Indeed Shifts 70% of Hires to AI Matching as Tech Postings Rebound

Indeed 将 70% 招聘转向 AI 匹配，科技岗位需求回暖 企业级 AI 推荐系统目前驱动了该平台 70% 的招聘，并通过以行为能力匹配取代刻板的关键词筛选，扩大了对软件工程的需求。 Anthony Moisant 结合他在美国海军核潜艇部队的服役背景，阐述了严格的治理与安全护栏如何积极加速技术迭代步伐而非阻碍创新。他利用平台数据打破了“软件工程正在过时”的恐慌，概述了 Indeed 如何通过构建“职业通货”（career currency）来缓解内部员工焦虑，并强调需要扎实的语义上下文来防止智能体泛滥。Moisant 还详细介绍了 Indeed 如何打破失效的 ATS 关键词筛选死循环，透露平台目前 70% 的招聘均源自基于行为的主动式 AI 推荐。 Anthony Moisant 借鉴美国海军核潜艇的严苛纪律，提出 Indeed 的“安全速度”哲学，强调严格的安全护栏与治理不仅不会拖慢创新，反而能构建团队信心、加速技术规模化落地。 (0:04) 针对软件工程消亡的悲观论调，Moisant 指出基础 LLM 近 60% 的推理负载均集中在编码任务，且在美国整体职位发布下降 7% 的背景下，软件开发岗位招聘逆势上升了 15%。 (4:08) 为缓解员工对 AI 取代岗位的恐惧，Indeed 致力于培训全员掌握前沿的 AI 编排能力以累积“职业筹码”，并借助黑客松与创新展示打破了原本多达 9 层的管理层级壁垒。 (10:26) Indeed 正在从传统关键词简历筛选转向综合行为与便携技能的智能挖掘，目前平台上约 70% 的录用直接来自算法的主动推荐，而非求职者的手动搜索投递。 (17:44)

Indeed Chief Information Officer and Chief Security Officer Anthony Moisant argues that enterprise AI deployment requires a 'safe speed' philosophy, where rigorous governance, guardrails, and data context accelerate innovation rather than restrict it. Drawing on nuclear submarine discipline and Indeed's labor platform data, Moisant contends that AI will elevate human roles and reshape job matching rather than simply displace the workforce. Submarine Discipline and the 'Safe Speed' Framework (0:04) Drawing from his training at the U.S. Navy's nuclear submarine school, Moisant rejects the premise that safety, compliance, and governance act as bureaucratic brakes on innovation. In submarine operations, strict safety protocols establish the operational confidence required for crews to move rapidly in life-or-death environments. At Indeed, establishing transparent guardrails, identity verification, and content moderation provides internal developers and external regulators with the confidence needed to scale algorithmic experimentation safely. Labor Market Signals and Developer Job Growth (4:08) Addressing widespread speculation that autonomous agents will render human coders obsolete, Moisant cites data from Indeed’s network of 650 million job seekers and 3.5 million employers. Following major AI milestones like the release of Claude Code, U.S. software development job postings rose by 15%, even while aggregate national job openings fell by 7%. Furthermore, Moisant notes that roughly 60% of foundational large language model inference workloads are dedicated to software engineering. Rather than eliminating jobs, Moisant predicts AI will follow the trajectory of electricity, evolving into an invisible, ubiquitous utility within a decade. By automating mechanical coding tasks, generative tools push software developers up the value chain toward architecture, critical judgment, and strategic product design. Preventative Maintenance, Contextual Guardrails, and Organizational Buy-in (8:38) As AI tools empower non-technical personnel across customer support, sales, and legal divisions to construct custom agents, Moisant warns of operational risks. Without strict semantic layers and data pipelines providing grounded business context, decentralized agents are prone to hallucinations and flawed outputs during critical operations like contract analysis. To counter internal employee anxiety and potential rejection of AI workflows, Indeed's leadership emphasizes 'career currency.' Rather than guaranteeing that job descriptions will remain unchanged, the organization commits to training workers on advanced AI stacks, using internal hackathons, innovation grants, and executive showcases to incentivize adoption. This initiative also reduced organizational latency, collapsing the nine managerial layers that historically separated frontline engineers from senior executives across the broader 70,000-person parent company. Overcoming the Hiring Paradox via Recommendation Engines (15:02) Moisant criticizes the prevailing hiring paradox, where job applicants use generative tools to mass-submit resumes while employer applicant tracking systems use automated keyword filters to instantly discard them. Moisant argues that early-career candidates should focus on building demonstrable project portfolios rather than relying on theoretical machine learning coursework, as hiring managers increasingly prioritize proven initiative and business judgment over raw academic pedigree. To bypass rigid keyword filters that ignore portable and behavioral skills, Indeed has pivoted from passive keyword searches toward behavioral AI recommendation models. Moisant reports that approximately 70% of hires on Indeed now originate through algorithmic matching rather than conventional search-and-apply actions, surfacing candidates based on broader capability patterns rather than narrow job descriptions. While Indeed reports that recommendation engines now drive the majority of its platform's hires, the broader enterprise market continues to struggle with quantifying clear return on investment from early-stage agentic workflows, leaving the long-term balance between automated hiring tools and recruiter evaluation an ongoing industry challenge.

