本报告所称 Agent 人才：Agent talent in this report:主要职责直接参与 Agent 系统的构建、运行、研究、评测、产品化或技术部署的人才。本文统计的 JD / 人才均与 Agent 系统强相关；下文分类名称省略 Agent 前缀以减少重复。Agent 系统能基于模型规划步骤、调用工具并完成任务。People whose primary responsibilities directly involve building, operating, researching, evaluating, productizing, or technically deploying Agent systems. The counted JDs and professionals are strongly Agent-related; the displayed group names omit the repeated Agent prefix. Agent systems use models to plan steps, call tools, and complete tasks. 十个岗位组：Ten job groups:Application Engineering、Research / Model Behavior、Runtime / Platform、Deployment Engineering、Product / Design、Post-training / Robustness、Evals / Quality、Orchestration / Workflow、Solutions Architecture、Safety / Governance。Application Engineering; Research / Model Behavior; Runtime / Platform; Deployment Engineering; Product / Design; Post-training / Robustness; Evals / Quality; Orchestration / Workflow; Solutions Architecture; and Safety / Governance.
Agent 人才是共享生产底层的多个岗位市场Agent talent is several job markets sharing a production foundation
本报告基于 259 个可见 Agent JD 与 513 名可识别在职 Agent 人才，覆盖美国 32 家公司。同一套十个岗位组贯穿招聘需求与人才存量，使公司重点可以直接对比。This report is based on 259 visible Agent JDs and 513 observed current Agent professionals across 32 U.S. companies. The same ten job groups connect hiring demand with talent stock, making company priorities directly comparable.
513
可识别在职 Agent 人才Observed current Agent professionals
公开可见下限Publicly visible lower bound
259
活跃 Agent JDActive Agent JDs
公开招聘需求Visible hiring demand
10
统一岗位组Unified job groups
259 JD · 513 人259 JDs · 513 people
53.2%
前两家公司人才占比Talent share held by the top two companies
273 / 513
01Agent 共同底层是接入、评测与稳定运行能力The shared Agent foundation is integration, evaluation, and reliable operation
Reliable operation、Continuous evaluation 与 Real-system integration 分别覆盖 87.3%、73.4% 和 69.5% 的 JD；LangChain / LangGraph 仅覆盖 9.3%。Reliable operation, Continuous evaluation, and Real-system integration appear in 87.3%, 73.4%, and 69.5% of JDs; LangChain / LangGraph appears in only 9.3%.
02Application Engineering 最大，但公司的首要缺口完全不同Application Engineering is largest, but each company's primary gap is different
Application Engineering 有 77 个 JD、覆盖 18 家公司；但 Google 的 22 个 JD 中有 11 个是 Deployment Engineering，Salesforce 的 27 个中有 8 个是 Orchestration / Workflow，OpenAI 还同时发布 6 个 Post-training / Robustness 与 5 个 Safety / Governance 岗位。Application Engineering has 77 JDs across 18 companies. Yet 11 of Google's 22 JDs are Deployment Engineering, 8 of Salesforce's 27 are Orchestration / Workflow, and OpenAI also posts 6 Post-training / Robustness and 5 Safety / Governance roles.
03最大人才库与最激进招聘方分属两批公司Talent reservoirs and the most aggressive hirers are different company sets
Microsoft 与 Salesforce 集中 273 / 513 名可识别在职人才；Scale AI 与 OpenAI 合计发布 68 个 JD，却只观察到 18 名在职人才，招聘需求与现有人才存量明显错位。Microsoft and Salesforce hold 273 / 513 observed current professionals. Scale AI and OpenAI post 68 JDs combined but have only 18 observed current professionals, revealing a clear mismatch between hiring demand and visible talent stock.
04主要人才来源不只在 AI Lab，也在云平台、咨询与企业软件The main talent sources extend beyond AI labs into cloud, consulting, and enterprise software
Amazon / AWS 是最大可见来源，共 21 人；Bain 向 Decagon 输送 6 人，Slack 与 MuleSoft 合计向 Salesforce 输送 8 人。不同 Agent 岗位对应不同来源生态。Amazon / AWS is the largest visible source with 21 people; Bain feeds 6 into Decagon, while Slack and MuleSoft together feed 8 into Salesforce. Different Agent roles draw from different source ecosystems.
