ServiceNow CEO Bill McDermott：AI 思考但工作流执行，企业平台比 LLM 构建便宜 10 倍ServiceNow CEO Bill McDermott: AI Thinks, Workflow Acts — Enterprise Platforms 10x Cheaper Than Building with LLMs
核心要点：在 AI 时代，像 ServiceNow 这样的成熟企业工作流平台比单纯用语言模型从零重建核心业务流程能提供更大价值和可靠性。
ServiceNow CEO Bill McDermott 是一位从青少年时期买下熟食店起步、后来执掌 SAP 和 ServiceNow 等大型平台公司的资深领袖。他从数十年以客户为中心的经验中总结道，虽然 AI 加速创新，但企业需要确定性的工作流、完整上下文和治理机制——这是纯 LLM 难以单独提供的，且成本高昂。
McDermott 指出，在考虑 GPU 基础设施、token 费用、人力资本转移以及跨 HR、财务、合规等部门的集成重建后，即使是 ServiceNow 平台上的简单应用，用语言模型复制也要贵大约 10 倍。"经营企业的人明白人会犯错，但他们永远不会原谅软件犯错。" 他将 ServiceNow 定位为 "AI 控制塔" 和 "中央神经系统"，能集成超大规模云、语言模型、记录系统，并通过 Armis 扩展安全能力，从而实现真正的 agentic 业务。
对话强调人与人之间的连接和责任感不可替代：AI 的作用是放大人类雄心，而非取代它。早期采用者正超越实验阶段，通过 agent 处理战术性工作，让人类专注于判断和关系，建立新业务模式。ServiceNow 自身已用 AI agent 处理 90% 的客服案例，并预计尽管增长但头寸保持平稳，得益于生产力提升。
直接引用："AI 思考，但工作流执行。" 这一理念解释了为何拥有深厚上下文和执行能力的平台业务即使在生成式 AI 重塑软件的时代仍具有强大护城河。
链接：https://www.youtube.com/watch?v=tNNFJa5pUEg
ServiceNow CEO Bill McDermott 是一位从青少年时期买下熟食店起步、后来执掌 SAP 和 ServiceNow 等大型平台公司的资深领袖。他从数十年以客户为中心的经验中总结道，虽然 AI 加速创新，但企业需要确定性的工作流、完整上下文和治理机制——这是纯 LLM 难以单独提供的，且成本高昂。
McDermott 指出，在考虑 GPU 基础设施、token 费用、人力资本转移以及跨 HR、财务、合规等部门的集成重建后，即使是 ServiceNow 平台上的简单应用，用语言模型复制也要贵大约 10 倍。"经营企业的人明白人会犯错，但他们永远不会原谅软件犯错。" 他将 ServiceNow 定位为 "AI 控制塔" 和 "中央神经系统"，能集成超大规模云、语言模型、记录系统，并通过 Armis 扩展安全能力，从而实现真正的 agentic 业务。
对话强调人与人之间的连接和责任感不可替代：AI 的作用是放大人类雄心，而非取代它。早期采用者正超越实验阶段，通过 agent 处理战术性工作，让人类专注于判断和关系，建立新业务模式。ServiceNow 自身已用 AI agent 处理 90% 的客服案例，并预计尽管增长但头寸保持平稳，得益于生产力提升。
直接引用："AI 思考，但工作流执行。" 这一理念解释了为何拥有深厚上下文和执行能力的平台业务即使在生成式 AI 重塑软件的时代仍具有强大护城河。
链接：https://www.youtube.com/watch?v=tNNFJa5pUEg
The Takeaway: In the age of AI, established enterprise workflow platforms like ServiceNow deliver far greater value and reliability than attempting to rebuild core business processes from scratch with language models alone.
ServiceNow CEO Bill McDermott, a veteran leader who rose from buying a deli as a teenager to running major platform companies including SAP and now ServiceNow, draws from decades of customer-centric experience. He argues that while AI accelerates innovation, businesses need deterministic workflows with full context and governance — something pure LLMs struggle to provide without enormous hidden costs.
McDermott highlights that replicating even a simple application on ServiceNow's platform with a language model would cost roughly 10 times more when factoring in GPU infrastructure, token expenses, human capital diversion, and rebuilding integrations across departments like HR, finance, and compliance. "People that run businesses understand that people make mistakes. They never will forgive software for making a mistake." He positions ServiceNow as the "AI control tower" and "central nervous system" that integrates hyperscalers, language models, systems of record, and now security via Armis, enabling true agentic businesses.
The discussion underscores that human connection and accountability remain irreplaceable: AI serves to amplify ambition, not replace it. Early adoption is accelerating beyond experiments, with first movers rethinking headcount, business models, and ROI through agents that handle tactical work, freeing humans for judgment and relationships. ServiceNow itself uses AI agents for 90% of customer service cases and expects flat headcount despite growth thanks to productivity gains.
Direct quote: "AI thinks, but workflow acts." This philosophy explains why platform businesses with deep context and execution capabilities hold strong moats even as generative AI reshapes software.
Link: https://www.youtube.com/watch?v=tNNFJa5pUEg
查看原文 →
ServiceNow CEO Bill McDermott, a veteran leader who rose from buying a deli as a teenager to running major platform companies including SAP and now ServiceNow, draws from decades of customer-centric experience. He argues that while AI accelerates innovation, businesses need deterministic workflows with full context and governance — something pure LLMs struggle to provide without enormous hidden costs.
McDermott highlights that replicating even a simple application on ServiceNow's platform with a language model would cost roughly 10 times more when factoring in GPU infrastructure, token expenses, human capital diversion, and rebuilding integrations across departments like HR, finance, and compliance. "People that run businesses understand that people make mistakes. They never will forgive software for making a mistake." He positions ServiceNow as the "AI control tower" and "central nervous system" that integrates hyperscalers, language models, systems of record, and now security via Armis, enabling true agentic businesses.
The discussion underscores that human connection and accountability remain irreplaceable: AI serves to amplify ambition, not replace it. Early adoption is accelerating beyond experiments, with first movers rethinking headcount, business models, and ROI through agents that handle tactical work, freeing humans for judgment and relationships. ServiceNow itself uses AI agents for 90% of customer service cases and expects flat headcount despite growth thanks to productivity gains.
Direct quote: "AI thinks, but workflow acts." This philosophy explains why platform businesses with deep context and execution capabilities hold strong moats even as generative AI reshapes software.
Link: https://www.youtube.com/watch?v=tNNFJa5pUEg