Indeed首席信息官兼首席安全官Anthony Moisant指出，借鉴核潜艇的严苛纪律建立安全护栏不仅不会减缓创新，反而是规模化AI应用建立信任与实现“安全速度”的前提；同时平台数据显示，AI未曾终结软件开发行业，且正将全球招聘从机械的简历筛选重塑为基于行为的主动推荐模式。 以核潜艇安全纪律保障创新速度 (0:04) Anthony Moisant结合其在美国海军核潜艇学校受训的背景，提出了AI领域的“安全速度”（Safe Speed）哲学。他明确反驳了将安全性、合规和治理视为业务阻碍的常规看法，强调在核潜艇这种容错率为零的高危环境中，严密的规范和防护协议恰恰是全员高速协作的信心来源。在就业市场中，算法推荐直接关系到个体的谋生发展，稍有偏差便可能波及个人与家庭，因此必须将透明护栏和问责机制置于底层架构之中，以此消除工程、法务及监管层面的顾虑，反向驱动创新落地。 身兼CIO与CSO双重职责的Moisant表示，CIO侧重于推进技术变革与业务实验，而CSO则负责统筹由网络安全、内容审核和身份验证构建的“信任层”，确保求职者与雇主真实可信。他认为两项职责相辅相成：平台的竞争力核心在于速度，而缺乏信任层的快速试错会损害产品完整性。将技术研发与安全治理纳入统一愿景，才能确保高频的产品实验不透支双边市场的信任资本。 市场数据反驳软件工程消亡论 (4:08) Moisant预测，十年之内独立的AI企业峰会将不复存在，AI将如同电力一样融入各类业务底层，成为隐形的公共基础设施。针对AI智能体将导致大量白领失业的悲观论调，Moisant引用Indeed覆盖6.5亿求职者与350万雇主的实际监测数据进行了反驳：在Claude Code等关键工程模型发布后，尽管全美整体招聘职位下滑了7%，但软件开发岗位的发布量反而逆势上升了15%。 Moisant指出，当前基础大语言模型近60%的推理工作负载都集中在软件工程领域。AI的介入并非消灭开发者，而是通过自动化枯燥重复的底层代码编写，剥离琐碎杂务，促使软件工程师向上迁移至系统架构设计、业务主观判断及产品创新等更高价值的维度。正如互联网使信息触手可及，AI带来的“智能富足”正在推动劳动价值向高阶认知演进。 防范代理泛滥与化解内部抵触 (8:38) 针对企业内客服、销售及法务等非技术人员利用AI自主构建智能体的现象，Moisant强调，虽然权限下放激发了创新，但若缺乏语义基础和业务语境，智能体野蛮生长将带来严重的运行隐患。组织必须像核潜艇执行严格的日常预防性维护清单一样，重金投入高质量数据管道、语义层及语境护栏，以防智能体在合同审查等关键任务中产生幻觉并给出违规建议。 面对AI推行初期员工对岗位安全感产生的“组织排斥”（organ rejection），Indeed管理层转而通过打造“职业货币”（career currency）建立信任。管理层坦承无法保证岗位职责一成不变，但承诺提供前沿AI培训，使员工简历具备行业认可的高阶AI协作信誉。此外，企业通过黑客松、现金奖励和高管直接对接等激励措施，瓦解了母公司7万名员工及Indeed内部1万名员工原本跨越9层的层级沟通隔阂，大幅压缩了战略信息向一线的传导延迟。 瓦解招聘黑箱与转向行为推荐 (15:02) 在早期人才选拔上，Moisant打破了“AI原生代毕业生即插即用”的设想，指出刚毕业的学生虽掌握机器学习理论，但普遍欠缺商业敏锐度、实操判断力及企业变现常识，因此Indeed更看重求职者利用AI工具完成的实际作品集与探索欲。面对求职者用AI批量海投简历、雇主用算法自动筛除而引发的“招聘黑箱”与虚假岗位乱象，平台正着力绕开僵化的关键词匹配，直接对接求职者真实能力与雇主经核实的需求。 Moisant以同事Zach为例：Zach的传统简历局限于日常本职，但Indeed系统通过其行为信号与跨企业成功模型，主动识别出其未在简历中展现的主持技能，为其精准推送了主持人职位并最终被录用。Moisant透露，目前Indeed平台上约70%的成功招聘均由AI的主动推荐促成，而非传统的关键词手动检索。这一机制促使招聘体系从静态纸面标签转向了动态的实际能力发现。 Anthony Moisant展示了将高可靠性安全架构与大规模AI应用融合的实践路径，并指出AI正促使招聘重心从死板简历转向全方位的行为能力推荐。然而，如何彻底消除求职端AI刷量与企业端算法自动拒信之间的恶性循环，仍有赖于全行业对透明匹配标准的大规模采纳。

## Background 该内容由频道 Ai4 于 2026 年 8 月 22 日发布在 YouTube 平台。视频记录了 Indeed 首席信息官（CIO）兼首席安全官（CSO）Anthony Moisant 与《财富》（Fortune）杂志编辑总监 Kristin Stoller 的深度对谈，围绕 Indeed 如何在保障人类信任与数据安全的前提下推行“安全速度”哲学，以智能体 AI（Agentic AI）驱动全球劳动力转型并赋能业务创新展开探讨。 ## TL;DR 访谈中，Anthony Moisant 结合核潜艇服役背景阐述了“安全速度”核心哲学，指出严格的治理护栏与信任基础是团队敢于全速创新的底气。他援引平台海量真实数据反驳了“AI 导致程序员失业”的论调，并警示在落地智能体时需警惕“智能体泛滥”，必须夯实业务上下文与语义层地基。为化解员工对 AI 替代的“器官排斥”，Indeed 承诺为员工注入“职业硬通货”，并通过创新机制压缩 9 层科层汇报延迟；目前平台 70% 的成功录用已由算法推荐驱动，标志着招聘范式正从被动搜索全面转向主动智能匹配。 ## Atmosphere & style 整场对话在轻松风趣与硬核严谨之间取得了绝佳平衡，开场关于“母亲秘制青酱”的幽默调侃迅速拉近了距离，随后无缝切入核潜艇生存法则与底层工程哲学。主持人提问敏锐直接，直面技术失业焦虑与招聘黑盒等现实痛点，Moisant 则以极客特有的坦诚与热忱从容化解，生动诠释了“安全也可以很性感”的工程师浪漫。现场交流节奏明快、观点交锋真实落地，既有高管层面的战略高度，又充满对一线员工与求职者的人文关怀。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=tEmbGUuKidY&t=4s) 核潜艇服役经历塑造了“安全速度”（Safe Speed）哲学：严密的安全护栏与治理框架并非官僚阻碍，而是构建系统信任与放手极速创新的基石。 - [02:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=122s) 在招聘等关乎人生命运的领域，提前将透明度与安全护栏内置于架构中，能够化解监管顾虑并让团队获得更长远的创新自由。 - [02:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=122s) CIO（聚焦业务创新杠杆）与 CSO（主导信任层与平台真实性）的双重角色并不对立，二者统一于在安全信任底座上提升组织试错与交付的“速度”。 - [04:08](https://www.youtube.com/watch?v=tEmbGUuKidY&t=248s) AI 未来将如同电力一样融入时代基础设施：十年后除了极少数底层模型开发者外，将不再存在专门讨论如何应用 AI 的泛行业大会。 - [04:08](https://www.youtube.com/watch?v=tEmbGUuKidY&t=248s) 大模型自动化低级苦工促使人才向价值链上游迁移，自 Claude Code 发布后全美软件开发职位逆势增长 15%，反驳了 AI 导致程序员失业的悲观论调。 - [08:38](https://www.youtube.com/watch?v=tEmbGUuKidY&t=518s) 警惕企业内部“智能体泛滥”（Agent Proliferation）风险，必须为分散的 Agent 构筑统一的业务上下文、语义层与数据管道，确保生成结果真实可靠。 - [10:23](https://www.youtube.com/watch?v=tEmbGUuKidY&t=623s) 化解员工对 AI 的“器官排斥”恐慌：拒绝虚假承诺岗位不变，转而投资员工掌握顶尖 AI 工具栈，赋予其具有行业溢价的“职业硬通货”。 - [10:23](https://www.youtube.com/watch?v=tEmbGUuKidY&t=623s) 通过 AI 创新机制与跨层黑客松直连决策层与一线问题解决者，有效压缩大型企业内多达 9 层的科层汇报与沟通延迟。 - [15:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=902s) 企业在选拔早期人才时应看重创造力、求知欲以及亲自动手构建并兑现商业价值的 Builder Mindset，而非迷信空谈理论的“AI 原生毕业生”。 - [17:44](https://www.youtube.com/watch?v=tEmbGUuKidY&t=1064s) AI 深度匹配引擎正在打破“求职者 AI 海投与雇主 AI 刷人”的黑盒悖论，通过多维行为洞察主动发掘简历之外的可迁移潜能与软实力。 - [18:30](https://www.youtube.com/watch?v=tEmbGUuKidY&t=1110s) Indeed 上 70% 的成功录用已直接由算法推荐驱动，标志着招聘范式已从过去的“主动搜索-投递”彻底转向主动“算法匹配”时代。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=tEmbGUuKidY&t=4s) 潜艇学校的生存法则：“安全速度”与信任基石 访谈伊始，主持人 Kristin Stoller（《财富》杂志编辑总监）打趣地提到，要接在刚刚演讲的 Zach 之后登场确实极具挑战。她与 Indeed 首席信息官（CIO）兼首席安全官（CSO）Anthony Moisant 互相调侃起来，Moisant 幽默地感叹“看来我们当中只有一个人有着真正爱他的母亲”，引得主持人笑称 Moisant 的母亲没能抽出时间赶来现场。Moisant 借机揭秘了当天最大的秘密——“我母亲做的罗勒青酱（pesto）全天下第一”，并嘱咐全场千万保密，这是家族独家私藏。随后，话题自然转入 Moisant 的硬核履历：他曾毕业于美国海军核潜艇学校。在核潜艇这种极高风险、任何微小疏忽都可能酿成灾难的环境中服役的经历，深刻塑造了他在 Indeed 推行 AI 战略时的核心理念——“安全速度”（Safe Speed）。 Moisant 阐述了潜艇生涯带给他在商业技术层面的关键顿悟：在传统商业认知中，人们常常把安全、信任与合规误认为是官僚主义，认为它们阻碍了创新与前进的速度。然而，潜艇上的严苛训练让他早早明白，恰恰是安全与严密的规则构建了绝对的“信心”，而正是“信心”才让人敢于全速前行并实现规模化扩张。Indeed 将这套工程哲学完全移植到了 AI 开发中：护栏（guardrails）、治理框架与问责机制绝非行政羁绊，而是建立信任的坚实地基。唯有建立起信任，工程师团队才能获得放手狂奔的底气，更快地迭代产品，最终帮助求职者更高效地找到工作。 ```mermaid flowchart LR accTitle: 安全速度哲学循环 subgraph 潜艇安全哲学在AI工程中的映射 A["设立严密护栏与治理<br/>(Guardrails & Governance)"] --> B["奠定系统与业务信任<br/>(Foundation of Trust)"] B --> C["建立团队前行信心<br/>(Builds Confidence)"] C --> D["实现规模化极速创新<br/>(Safe Speed at Scale)"] D --> A end ``` --- ### [02:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=122s) 双重头衔的协同：在创新提速与求职者生命线之间护航 针对企业内部对这套“安全哲学”的反馈，Moisant 指出，工程师本身就极度渴望明确边界与护栏所在；公司的法务团队同样需要透彻理解安全边界，而外部监管机构更是在高度关注决策透明度。在招聘就业领域引入 AI 具有非同寻常的伦理分量：如果 AI 系统做对了，就是为求职者推开改变命运的大门；如果算法产生偏差或做错了，则是对某个人甚至其背后的家庭和社会关上一扇关键的机遇之门，后果可能是灾难性的。