02 · 公司瓶颈02 · Company bottlenecks
同样在招 Agent，各家公司要的人完全不同The same Agent label hides very different hiring needs
同样在招 Agent，OpenAI 要的是安全接入真实系统，Anthropic 要的是测清模型行为，Scale AI 要的是评测工厂，Salesforce 要的是企业工作流持续验证，ServiceNow 要的是状态层运行。它们共享 Agent 标签，但真正竞争的能力栈并不相同。The same Agent label hides different needs: OpenAI centers on safe integration into real systems, Anthropic on model-behavior evaluation, Scale AI on evaluation pipelines, Salesforce on workflow validation, and ServiceNow on state-layer operation. They share an Agent label but compete for different capability stacks.
OpenAI33 JD
安全系统接入Safe system integration
Real-system integration 覆盖 33 / 33，Safety / permissions 覆盖 32 / 33。OpenAI 的反常点在于：几乎每个 Agent JD 都要碰到真实系统边界，需求已经超出 demo 构建。Real-system integration covers 33 of 33 JDs, and Safety / permissions covers 32 of 33. OpenAI's outlier is that almost every Agent JD touches real-system boundaries, with demand moving beyond demo-building.
Real-system integration
100.0%33 / 33
Reliable operation
100.0%33 / 33
Safety / permissions
97.0%32 / 33
Continuous evaluation
57.6%19 / 33
Agent orchestration
33.3%11 / 33
数据来源: Metix AIData source: Metix AI
Anthropic16 JD
模型行为评测Model-behavior evaluation
Continuous evaluation 覆盖 12 / 16，高于 Real-system integration 的 6 / 16。它和 OpenAI 的差异在这里最清楚：Anthropic 更像评测、环境与模型行为队形。Continuous evaluation covers 12 of 16 JDs, above Real-system integration at 6 of 16. This is the clearest split from OpenAI: Anthropic looks more like an evaluation, environment, and model-behavior formation.
Real-system integration
37.5%6 / 16
Reliable operation
87.5%14 / 16
Safety / permissions
68.8%11 / 16
Continuous evaluation
75.0%12 / 16
Memory / state
12.5%2 / 16
数据来源: Metix AIData source: Metix AI
Scale AI35 JD
评测工厂Evaluation factory
Continuous evaluation 覆盖 35 / 35，是六项能力中唯一拉满的一项；同时 Agent orchestration 覆盖 19 / 35，说明它在招的是可规模化评测与反馈流水线。Continuous evaluation covers 35 of 35 JDs, the only saturated capability among the six; Agent orchestration also covers 19 of 35, pointing to scalable evaluation and feedback pipelines.
Real-system integration
80.0%28 / 35
Reliable operation
85.7%30 / 35
Safety / permissions
57.1%20 / 35
Continuous evaluation
100.0%35 / 35
Agent orchestration
54.3%19 / 35
数据来源: Metix AIData source: Metix AI
Google22 JD
云与产品编排Cloud/product orchestration
Google 的 Agent orchestration 覆盖 12 / 22，高于全市场 42.1% 的基准；但 Real-system integration 只有 12 / 22，不像 OpenAI 那样饱和。Google's Agent orchestration covers 12 of 22 JDs, above the 42.1% market baseline; Real-system integration is only 12 of 22, unlike OpenAI's saturation.
Real-system integration
54.5%12 / 22
Reliable operation
86.4%19 / 22
Safety / permissions
63.6%14 / 22
Continuous evaluation
68.2%15 / 22
Agent orchestration
54.5%12 / 22
数据来源: Metix AIData source: Metix AI
Salesforce27 JD
工作流持续验证Workflow validation
Continuous evaluation 覆盖 26 / 27，Real-system integration 覆盖 24 / 27。Salesforce 的 Agent 需求集中在企业工作流接入后的持续测量。Continuous evaluation covers 26 of 27 JDs, and Real-system integration covers 24 of 27. Salesforce concentrates demand on connecting enterprise workflows and measuring them continuously.
