:20] right? It needs to be part of the buyer's journey.  And so, you really have to understand that buyer's [38:24] journey and do what's relevant and meaningful and  timely for what you're trying to do. And I think [38:30] that's really the key of it. So, you know, I could  send somebody a bottle of wine, yay, but they may [38:35] not drink, right? And all those things, but if I  instead do something that's very relevant to them, [38:39] then I think it's going to be much better off.  And you can send me one if you want, Eric. It'll [38:44] be bourbon. So yeah, that works too. All right, so  are there any questions from this section? Because [38:49] this section has not yet asked the question. I  always try to be fair to everyone. All right, [38:54] right in the front. So we're going to have to  sprint around. I apologize. I'll stall in the [38:58] meantime. Uh so I mentioned before that I'm an  author of a book called Blitzcaling. I carry [39:04] with me the signed bookmarks. These are not via  auto pen or printed. I actually sign them with [39:08] a Sharpie myself. I find it encourages people to  buy the book. So if you want one of these, come [39:12] to me afterwards. All right, that stalling worked  perfectly. Go. Thank you. Uh, I'm Erin. I missed [39:19] XOO as well. And I just have a quick question. Uh,  pretty impressive plant. Pretty impressive. It's [39:23] not at all. Ask her. [laughter] We don't work that  way. Um, my question to you is um I'm like very [39:32] inspired by your story of like the automation, how  you go from weeks to minutes, etc. And I feel like [39:38] when it comes to implementing AI, it's an easier  putt. Although like old was saying it's harder at [39:43] you know enterprise companies to kind of automate  things. It's not as you know like the whole agents [39:48] thing and stuff like that. But for you guys for  you all I think it's easier to implement AI to [39:55] make something easier but how about rethinking  something completely from and I know this is [40:02] not the first time you heard this but I'm just so  intrigued about that portion and your experience [40:07] rethinking things from scratch. Yeah. So I mean  for us my favorite tool in my entire toolbox, [40:14] my favorite AI tool is a whiteboard. Okay. So  the the first I've been been at MXD process for [40:20] about three years. When I showed up, they had zero  standards. They you know was they they basically [40:26] used HubSpot was nothing more than a database and  that was it. And I must have went through about 30 [40:31] iterations of the new work process. And I always  kept looking say okay number one a how can I how [40:37] can I do this differently? You know, first of all,  I grounded myself in the buyer's journey. Then, [40:41] how do I do this differently? And how can  technology help me skip steps or do things [40:45] in a smarter way? And I think that's actually  the most important tool any person can have in [40:51] their office is a whiteboard and think, take the  time to think because it's how the way you become [40:57] an AI first organization is really, and I know  everybody's talking about that now, right? But [41:02] it's it's just transform your business so things  are easier for your customer. It's easier for you [41:07] and you can close the sale faster. But you've got  to spend the thinking time to do that before you [41:11] ever look at technology one or like this kind of  chatbot or co-work or anything else. If you rely [41:17] on AI in order to do that thinking for you, you're  going to get the same answer as everybody else and [41:23] it's going to be doing it the same way it's always  been. So unless you take that time for yourself [41:27] and really think strategically through your  business, you're not going to change. Everyone [41:33] uses the same model. So what you bring to it is  what differentiates you. Go ahead. Yeah, I mean [41:38] I I couldn't agree more and I I love the question.  Uh, I think one of the most important things when [41:46] using AI at an organization at large, let alone  in go to market is to be really intentional and [41:53] clear upfront with yourself and with your sellers.  Where on the spectrum of how does this change my [41:59] job are you falling? And one end of the spectrum  is AI is like so far in the background that like [42:06] nobody has to even know that it's there,  right? Like things just like feel easier, [42:11] right? So maybe it's how like prospects book demos  on your website. It gets them to like the right [42:17] seller based on the right territory faster. It's  like nobody has to know. This thing just like is [42:22] easier and works better. Uh that's one end. The  other end is like fundamentally restructuring not [42:28] only how you do your job but like how you think  about doing your job. So like simplest example is [42:34] something like uh cursor co-pilot, right? Like  a developer's job is no longer to write code, [42:40] it's to review code. that's like a fundamentally  different thing and the same thing applies to [42:44] sellers. So being really upfront of like where on  that spectrum is this tool roll out I'm proposing [42:51] and making sure that like that is what people want  and then enabling them as such and supporting them [42:57] like to the right degree based on where you  fall I think is if you're not doing that up [43:02] front you're going to end up realizing where you  are on the back end probably because something's [43:08] gone wrong. People say like, "Oh, why did you do  that if I don't even notice it?" Or like, "Wait, [43:12] this is a different job to be done." You know,  to me, AI is about what capabilities, what new [43:19] capabilities can we bring to the organization  versus what processes can we replace with AI. [43:25] It's a very big difference between the two. Yeah.  Got it. We have time for one final very quick [43:31] question. Go ahead. This gentleman over here  raised his hand. When you're hiring new reps, [43:36] and he's got a loud voice. I like that. Sorry.  Thank you. When you're hiring new reps, um, [43:41] what are you looking for related to either AI or  or anything? What's the main thing with technology [43:47] you're looking for? White, by the way, it's  it's not even technology. I want somebody that's [43:50] curious and I want somebody that's smart. Okay?  Because if I have those two things and they're and [43:55] so if they're curious, they're willing to learn,  they're willing to adapt new ways of working, [43:59] all those kinds of things. And number two is  they've got to be willing and you know, our [44:03] business is really complicated. I need somebody  that's smart enough that can pick that up, [44:07] right? I don't, you know, chemical engineering  background is great or those kinds of things, [44:11] but I think it's those two things. And then third,  I think because I always have to talk in threes [44:15] is around just human empathy, right? And I think  those are that's very important for selling period [44:20] because people buy from people that they like.  And so as I look at that, it's really it's yeah, [44:24] it's curiosity, it's smarts, and it's going to  be empathy are the three things I hire on. We [44:29] actually test for it. So I have nothing  else to add. Curiosity and smart. Yeah, [44:34] that's it. I I will add this this is an area we've  actually invested a lot in at at Sedexo and really [44:41] designed competency profiles for our sellers and  and we're recruiting to those profiles as well, [44:48] especially with the buyer journeys changing,  right? Earlier I I I shared that more than 90% [44:55] of buyers are already using AI in their buying  journey. So, and and there's still overwhelming [45:01] research that people buy from people. So, and  years ago, you know, with the with the introducing [45:08] of the internet, you know, people were saying  the the job of a seller is going to go away. [45:13] It's not going to go away. But what your buyers  expect from your sellers is changing, right? Uh [45:20] um yes. Uh when when your seller is in front of  your your buyer, it's no longer about giving them [45:27] the information, but it's about giving them the  trust and the confidence. They're they're they're [45:32] buying the confidence. So, those are the types of  competencies we look for in our sellers. Are is [45:37] this person going to be the person that's going  to be able to show up in a room in front of a [45:42] group of buyers and complex buying communities and  really bring all of Sedexo to them and establish [45:49] that confidence with the buyer. Awesome. Let's  have a big round of applause for our panelists, [45:54] OJ, Charlie, and Eric. Thank you guys  so much. Enjoy the rest of the show.