因此，提前将透明度与护栏内置于架构中，不仅能够打消监管机构的疑虑，反而让企业从后续繁冗的合规审查中彻底解放出来，获得了前所未有的创新自由。 在谈及自己如何同时胜任 CIO 与 CSO 这两项看似存在天然冲突的职位时，Moisant 剖析了双重角色的互补协同： - **CIO 职责**：牢牢聚焦于通过技术变革驱动业务增长与业务创新，寻求商业杠杆。 - **CSO 职责**：主导“信任层”（Trust Layer）的构筑，涵盖网络安全、物理安全、信任与安全（Trust & Safety）、内容与行为审核，以及在平台上验证求职者和雇主双向真实身份的整套机制。 Moisant 强调，这两个岗位绝不是对立的拉扯，它们统一于公司的终极核心竞争力——“速度”。业务需要高频试错与快速学习，而这一切必须建立在求职者与雇主所依赖的高质量、高信任平台之上，双帽兼戴恰好达成了安全与创新的无缝共振。 --- ### [04:08](https://www.youtube.com/watch?v=tEmbGUuKidY&t=248s) 破解智能体泡沫：从十年前瞻到反直觉的程序员就业数据 展望智能体 AI（Agentic AI）在工作场所的演进轨迹时，Moisant 首先向现场观众抛出了一个引人深思的问题：“十年之后，在座有谁认为还会存在专门的‘AI 研讨大会’？”他以历史上的“电力”做类比：在电气化革命初期，人们确实会举办电力研讨会，但如今没有任何人会把电力当作独立的议题来讨论，它早已化作支撑一切运转的基础设施与时代织构。十年之后，除了极少数钻研底层大模型架构的专业从业者之外，AI 将像空气和水一样嵌入所有业务领域，不再会有专门讨论“如何应用 AI”的泛行业大会。回到当下，Moisant 认为智能体创造真实商业价值仍处于极早期，尽管多数企业尚未看到显著的投资回报率（ROI），但持续重度投入多年的 Indeed 已经捕捉到了惊人的业务红利，不过全行业整体依然处在爬坡的学习曲线上。 面对业界关于“未来公司将不再需要人类员工、岗位全由智能体取代”的极端论调，Moisant 拿出 Indeed 平台上涵盖的 6.5 亿求职者与 350 万雇主的庞大真实市场数据予以明确反驳： | 指标维度 | 市场普遍悲观预期 | 实际平台监测数据与行业现实 | |:--- |:--- |:--- | | **软件开发岗位趋势** | 认为 AI 将导致程序员彻底失业 | 自 2 月份 Claude Code 发布以来，全美软件开发职位逆势**增长 15%**，而同期全美整体职位发布量**下滑 7%** | | **大语言模型推理负载** | 广泛分散于各行各业各类任务 | 基础模型的推理使用场景中，近 **60%** 的计算负荷集中在软件开发与工程领域 | | **劳动力演进本质** | 自动化任务 = 岗位永久消失 | 自动化消除低级苦工（Toil），促使人才向价值链上游迁移，释放判断力与创造力，催生更多新岗位与新商业形态 | Moisant 总结道，正如互联网普及使得信息与连接变得无限充裕一样，AI 正在使“智能”变得充裕。每次工业革命都是在充裕性中创造出更多增量而非单纯剥夺机会，软件工程师的激增正是因为 AI 剥离了阻碍创造力的繁重杂务，让人类能够去做曾经想都不敢想的复杂创新。 --- ### [08:38](https://www.youtube.com/watch?v=tEmbGUuKidY&t=518s) “安全也可以很性感”：智能体泛滥与语义层地基 当被问及当前在 Indeed 最让他兴奋的技术创新时，Moisant 笑称自己是个彻头彻尾的“极客”和“书呆子”（geek & nerd），最让他热血沸腾的反而是多数人觉得乏味无趣的底层工程。他再次借用核潜艇的例子：潜艇运行的核心是日常预防性维护，每天必须像钟表一样精准执行详尽的检查清单、重复性任务与护栏搭建，这虽然看起来一点都不酷，但却是保命和运转的前提。 当前 AI 落地的一大严峻隐患是“智能体泛滥”（Agent Proliferation）。随着大模型工具的普及，不仅是工程师，客服代表、销售人员、高管团队甚至法务人员都在构建专属智能体，且表现惊艳。但致命风险在于：如果这些分散构建的智能体缺乏统一的上下文地基与严密护栏，在法务合同审查等高精场景中，系统根本无法保证生成的回答是真实扎根于业务事实的。 - 必须为企业级智能体筑牢坚固的地基：引入严格的业务上下文（Business Context）、统一的语义层（Semantic Layers）以及高标准的数据管道。 - 主持人调侃道，这正证明了“安全其实也可以很性感（Safe can be sexy）”，Moisant 则欣然笑纳，戏称这可以作为团队新的宣传标语车贴。 --- ### [10:23](https://www.youtube.com/watch?v=tEmbGUuKidY&t=623s) 击碎“器官排斥”：赋予员工职业硬通货与压缩九层汇报延迟 针对业界常年挂在嘴边的“AI 将人类解放出来从事更有意义的工作”这一说法，Kristin 犀利质疑到底由谁来定义何为“有意义”。Moisant 坦率承认这是极度棘手的现实难题。Indeed 推进 AI 初期曾遭遇过严重的“器官排斥（Organ Rejection）”：将 AI 植入组织肌体时，员工本能地产生剧烈的危机感与生理性排斥，陷入“我的岗位还在不在？我是否还有价值？”的恐慌。团队很快意识到这是自身早期逻辑与沟通上的重大疏漏。 为了消除恐慌、重塑信任，Moisant 提出了“职业硬通货（Career Currency）”的概念： - **拒绝虚假承诺**：公司绝不自欺欺人地承诺某个岗位永远不发生颠覆性变化甚至消失，因为技术浪潮不可阻挡。 - **确立赋能承诺**：公司承诺对每一位员工进行全方位投资，确保他们熟练掌握最顶尖的 AI 工具栈，成长为无所不能的“超人（superhuman）”。 - **打造履历溢价**：塑造让业界公认的“Indeed 培养印记（Indeed pedigree）”，确保走出 Indeed 的员工无论走到哪里都是炙手可热的复合型顶尖人才。 在如何真诚激发员工应用 AI 意愿的讨论中，Moisant 强调，相比于单纯的现金刺激，更重要的是领导层“及时捕捉员工的闪光点（catch people doing good）”，给予坚定的资源支持和高层曝光度。Indeed 设立了涵盖黑客松的年度 AI 创新大奖，获奖者不仅能拿到奖金和培训积分，更能获得直面核心高管的宝贵机会。Moisant 坦言，母公司 Recruit 拥有 70,000 名员工，Indeed 自身也有约 10,000 人，日常管理中高管愿景传达到一线键盘工程师需要穿透整整 9 层管理架构；反过来，一线发现影响求职者体验的关键痛点，向上汇报同样受阻于这 9 层壁垒。密集的 AI 创新机制与跨层交流，正在奇迹般地压缩这 9 层科层体系带来的组织沟通延迟。 ```mermaid flowchart TD accTitle: 组织科层延迟压缩 subgraph 传统科层结构的9层壁垒 E1["C-Suite 决策高层"] -->|战略愿景逐层衰减| L1["中间管理层 (共 9 层)"] L1 -->|执行阻滞| F1["一线键盘操作员工"] F1 -->|业务痛点反馈受阻| L1 L1 -->|信息延迟失真| E1 end subgraph AI创新机制下的短路连接 E2["C-Suite 决策高层"] <-->|"AI黑客松 / 创新大奖 / 智能体直连<br/>(极速压缩延迟)"| F2["一线实际问题解决者"] end ``` --- ### [15:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=902s) 走出象牙塔的构建者，与打碎招聘“黑盒悖论” 谈及处于职场起步阶段的应届毕业生的困局，Moisant 击碎了当下企业盲目追捧所谓“AI 原生毕业生”的幻象。他指出，现实中根本不存在一毕业就能精通 AI 系统编排又兼具商业智慧的神童；高校毕业生往往空有书本与学术理论，却缺乏阅历、缺乏看穿业务拐点的商业变现直觉。面对这一断层，Indeed 在选拔早期人才时极其看重三项核心特质： - **创造力（Creativity）**：跳出既定框架构想解法的灵性。 - **求知与成长驱动力（Drive to learn and grow）**：对新技术演进保持持久的饥饿感。 - **构建者本色（Builder mindset）**：无论是学软件工程、项目管理还是市场营销，最重要的不是空谈理论，而是“在大学期间真正用 AI 亲手构建过什么项目，并拿出作品集证明其能兑现出商业价值”。 在对话尾声，主持人抛出了困扰招聘行业的“AI 招聘黑盒悖论（AI Hiring Blackbox Paradox）”：求职者利用 AI 批量定制并疯狂海投简历，雇主端招聘人员同样用 AI 算法大批量自动化刷人，两端全被机器接管，真实的人类在其中完全隐身，导致了“投递上千家无音讯”的“幽灵职位（Ghost Jobs）”灾难。 Moisant 呼应了此前 Zach 的见解，坦言人类招聘官在大规模场景下识别精准人才与机遇的表现并不理想。Indeed 正在全力打造其专属 AI 平台，旨在通过更深入的智能匹配，将机械性的大规模人才初筛从人类手中剥离，以期彻底打破这种算法空转互搏的双向消耗，重构真实高效的招聘闭环。 ### [17:44](https://www.youtube.com/watch?v=tEmbGUuKidY&t=1064s) 从简历筛选到行为洞察：Zach 的意外转型与 AI 推荐引擎 传统招聘流程往往存在严重局限，招聘人员往往过度依赖硬性技能的机械筛选，而忽视了候选人身上那些更具通用性、可迁移性的底层行为特征与软实力。Indeed 首席信息官兼首席安全官 Anthony Moisant 分享了一个生动的真实案例：一位名叫 Zach 的求职者，原本有一份自己专注的主要职业方向，其简历内容也是完全围绕该领域编写的。然而，Indeed 现在的 AI 平台所挖掘的维度远比一纸死板的简历要深得多。系统不仅会提取简历上的显性文本，还会持续分析用户在平台上的实时互动行为，并比对不同雇主之间的用人成效模式。 事实证明，Zach 实际上拥有极其出色的活动主持（MC, Master of Ceremonies）才能。尽管他根本没有在平台上主动搜索过任何主持类职位，但 Indeed 强大的 AI 推荐引擎敏锐捕捉到了这种潜质，并主动将一份主持工作推送到了他的面前。Zach 顺势提交了申请，最终顺利拿下了这份工作。这个例子深刻体现了 AI 如何打破传统认知的壁垒，帮助招聘者跳出“以为自己知道候选人适合什么”的既有成见，发掘出隐藏在履历之外的人才价值。 ```mermaid flowchart LR accTitle: Indeed AI 深度匹配机制 subgraph S1["传统模式"] A["求职者简历"] --> B["硬性技能机械筛选"] B --> C["局限在特定职位"] end subgraph S2["Indeed AI 引擎"] D["求职者档案与作品集"] --> E["全站行为分析与跨雇主匹配"] E --> F["主动智能推荐<br/>(发掘未主动搜索的潜能)"] F --> G["成功录用"] end ``` ### [18:30](https://www.youtube.com/watch?v=tEmbGUuKidY&t=1110s) 70% 招聘来自算法推荐：告别“主动搜索”，拥抱精准人岗匹配 这种技术演进彻底颠覆了求职者的求职习惯和雇主的招聘范式。过去数十年来，在 Indeed 找工作的标准流程始终是“搜索职位，然后申请职位”。虽然目前仍有许多人保持着这一惯性习惯，但数据表明底层逻辑已经发生质的跃迁——如今 Indeed 上高达 70% 的成功录用都是直接由算法推荐驱动的，而非来自求职者的手动搜索。 面对这一根本性转变，Anthony Moisant 对广大求职者提出了关键建议：充分且如实地填写个人档案，保证档案信息的准确性，并尽可能展示完整、详实的作品集（Portfolio）。算法只有掌握了丰富而真实的上下文数据，才能输出高精度的智能匹配，进而从根本上化解求职者与招聘方之间的信息不对称问题。在分享结束后，主持人 Kristin Stoller 对这一未来图景表达了由衷期待，并与 Anthony Moisant 一道向全场观众致谢，圆满结束了本次对谈。 - **70% 招聘转化率来自推荐**：Indeed 当前的成功录用中，有 70% 源自智能推荐引擎的主动分发，标志着招聘从被动“搜索-投递”时代步入主动“算法匹配”时代。 - **档案与作品集完整度至关重要**：求职者最大化利用 AI 推荐机制的核心在于完善个人档案的真实性与深度，并开放作品集供模型理解分析。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=tEmbGUuKidY&t=4s) 看来我们当中只有一个人有着真正爱他的母亲。 — Anthony Moisant > [00:04](https://www.youtube.com/watch?v=tEmbGUuKidY&t=4s) 我母亲做的罗勒青酱（pesto）全天下第一，千万保密，这是家族独家私藏。 — Anthony Moisant > [00:04](https://www.youtube.com/watch?v=tEmbGUuKidY&t=4s) 护栏、治理框架与问责机制绝非行政羁绊，而是建立信任的坚实地基。唯有建立起信任，工程师团队才能获得放手狂奔的底气。 — Anthony Moisant > [04:08](https://www.youtube.com/watch?v=tEmbGUuKidY&t=248s) 十年之后，在座有谁认为还会存在专门的‘AI 研讨大会’？ — Anthony Moisant > [08:38](https://www.youtube.com/watch?v=tEmbGUuKidY&t=518s) 安全其实也可以很性感（Safe can be sexy）。 — Kristin Stoller > [10:23](https://www.youtube.com/watch?v=tEmbGUuKidY&t=623s) 我的岗位还在不在？我是否还有价值？ — Anthony Moisant > [10:23](https://www.youtube.com/watch?v=tEmbGUuKidY&t=623s) 更重要的是领导层及时捕捉员工的闪光点（catch people doing good），给予坚定的资源支持和高层曝光度。 — Anthony Moisant > [15:02](https://www.youtube.com/watch?v=tEmbGUuKidY&t=902s) 在大学期间真正用 AI 亲手构建过什么项目，并拿出作品集证明其能兑现出商业价值。 — Anthony Moisant > [18:30](https://www.youtube.com/watch?v=tEmbGUuKidY&t=1110s) 如今 Indeed 上高达 70% 的成功录用都是直接由算法推荐驱动的，而非来自求职者的手动搜索。 — Anthony Moisant ## Mentioned Anthony Moisant, Kristin Stoller, Zach, Indeed, Fortune, Ai4, Recruit, 美国海军核潜艇学校, Claude Code, Agentic AI ## Video description Anthony Moisant - CIO, CISO, Indeed Kristin Stoller - Editorial Director, Fortune How does the world’s #1 job site in the world move at the speed of AI without breaking the foundation of human trust? As both CIO and CSO at Indeed, Anthony Moisant operates at the critical intersection of speed and safety. He is the architect of a "safe speed" philosophy, driving radical innovation through Agentic AI while safeguarding the human trust of the hundreds of millions of job seekers and employers that depend on the platform. Join Indeed CIO and CSO Anthony Moisant in conversation with Kristin Stoller, editorial director at Fortune to discuss how Indeed is transforming a global workforce into an AI-first, agentic organization while balancing the relentless drive for innovation with the absolute necessity of security and data integrity. This isn't just about new tools, it's about a fundamental shift in how employees interact with AI to automate the mundane and unlock a new tier of human creativity and strategy. [Source](https://www.youtube.com/watch?v=tEmbGUuKidY)

[0:04] Anthony Moisant: 看来我们当中只有一个人有着真正爱他的母亲。 [0:04] Anthony Moisant: 我母亲做的罗勒青酱（pesto）全天下第一，千万保密，这是家族独家私藏。 [0:04] Anthony Moisant: 护栏、治理框架与问责机制绝非行政羁绊，而是建立信任的坚实地基。唯有建立起信任，工程师团队才能获得放手狂奔的底气。 [4:08] Anthony Moisant: 十年之后，在座有谁认为还会存在专门的‘AI 研讨大会’？ [8:38] Kristin Stoller: 安全其实也可以很性感（Safe can be sexy）。 [10:23] Anthony Moisant: 我的岗位还在不在？我是否还有价值？ [10:23] Anthony Moisant: 更重要的是领导层及时捕捉员工的闪光点（catch people doing good），给予坚定的资源支持和高层曝光度。 [15:02] Anthony Moisant: 在大学期间真正用 AI 亲手构建过什么项目，并拿出作品集证明其能兑现出商业价值。 [18:30] Anthony Moisant: 如今 Indeed 上高达 70% 的成功录用都是直接由算法推荐驱动的，而非来自求职者的手动搜索。

[0:04] Kristin Stoller: Operating inside a nuclear submarine is a high-stakes environment where a single mistake can lead to fatal tragedy. [0:04] Anthony Moisant: Building transparent guardrails from the outset reassures internal engineers, legal teams, and external regulators, which paradoxically accelerates innovation rather than stifling it. [2:57] Anthony Moisant: Speed requires rapid experimentation and quick takeaways from failed tests, but experimentation without a secure trust layer compromises product integrity. [4:08] Anthony Moisant: Just as the internet democratized connectivity and made information abundant, AI makes intelligence abundant, unlocking higher-tier judgment and new business capabilities. [4:08] Anthony Moisant: Despite doom-laden predictions that software development is obsolete, U.S. software development job postings rose by 15% following major AI engineering milestones like the Claude Code release, even while aggregate U.S. job postings fell by 7%. [4:08] Anthony Moisant: Automating repetitive coding tasks strips away mundane toil, allowing developers to move up the value chain toward architecture, judgment, and product creativity. [8:38] Anthony Moisant: safe can be sexy [10:26] Anthony Moisant: While leadership cannot promise that individual job descriptions will remain static, they commit to training every employee to become 'superhuman' using cutting-edge AI stacks. [10:26] Anthony Moisant: In a parent organization of 70,000 employees (and 10,000 at Indeed), strategic directives historically took nine managerial layers to reach hands-on keyboard staff. AI workflows and direct innovation showcases collapse this communication latency [15:02] Anthony Moisant: Instead of looking for purely theoretical skills, Indeed seeks candidates with intrinsic drive, curiosity, and a tangible portfolio of things they have actively built using AI tools. [17:44] Anthony Moisant: based on what they think they know. [17:44] Anthony Moisant: approximately 70% of hires on Indeed now originate directly through AI recommendations rather than manual search queries.