Real-system integration
88.9%24 / 27
Reliable operation
92.6%25 / 27
Safety / permissions
48.1%13 / 27
Continuous evaluation
96.3%26 / 27
Agent orchestration
48.1%13 / 27
Memory / state
3.7%1 / 27
数据来源: Metix AIData source: Metix AI
ServiceNow27 JD
状态层运行State-layer operation
Memory / state 覆盖 6 / 27，是全市场 6.6% 基准的 3.4 倍；ServiceNow 的差异更像长流程、工单状态与平台运行。Memory / state covers 6 of 27 JDs, 3.4 times the 6.6% market baseline. ServiceNow's difference looks closer to long-running workflows, ticket state, and platform operation.
Reliable operation 覆盖 8 / 8，Safety / permissions 覆盖 6 / 8，说明 Agent 需求紧贴数据平台的运行边界。Reliable operation covers 8 of 8 JDs, and Safety / permissions covers 6 of 8, tying Agent demand closely to the operating boundary of a data platform.
8 / 8Reliable operation6 / 8Safety / permissions
03 · 十组需求地图03 · Ten-group demand map
最大需求是做应用，最有辨识度的需求却在公司自己的瓶颈Applications drive the most volume; company-specific bottlenecks create the real differentiation
十个岗位组共同构成 Agent 招聘需求地图。Application Engineering 有 77 个 JD，Research / Model Behavior 有 50 个，Runtime / Platform 有 34 个，三组合计占 62.2%；但公司层面的招聘重心并不随大盘走。The ten job groups form the Agent demand map. Application Engineering has 77 JDs, Research / Model Behavior 50, and Runtime / Platform 34, together accounting for 62.2%; company-level priorities, however, do not simply follow the market total.
十个岗位组的公开招聘需求Visible hiring demand across ten job groups
Application Engineering 是最大需求组，但岗位最多不等于最稀缺。后文将把同一岗位组框架与 513 名可识别人才存量对齐。Application Engineering is the largest demand group, but the most postings do not necessarily imply the greatest scarcity. A later section aligns the same job-group framework with 513 observed professionals.
Application EngineeringApplication Engineering
77 JD · 18 家公司· 18 companies
Research / Model BehaviorResearch / Model Behavior
四家重点公司的十岗位组招聘构成Ten-group hiring mix across four focus companies
完整 JD 大盘呈现四种招聘配方：OpenAI 的 33 个岗位覆盖 10 个岗位组中的 8 个，Application Engineering 最多（10），同时加码 Post-training / Robustness（6）与 Safety / Governance（5）；Anthropic 的 16 个中 Research / Model Behavior 有 6 个；Scale AI 的 35 个中 Application Engineering 有 11 个、Research / Model Behavior 有 9 个；Google 的 22 个中 Deployment Engineering 独占 11 个。The complete JD base reveals four hiring recipes. OpenAI's 33 roles cover 8 of the 10 job groups, led by Application Engineering (10) while also investing in Post-training / Robustness (6) and Safety / Governance (5). Anthropic has 6 Research / Model Behavior roles among 16. Scale AI has 11 Application Engineering and 9 Research / Model Behavior roles among 35. Google assigns 11 of 22 roles to Deployment Engineering.
OpenAI
33JD
Application EngineeringApplication Engineering10
Research / Model BehaviorResearch / Model Behavior4
259 个 JD 形成十条竞争赛道259 JDs form ten competition lanes
十条竞争赛道对应十类主要交付物。公司真正争夺的是同一工作对象下的候选人；贴着 Agent 标签但工作对象不同的岗位，竞争边界也不同。能力词只说明瓶颈差异，不改变岗位竞争边界。The ten competition lanes correspond to ten primary deliverables. Companies compete for candidates within the same work object; roles carrying the Agent label but tied to different work objects have different competition boundaries. Capability terms explain bottleneck differences without changing the competition boundary.
01
Application EngineeringApplication Engineering
构建面向用户或具体业务场景的 Agent 产品、功能与应用。Builds user-facing or domain-specific Agent products, features, and applications.
