在投入数十亿美元构建人工智能之后，大型科技公司如今正大举投入，建设所需的人力队伍，以证明这些投资正在产生回报。
After spending billions building artificial intelligence, Big Tech is now spending heavily to build the human workforce needed to prove the investments are paying off.
各公司正争相寻找能够将这项技术应用于企业内部的专业AI工程师。这些专家被称为前沿部署工程师（FDE），他们将先进的AI和软件专业知识与对企业运作方式的理解结合在一起。他们可能是关键的一环——将强大的大语言模型转化为能够带来可衡量商业成果的企业系统所必需的人才。
Companies are scrambling to find specialized AI engineers who can put the technology to work inside businesses. These specialists — called forward deployed engineers, or FDEs — combine advanced AI and software expertise with an understanding of how corporations operate. They could be the key humans in the loop required to turn powerful large language models into enterprise systems that can deliver measurable business results.
本周，市场对AI热潮在经济上不可持续的担忧加剧，从戴尔到CoreWeave、AMD和美光等许多顶级AI概念股和存储概念股都遭遇了显著的短期损失。
Market concerns that the economics of the AI boom are not sustainable have intensified this week with many top AI-themed and memory-themed stocks from Dell to CoreWeave, AMD and Micron suffering significant short-term losses.
亚马逊云科技已承诺投入10亿美元用于一项前沿部署工程计划。OpenAI推出了一项部署业务，初期投资超过40亿美元。Anthropic将投入1亿美元，用于一项到2027年底培训1万名前沿部署工程师的计划。埃森哲和谷歌云正在组建一支1000人的前沿部署工程团队。
Amazon Web Services has committed $1 billion to a forward-deployed engineering initiative. OpenAI has launched a deployment business backed by more than $4 billion in initial investment. Anthropic is putting $100 million into a program to train 10,000 forward-deployed engineers by the end of 2027. And Accenture and Google Cloud are building a 1,000-person forward-deployed engineering workforce.
“华尔街开始对这些动辄数十亿、在AI上花钱却得不到任何商业成果的公司失去耐心，”高管猎头公司Christian & Timbers创始人杰夫·克里斯蒂安表示。
"The Street is starting to become impatient with these billions of companies who are spending on AI without getting any business outcomes," said Jeff Christian, founder of executive search firm Christian & Timbers.
前沿部署工程师直接与客户合作，梳理现有技术系统、安全要求和组织障碍，将有前景的AI实验转化为企业真正可用的应用。
Working directly with customers, FDEs navigate existing technology systems, security requirements, and organizational obstacles to turn promising AI experiments into applications companies can actually use.
在Caylent——一家将前沿部署工程师派驻客户现场的AWS咨询合作伙伴——首席执行官瓦尔·亨德森表示，这一职位所需的不仅仅是技术专长。
At Caylent, an AWS consulting partner that embeds FDEs with customers, CEO Val Henderson said the role requires more than technical expertise.
“毫无疑问，最难培养的技能就是工程师与客户之间的沟通能力（即‘客户情商’），”她说道。她还补充说，工程师还需要理解客户的需求和企业的运作方式，能够在项目遇到意外问题时做出明智的判断。
"The hardest thing to teach is the customer EQ, without question," she said, adding that engineers also need to understand customers and businesses, navigate organizational dynamics, and exercise judgment when projects encounter unexpected problems.
随着人才市场的竞争日益激烈，招聘需求也在迅速增加。根据劳动力市场分析公司 Lightcast 的数据，2023 年美国针对 FDE（Forward-Deployed Engineers，即负责将人工智能模型部署到实际应用中的工程师）的招聘职位仅有不到 200 个，到 2025 年这一数字上升到了约 1,200 个；而在 2026 年的前九个月里，这一数字更是达到了 7,300 个以上。同期，这些职位的招聘广告中给出的平均薪资为 195,456 美元，比 2023 年增长了 40%，是所有美国招聘职位平均薪资的三倍多。
A surge in job postings, salaries with scarce talent available Experienced talent is scarce, and developing that expertise can't happen overnight in a labor market race that recent hiring numbers suggest is accelerating. U.S. job postings for FDEs surged from fewer than 200 in 2023 to approximately 1,200 in 2025 and more than 7,300 in the first nine months of 2026, according to data from labor market analytics firm Lightcast. Median advertised salaries reached $195,456 during that period, up 40% from 2023 and more than three times the median advertised salary across all U.S. job postings.
这种对 FDE 工程师的竞争不仅限于那些开发人工智能模型的公司，咨询公司也是主要的招聘方之一，这反映了企业对能够帮助企业实施人工智能技术的工程师的迫切需求。Lightcast 统计显示，在 2026 年的前九个月里，德勤（Deloitte）发布了 872 个 FDE 招聘职位，埃森哲（Accenture）发布了 688 个，谷歌（Google）发布了 499 个，毕马威（KPMG）发布了 447 个。这些数据仅代表发布的招聘信息，并不代表实际完成招聘的数量。
The competition extends well beyond the companies developing AI models. Consulting firms are among the biggest recruiters, reflecting demand for engineers who can help corporate clients implement the technology. Lightcast counted 872 unique FDE job postings from Deloitte and 688 from Accenture during the first nine months of 2026, followed by Google with 499 and KPMG with 447. The figures represent advertised positions, not completed hires.