[0:04] Thank you everyone. Uh it's going to be really  hard to beat Zach. I don't know how we're even [0:09] going to begin to do that, but we're gonna we're  gonna try, right, Anthony? Incredibly difficult. [0:14] Uh one of us has a mother that actually loves  them, apparently. I know. Yeah, Anthony. Your [0:19] moment, couldn't couldn't make it. Couldn't take  the time. Well, now the cat's out of the bag on [0:23] the biggest secret of the day, which is my mom  makes the best pesto. So, uh you all know that, [0:27] but please don't tell anyone. We're trying to  keep it local in the family. Um, but Anthony, [0:32] speaking of fun facts, one of my favorite  facts about you is that you are a graduate [0:38] of the Navy's nuclear submarine school,  a place where I'm sure that, you know, [0:42] mistakes can be very dangerous. It's a it's a  very risky environment. And now I want to connect [0:48] that to AI because you are running the safe speed  philosophy um about AI at Indeed. tell me about, [0:55] you know, the lessons you might have learned  from your time, you know, at the submarine [0:58] school and how you can kind of apply those  today and what you're doing. Yeah, you know, [1:02] it's fascinating. Uh, in in as a submariner, uh,  you are definitely in a put in a place where uh, [1:09] life safety is critically important. Mistakes can  can end in a lot of tragedy. And uh I think from a [1:16] business perspective sometimes we get confused  in the tension and think that safety, trust, [1:21] security is actually bureaucracy or gets in the  way of speed, gets in the way of of innovation. [1:27] And one of the things that I learned early on  in in uh on on the submarine is actually this [1:32] is what builds confidence and confidence is what  builds speed. Confidence is allows you to move at [1:38] speed and at at scale. Um so we've sort of taken  the same philosophy here at Indeed and from from [1:44] an engineering perspective of you know guard  rails governance accountability isn't actually [1:50] bureaucracy it's the foundation of trust and  that foundation of trust builds confidence which [1:56] allows us to build faster uh and create better  outcomes and get folks jobs. Yeah, I love that. [2:02] And how do your how does it land internally?  How do your employees react to that? You know, [2:06] it's interesting. You would I would think that  uh you know engineers want to know where the [2:11] guard rails are. In fact, sure. Uh our lawyers  want to understand how we're thinking about [2:16] the guard rails as well. Regulators are wanting  to understand from a transparency perspective, [2:22] how do we make decisions when when we're helping  people get jobs, when we're leveraging AI, [2:26] if we do that right, we're opening the door, like  you heard from Zach earlier, we're opening a door [2:31] for somebody. If we do that wrong, we're closing  the door on an opportunity. And that can have a [2:36] disastrous effect on an individual, on a family,  on society. So, it's really, really important that [2:41] we've got all of these guard rails in place that  we're transparent. And, you know, again, if we can [2:47] be transparent, if we can build the guardrails  up front, regulators understand. Sure. And when [2:52] regulators understand, it actually frees us up  from an innovation standpoint. Oh, excellent. Now, [2:57] speaking of innovation, you have two titles. How  do you talk to us about those and how you balance [3:02] both of those and what it means? Right. So I think  from a CIO perspective, so I am the CIO for Indeed [3:08] and uh this is really firmly focused on creating  innovations for the business, growing the business [3:13] through technology transformation and then I'm  also responsible with my chief security officer [3:18] hat for the trust layer. So everything from cyber  security to physical security, trust and safety, [3:26] moderation. How do we how do we validate  and verify humans that are on the platform, [3:30] job seekers that are on the platform? How do we  validate employers that are on the platform? Um, [3:36] and I think it's a it's uh it's it's it is  two very distinct jobs, but you can't have [3:42] one without the without the other. Very important.  And both of them are actually not opposed from the [3:48] underlying competitive advantage we have from a  business perspective which is speed. Mhm. So we [3:54] want to innovate, we want to experiment, we want  to learn from experiments fast, but we also need [3:59] to make sure that we've got a high quality product  for our job seekers and employers and that's where [4:03] trust comes in. So this is definitely a a synergy  that happens between the two two hats that I wear. [4:08] Yeah, that's nice. Now set the scene for us here.  Where is Agentic AI at in the workforce today? [4:15] What does it look like right now? And then  where do you think it's heading, let's say, [4:18] 10 years from now? Yeah, I think one of the things  I'll start 10 years out. I I'm curious if anyone [4:25] here believes that there's going to be AI for a  conference 10 years from now. Yeah, raise your [4:30] hand. Raise your hand if you think it's going  to still exist. I think what's happening is we, [4:34] you know, I think people probably had a this, you  know, electricity conference in the past. We no [4:41] longer talk about electricity as own thing. It's  just the means to an end. And I think AI is going [4:46] to be more of the foundation, more of the fabric.  