负责权限、Guardrails、安全执行、风险控制与治理。Owns permissions, guardrails, secure execution, risk controls, and governance.
11个 JDJDs5家公司companies
OpenAI 5Google 2Salesforce 2Glean 1Microsoft 1
查看代表岗位View representative roles
Product Manager, Agent Security & Governance
Agentic Safety and Ecosystem Architect, Trust and Safety
Principal Product Manager, Agent 365 Security & Governance
Principal Software Engineer, Codex Cyber
05 · 共同底层05 · Shared foundation
框架不再是主角：Agent 招聘买单的是“接得上、测得准、跑得稳”Frameworks are no longer the main story: Agent hiring pays for systems that connect, evaluate, and run reliably
LangChain / LangGraph 是出现最多的具名框架，却只覆盖 24 / 259 个 JD。Reliable operation、Continuous evaluation 与 Real-system integration 分别覆盖 226、190 和 180 个 JD，是 7.5—9.4 倍的覆盖差距。LangChain / LangGraph is the most-mentioned named framework, yet it appears in only 24 / 259 JDs. Reliable operation, Continuous evaluation, and Real-system integration appear in 226, 190, and 180 JDs—a 7.5–9.4× coverage gap.
生产能力远比框架关键词更普遍Production capabilities are far more common than framework keywords
框架仍是实现路径，但已经不能代表 Agent 人才的共同要求。跨公司真正稳定出现的是运行、评测与集成能力。Frameworks remain implementation paths, but no longer represent the shared requirement for Agent talent. Operation, evaluation, and integration are the capabilities that consistently recur across companies.
跨公司的生产能力Cross-company production capabilities
这些是跨十个岗位组统计的多标签能力信号These are multi-label capability signals measured across all ten job groups
Reliable operationReliable operation
87.3% · 226
Continuous evaluationContinuous evaluation
73.4% · 190
Real-system integrationReal-system integration
69.5% · 180
具名框架Named frameworks
最高仅覆盖 24 个 JDThe highest reaches only 24 JDs
LangChain / LangGraph
9.3% · 24
CrewAI
2.7% · 7
AutoGen
1.9% · 5
LlamaIndex
1.9% · 5
数据来源: Metix AIData source: Metix AI
06 · 存量与稀缺06 · Stock and scarcity
人才最多与招聘最凶分属两批公司Talent reservoirs and the most aggressive hirers are different company sets
Microsoft 与 Salesforce 合计拥有 273 / 513 名可识别 Agent 人才，占 53.2%；Scale AI 与 OpenAI 则有 35 和 33 个活跃 JD，但可识别存量只有 8 和 10 名。在同一岗位组框架下，Research / Model Behavior 是最清晰的稀缺方向。Microsoft and Salesforce together hold 273 / 513 observed Agent professionals, or 53.2%. Scale AI and OpenAI have 35 and 33 active JDs but only 8 and 10 observed current professionals. Within the same job-group framework, Research / Model Behavior is the clearest scarcity signal.
可识别 Agent 人才公司分布Observed Agent talent by company
Microsoft 与 Salesforce 是最主要的存量池；高需求的 OpenAI 与 Scale AI 更依赖外部候选人市场。Microsoft and Salesforce are the main reservoirs; high-demand OpenAI and Scale AI depend more on the external candidate market.
Microsoft
139 · 27.1%
Salesforce
134 · 26.1%
Google
44 · 8.6%
Meta
36 · 7.0%
Decagon
33 · 6.4%
ServiceNow
31 · 6.0%
Adobe
15 · 2.9%
Hippocratic AI
14 · 2.7%
Sierra
11 · 2.1%
OpenAI
10 · 1.9%
数据来源: Metix AIData source: Metix AI
每 100 名可识别人才对应的活跃岗位Active postings per 100 observed professionals
Research / Model Behavior 为 138.9（50 个岗位 / 36 名人才），是最清晰的稀缺信号。Safety / Governance 为 157.1（11 / 7）、Post-training / Robustness 为 140.0（14 / 10），也显示出更早期的紧张迹象。Research / Model Behavior reaches 138.9 (50 postings / 36 professionals), the clearest scarcity signal. Safety / Governance reaches 157.1 (11 / 7) and Post-training / Robustness 140.0 (14 / 10), also showing earlier signs of pressure.