It's going to be embedded in everything that we're [4:50] doing in every domain. I don't believe 10 years  into the future, we're going to treat this outside [4:55] of the practitioners angle of how do you build  new AI platforms and things of that nature. From [5:01] from a leverage of AI perspective, I don't think  we're going to have conferences like this anymore. [5:05] Uh perhaps 10 years into the future. Um what what  I'm seeing and what we're seeing from an agentic [5:11] standpoint right now is it's it's really early  days of creating real value for businesses. Y I [5:17] think uh many studies out there would suggest that  uh most companies are not creating uh a ton of ROI [5:25] yet but they are paving the pathway and I think  uh some companies are able to get ROI. Indeed [5:32] happens to be one that that has been lucky enough  to invest for several years. We are seeing pretty [5:37] amazing uh outcomes for for our business. U but  I think we're still on the learning curve. Sure. [5:42] We're still on the learning curve. Yeah. But I I  know you know there's some who predict that maybe [5:46] you know in the future we won't have employees or  a workforce or any of that. It'll all be agents. [5:52] How do you feel about that? Is it it's not what  we're seeing in the data. It's actually not what [5:56] we're seeing with our company. So put this into  context. We have about 650 million uh job seekers [6:01] on our platform. That's a lot of a lot of people  engaging with the site every single day. Three and [6:06] a half million employers. So I think we've got  we absolutely have the biggest marketplace out [6:11] there of this kind of intelligence, this kind of  signal. What's happening with jobs? And I think [6:17] much like the sort of the revolution when you  you've taken the internet democratized and made [6:24] um abundant information and connectivity. AI is  making intelligence abundant and with intelligence [6:34] we're actually opening up new ways of thinking  new things that we never thought were possible [6:38] and that's actually fueling growth actually.  So I'll give you an example a a quick stat. Uh [6:45] about a year and a half ago, two years ago, uh I'm  sure many of you heard that software development [6:50] is dead. Software model, we're never going to have  software developers anymore. It's all going to be [6:56] uh AI. M well actually funny fact uh since the  since the uh claude code release in February [7:05] software development jobs are up 15% in the  United States whereas the rest of job postings [7:12] in the United States are down 7%. You think about  what that would imply LLM so effectively when you [7:20] look at the leverage of LLM these foundational  models almost 60% of inference use cases are [7:25] software development in nature so meaning it's  like the biggest it's the biggest atomic issue [7:30] that's being used in AI right now is towards  software development software engineering so [7:35] why are we seeing growth in software developers  and I think it comes down to now we can do more [7:42] now we can move up the value chain now we can  express more creativity, more judgment. AI is [7:48] allowing us to strip away some of the toil that  was in the way of creativity. It was stripping [7:53] away the toil that was in the way of building new  businesses, creating new capabilities. So, uh what [7:58] we see in the data and what's supported in many  of our customers is as we automate away tasks, [8:04] jobs change. Skills necessary to be successful  change, but jobs are not going away. In fact, [8:12] we're creating more jobs than we're losing at this  point, and there's no reason to expect otherwise. [8:18] Uh, every industrial revolution created more and  more from the abundance that was created before, [8:25] and we're going to see that from the abundance  of information. I I I choose to believe that. [8:29] That's interesting. So, there's hope for all  you software developers out there. You've heard [8:32] it here. Absolutely. And we're and we're hiring,  by the way. Yes. Um, so so what about at Indeed, [8:38] then? You know, what what are you working on  right now that excites you? How are you using [8:41] AI in innovative ways? I think that um so I'm I'm  a I'm a geek in many in many ways. I'm a little [8:49] bit of a nerd. What excites me doesn't excite  a lot of people. Uh and I think very similar to [8:54] my learnings on a submarine is preventative  maintenance on a submarine is an absolute [9:00] necessity. When you're going after a nuclear  submarine, you've got to make sure that you're [9:04] going through your checklist every single day.  you're doing the sort of repetitive processes, [9:10] the repetitive tasks, creation of the guard rails  like clockwork and it's not the sexiest thing. And [9:17] so what excites me is actually not the sexiest  thing. It's the these are the foundations that [9:22] a lot of people take for granted. Uh I'll give  you an example of this. Um, one of the one of the [9:28] dangers of agent proliferation is as you start to  democratize the building and innovation throughout [9:34] a company and it's no longer software engineers.  It's actually your CS people, your sales reps, [9:40] your executive teams. We have lawyers that that  are building agents and they're doing a phenomenal [9:44] job. But one mistake could be the proliferation  of agents where the agents have no guard rails. [9:50] The agents have no foundation of context.  So, when you're asking an agent a question, [9:55] let's say a contract review that a lawyer built  an agent for, how do we know that agent's actually [10:00] giving a good grounded response, right? And in  order to do that, you've got to frame all of your [10:05] agentic uh foundation in business contexts,  semantic layers, sort of data pipelines that [10:12] are of high quality. So, I jump out of bed on  some of the most geeky things out there. No, [10:17] I think safety safety is definitely also sexy.  So, safe safe can be sexy. Exactly. That's a new [10:23] That's our new our new bumper sticker. Exactly.  Uh, another phrase that I hear that I think I am [10:29] really really sick of and I'm curious if if you  are as well, but everyone, you know, for the past [10:34] year has talked about, you know, AI is freeing  humans up to do so much more meaningful work. [10:38] And I kind of pause because I'm like, you know,  what is meaningful? Who gets to decide? Like, [10:43] how do you how do you think about that? And  how are you you thinking about that at Indeed? [10:50] Yeah, that that's a really tricky question and  probably one of the things that we struggled with [10:54] early on is what I would say organ rejection.  The body rejects the organ. Yeah. You know, [11:00] you bring AI into into the sort of the body of  an organization and there's just an immediate [11:05] visceral reaction of what's going to happen with  my role, what's going to h am I valuable now? Can [11:12] I be valuable? Am I going to be valuable in the  future? And uh and I don't think we got that right [11:16] originally. We did not pay enough attention to it.  Uh but we quickly identified that was a gap in our [11:23] logic. It was a gap in our thinking. So we went  after okay how do we create the most amount of you [11:30] know like I said trust creates greater confidence  and confidence creates more speed. So how do we [11:36] create more confidence in the organization? A lot  of this was we need to create a career currency [11:42] for the people working at Indeed. We need them  to know that we cannot ever promise that a job [11:48] is not going to change so measurably that that it  goes away. We could not possibly say that. What [11:54] we can say is we're going to invest in every  employee making sure that they are superhuman [11:59] when it comes to the technology stack that we  have available to us that they've got all the [12:03] best enablement around AI that if you are coming  from Indeed every company in the world will see [12:10] any person coming from Indeed and know because of  that pedigree because of that enablement that the [12:16] career currency carries forward and you'd be lucky  to hire an Indidian into one of your businesses. [12:22] Yeah, I think a lot of the fear and  resistance also comes from, you know, [12:26] incentivizing people to actually want to use AI in  the workplace. And on my panel yesterday, we were [12:32] talking about how do you get people to that point  of being excited about it? Is it knowing that [12:36] they're going to have this Indeed pedigree? Is it  a raise? Is it a bonus? Is it a promotion? How do [12:41] you what do you think is the best way to get the  workforce excited? Yeah, I think it's not at least [12:47] for Indeed in my in my experience, it's not even  you know what incentivize a a software engineer [12:52] is sometimes different than a lawyer uh someone  that's got a book of business on the sales front, [12:58] the CS front, etc. But there are a couple things  that we recognize. One is uh you have to express [13:05] what you want the organization to look like and  how you want people to behave and you need to [13:10] catch people doing good. So reward, recognition,  u I think proximity to leadership and and sort of [13:17] that constant drum beat of this is important, this  is valuable, and that I've got your support like [13:22] the the thing that I'm going to do as an executive  on my team, I'm going to make sure you've got the [13:27] time to invest, to grow, to develop. Um, we do do  bonuses. We've got innovation awards. We in fact [13:33] had a we have an annual award, which is sort of a  a bunch of hackathons leading up to an annual AI [13:40] innovation award. Oh, nice. Bringing people into  uh to a central location. Uh there's cash bonuses. [13:48] There are training uh credits that are being uh  deployed to the the people that have won. But more [13:54] importantly, there's exposure to leadership. You  know, in a company, you know, our parent company [14:00] that we've got 70,000 employees, indeed it's  about 10,000 employees. as an executive in our [14:05] sea suite, it is very rare for me to have the  opportunity to spend time with a lot of people [14:10] that are closest to the media's problems, those  those engineers, the CS reps, the sales reps. And [14:15] so what we found is exposure is is probably the  biggest opportunity there. And and I'd say another [14:24] benefit that that we get is latency between