Research / Model BehaviorResearch / Model Behavior
138.9 · 50 / 36
Evals / QualityEvals / Quality
100.0 · 13 / 13
Application EngineeringApplication Engineering
98.7 · 77 / 78
Runtime / PlatformRuntime / Platform
94.4 · 34 / 36
Solutions ArchitectureSolutions Architecture
38.7 · 12 / 31
Deployment EngineeringDeployment Engineering
27.5 · 19 / 69
Orchestration / WorkflowOrchestration / Workflow
24.5 · 13 / 53
Product / DesignProduct / Design
8.9 · 16 / 180
数据来源: Metix AIData source: Metix AI
城市集中City concentration
San Francisco
拥有 133 名可识别人才（25.9%），但招聘需求占比更高：旧金山 91 个岗位（35.1%），湾区合计 58.3%。纽约也呈现 15.1% 岗位对 5.3% 人才的错配。It holds 133 observed professionals (25.9%), while demand is more concentrated: San Francisco has 91 postings (35.1%) and the Bay Area totals 58.3%. New York also shows 15.1% of postings versus 5.3% of talent.
岗位组存量Job-group stock
35.1%
Product / Design 是最大可识别存量（180 人），其次是 Application Engineering（78）与 Deployment Engineering（69）。公开履历对不同岗位组的可见度并不相同，因此供需比更适合判断优先级。Product / Design is the largest observed stock (180 people), followed by Application Engineering (78) and Deployment Engineering (69). Public profile visibility differs by job group, so demand-stock ratios are more useful for prioritization.
07 · 来源与流动07 · Sources and flows
抢人路线从 AI Lab 之外开始The sourcing map starts beyond AI labs
408 名人才可识别进入当前公司的外部上一站。Amazon / AWS 以 21 人成为最大来源；Bain 向 Decagon 输送 6 人，Slack 与 MuleSoft 合计向 Salesforce 输送 8 人。云平台、咨询交付与企业软件生态分别对应不同能力来源。An external prior employer is observable for 408 professionals. Amazon / AWS is the largest source with 21 people; Bain feeds 6 into Decagon, while Slack and MuleSoft together feed 8 into Salesforce. Cloud platforms, consulting delivery, and enterprise software ecosystems supply different capabilities.
进入当前 Agent 团队前的主要雇主Leading prior employers before the current Agent team
Amazon / AWS 的可见来源量是第二名 Microsoft 的 2.3 倍；Meta / Instagram、Google、LinkedIn 与 Bain 共同说明，来源池不应只看 AI Lab。Amazon / AWS contributes 2.3 times the visible source volume of second-ranked Microsoft. Meta / Instagram, Google, LinkedIn, and Bain show why sourcing should extend beyond AI labs.
Amazon / AWS
21
Microsoft
9
Meta / Instagram
8
LinkedIn
6
Bain & Company
6
Google
6
Slack
4
Cisco
4
MuleSoft
4
数据来源: Metix AIData source: Metix AI
六条最清晰的人才路径Six clearest talent routes
Amazon / AWS → Microsoft云平台 → 企业 AI 平台Cloud platform → enterprise AI platform
仅展示可识别的最近一次外部雇主流动；经历信息不完整的人才不计入路径。Only the latest observable external-employer move is shown; professionals without sufficiently complete histories are excluded from routes.
01
先锁定 JD 竞争组Fix the JD competition group first
先在十个统一岗位组中确定主要交付物，再列公司名单并比较同组公司的岗位与候选人池。First identify the primary deliverable within the unified ten-group taxonomy, then build the company list and compare companies and candidate pools inside that group.
02
再按瓶颈能力扩展证据Then expand evidence around the bottleneck
用运行、评测、集成、权限、状态与领域交付证据扩展搜索；单一框架名会过早缩窄候选人池。Expand the search with evidence of operation, evaluation, integration, permissions, state, and domain delivery; a single framework name narrows the pool too early.