a  seauite and people that are closest to a problem [14:30] happens because we don't actually communicate  very often and it goes through lots of layers [14:36] of leadership. As an example, my vision takes  nine layers to go down to translate to the person [14:41] that's on keyboard, right? and the person who has  hands-on keyboard identifying a problem that is [14:47] really getting in the way of the best outcomes  for a job seeker, they have to go through nine [14:52] layers to actually be heard. And so we think that  some of our AI programs and a lot of things that [14:57] we're doing is actually shrinking the latency  and and uh in in that process as well. Well, [15:02] let's talk about that person on the bottom rung  of the career ladder, the early career. I'm [15:06] sure we've all gotten the whiplash of, you know,  graduates are screwed, graduates are not screwed, [15:10] early career is dead, early career is fine.  Well, how do you think about it and what are you [15:14] looking for in your your new hires and your early  career talent? Yeah, I think everyone says my view [15:21] everyone says they want this AI native AI, you  know, intimate person that is coming out of school [15:27] and just deeply knows how to wield and orchestrate  AI. Unfortunately, it doesn't exist. Even a lot of [15:35] people that are coming from school today um they  have a lot of book knowledge, academic knowledge [15:40] around AI and AI and systems, but it's not married  up with people that have wisdom of the years. It's [15:46] not married up with someone that can see around  corners and understand how a business monetized [15:52] as an example. So, um what we look for is I  think sort of the in the tangible intangibles [15:59] if you will is like do you have creativity? Do  you have drive? Are you driven to learn and to [16:05] grow? And are you a builder? And so I I'd say  that for a whether you're a software engineer, [16:12] project manager, a marketer coming out of school,  one of the best things that we can get people to [16:17] do is actually build with AI and demonstrate their  portfolio to the companies. Actually, I think more [16:24] and more companies are going to be wanting and  interested in when you show up for a job. Yeah. [16:29] Show me what you've done. Show me what you've  built. show me how what you've built can translate [16:34] into business value. And I think we're starting  to see that from a lot of the entrepreneurs that [16:39] are coming out of the CS programs right  now. And it's it's pretty exciting. Oh, [16:42] exciting. The other thing I wonder about too is um  I like to call it the AI hiring blackbox paradox [16:48] here is that you know job seekers are using AI  to make their resumes, send them out to as many [16:54] people as possible. Recruiters are using AI to  screen it out. There's no real human there. How do [17:00] you solve for that? especially on a platform like  Indeed which is is trying to solve the hiring loop [17:05] here. Yeah. The I I think there's that paradox. I  think also there's a black hole problem. There is [17:12] a I don't know if everyone is familiar with this  idea of a ghost job. Yeah. I apply and I never [17:17] hear back. I apply to a thousand places. I never  hear back. One of the ways that we solve this and [17:22] I had just a wonderful conversation with Zack.  I wish we I wish he was on our panel actually [17:26] earlier is uh he actually hit it home with what  we are doing with our AI platform. More and more [17:32] we're trying to strip away the identification  of talent opportunity from humans because they [17:38] can't do a good job actually they cannot do a good  job uh at scale and a lot of the technologies out [17:44] there are really around filtering in sort of hard  skills and they're not looking at other behaviors [17:49] they're not looking at things that I think will  be more portable. Um, so Zach had this experience. [17:54] He was look, he sort of has a a a primary job that  that he is interested in. That's how his resume is [18:00] written. Our AI platform actually goes much deeper  than the resume. And we extract not just resume, [18:07] but we look at behaviors on the site. We look at  what is working in one employer versus another [18:12] employer. And he's a fantastic MC. And our  recommendation engine, he was not searching [18:18] for an MC job. our recommendation engine suggested  the job to him. He applied for the job and he got [18:26] the job. And I think more often than not, that's  what we're going to work on is recruiters not [18:30] filtering out based on what they think they know.  In fact, 70% of what we do today. It used to be [18:36] you go to Indeed, you search for a job and then  you would apply for a job. And a lot of people [18:41] are doing that right now, but 70% of our hires  are coming from recommendations at this point. [18:46] That's actually filling out your profile,  making sure your prof your profile is accurate, [18:52] making sure that your portfolio is available  allows us to get better recommendations and that [18:56] that's going to solve the problem. Excellent.  Well, I very much look forward to it. Thank [19:00] you so much, Anthony, and thank you all for  joining us. Thank you, everyone. Thank you.