03
最后选择来源生态Choose the source ecosystem last
Runtime 优先看云平台，客户工作流优先看咨询与交付，企业 Agent 平台优先看企业软件生态。Prioritize cloud platforms for runtime, consulting and delivery for customer workflows, and enterprise software ecosystems for enterprise Agent platforms.
08 · 附注08 · Notes
本报告聚焦公开可见的 Agent 岗位与人才证据This report focuses on publicly visible Agent job and talent evidence
覆盖说明Coverage notes
公开在招 Agent 岗位共 259 个，用于观察公司招聘重点。The report uses 259 publicly visible Agent postings to observe company hiring priorities.
可识别在职 Agent 人才共 513 人，用于观察人才存量、来源与流动。The report uses 513 observed current Agent professionals to study talent stock, sources, and movement.
岗位组：同一套十个岗位组连接招聘需求与人才存量，便于比较公司优先级。Job groups: the same ten job groups connect hiring demand with talent stock, making company priorities comparable.
流动：408 人可识别最近一次外部上一站，覆盖可识别人才的 79.5%。Flows: the latest prior external employer is observable for 408 people, covering 79.5% of the observed talent pool.
岗位组定义Job-group definitions
十个岗位组分别覆盖 Application Engineering、Research / Model Behavior、Runtime / Platform、Deployment Engineering、Product / Design、Post-training / Robustness、Evals / Quality、Orchestration / Workflow、Solutions Architecture 与 Safety / Governance。The ten job groups cover Application Engineering, Research / Model Behavior, Runtime / Platform, Deployment Engineering, Product / Design, Post-training / Robustness, Evals / Quality, Orchestration / Workflow, Solutions Architecture, and Safety / Governance.
岗位组以主要工作对象与职责为主，公司品牌、标题关键词和地点只作为辅助证据。Job groups prioritize primary work object and responsibilities, with company brand, title keywords, and location as supporting evidence.
能力、框架与领域词用于解释公司瓶颈；候选人竞争仍以主要岗位组为边界。Capabilities, frameworks, and domain terms explain company bottlenecks; candidate competition remains bounded by the primary job group.
解释边界Interpretation boundaries
513 是公开可见下限，不代表相关公司的完整 Agent 团队人数；不同岗位组的公开履历完整度不同。513 is a publicly visible lower bound, not the full Agent headcount at the covered companies; public profile completeness differs by job group.
岗位数代表可见招聘需求，不等于预算、录用人数或净扩张。Posting counts represent visible hiring demand, not budgets, hires, or net expansion.
可见岗位或人才较少的公司/岗位组，只用于说明招聘重点，不单独推断公司间人才竞争强度。Companies or job groups with fewer visible jobs or professionals are used to describe hiring focus, not to independently establish talent-competition intensity.
供需压力用于比较本报告覆盖范围内的岗位组优先级，不应解释为全市场空缺率。Demand-stock pressure compares job-group priorities within this report coverage and should not be read as a market-wide vacancy rate.
报告回答的问题Questions answered
现有 Agent 人才有多少，集中在哪些公司、城市和岗位组。How much current Agent talent is observed and where it concentrates by company, city, and job group.
十个岗位组的需求总量，以及各公司的主要招聘重点。Total demand across the ten job groups and each company's primary hiring focus.
招聘需求与人才存量在哪些岗位组上错配。Where hiring demand and current talent stock are mismatched by job group.
主要人才来源公司与可见流动路径是什么。Which employers are the main talent sources and what movement routes are visible.
需要把岗位竞争组转成可执行的人才搜索？Need to turn job competition groups into an actionable talent search?
Metix AI 可按目标公司的具体 JD、职级、地点与瓶颈能力，生成私有候选人池、竞争公司清单与验证证据。Metix AI can use a target company's specific JDs, seniority, location, and bottleneck capabilities to build a private candidate pool, competitor list, and validation evidence.
人才报告 · 不展示个人信息 · 由 Metix AI 提供Talent report · no personal information shown · provided by Metix AI