Lightcast 的全球研究负责人埃琳娜·马格里尼（Elena Magrini）表示：“对 FDE 工程师的需求正在快速增长，不断上涨的薪资表明雇主们正在争夺有限的人才资源。”
"Demand for forward deployed engineers is rising fast, and growing wages suggest employers are competing for a limited pool of talent," said Elena Magrini, head of global research at Lightcast.
不过，这些数据也指出了一个潜在的解决方案：“FDE 这个职位其实并不完全是一个全新的角色；它所要求的技能与软件工程有大约一半的重叠，因此我们可以利用更广泛的人才库来满足招聘需求。”
But the data also points to a potential solution. "FDE is not an entirely new role. It shares roughly half of its skills with software engineering, creating an opportunity to tap into a much broader talent pool," said Magrini.
科技企业及其企业客户正在通过多种方式来培养工程师在人工智能模型部署方面的专业能力。
Technology companies and their corporate clients are investing in multiple ways to develop deployment expertise.
AWS正在组建自己的FDE（Front-End Deployment Engineering）团队，该团队既吸收了亚马逊内部及AWS内部的资深工程师，也对外部人才进行了招聘。据AWS负责前端部署工程的副总裁Asa Kalavade介绍，该团队最初由数百名工程师组成。
AWS is building its own FDE organization, drawing on experienced engineers within Amazon and AWS while also recruiting externally. The company is starting with hundreds of engineers, according to Asa Kalavade, vice president of forward-deployed engineering at AWS.
Kalavade表示：“我们目前正在投入10亿美元，因为客户迫切需要这种实际的帮助，以快速将他们在AI领域的投资转化为实际成果。”
"We are investing $1 billion now because customers need this hands-on help turning their AI investments into results at speed," Kalavade said.
OpenAI则采取了更为广泛的部署策略：该公司获得了40亿美元的初始投资，并在今年早些时候达成协议，收购了专注于AI咨询和工程服务的公司Tomoro；Tomoro预计将为OpenAI带来约150名经验丰富的FDE专家和部署专家。
OpenAI is pursuing a broader enterprise deployment strategy, backed by the $4 billion in initial investment and an agreement earlier this year to acquire Tomoro, an applied AI consulting and engineering firm expected to bring approximately 150 experienced FDEs and deployment specialists into the operation.
AI技能缺口已成为一个普遍的市场问题 Anthropic采取了不同的方法：它直接培训其客户和合作伙伴现有的工程师。通过其“Claude Frontier Academy”项目，Anthropic计划在2027年底前培养出1万名FDE专家。首批参与培训的工程师来自Accenture、Bain、Capgemini、Commonwealth Bank of Australia、Deloitte、McKinsey、Morgan Stanley和Novo Nordisk等机构。
AI skills gap is a 'market problem' Anthropic is taking a different approach, training engineers already employed by its customers and partners. Through its Claude Frontier Academy, the company aims to train 10,000 FDEs by the end of 2027. Its first residency cohorts include engineers from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley, and Novo Nordisk.
Anthropic的全球业务发展与合作伙伴关系负责人Steve Corfield表示：“AI技能缺口影响着每一位客户和合作伙伴，因此这是一个市场问题，而Anthropic致力于帮助解决这一问题。”
"The AI skills gap affects every customer and every partner, so it is a market problem, and Anthropic is committed to helping close it," said Steve Corfield, Anthropic's global head of business development and partnerships.
与此同时，Accenture和Google Cloud也在积极扩充其FDE团队规模，计划培养1000名FDE专家；他们主要通过重新培训现有的软件工程师来满足客户在AI部署方面的需求。
Accenture and Google Cloud, meanwhile, are building a 1,000-person FDE workforce, relying partly on retraining experienced software engineers for customer-facing AI deployment work.
“现实情况是，经验丰富的FDE无法以客户所需的规模被直接招募进来，”安永（Accenture）的FDE及软件与平台工程负责人Ram Ramalingam表示。Ramalingam指出，根据候选人的背景不同，该公司对FDE的培训和入职过程可能需要30到90天。他还补充说，参与者本身已经具备丰富的软件工程经验，该计划旨在为其增添专门的部署能力，而非从零开始培训新手。
"The reality is that experienced FDEs cannot simply be recruited at the scale clients need," said Ram Ramalingam, Accenture's FDE and software and platform engineering lead. Depending on their backgrounds, the company's FDE training and onboarding can take 30 to 90 days, Ramalingam said. He added that participants already have substantial software engineering experience; the program is designed to add specialized deployment capabilities rather than train beginners.
AWS也看到了在客户组织内部培养专业能力的机会。Kalavade表示，许多企业已经拥有能力出众的工程师，但需要帮助了解如何将AI项目投入生产环境。AWS的FDE会与这些团队并肩合作，以便客户最终能够自行维护和扩展系统。
AWS also sees an opportunity to develop expertise inside customer organizations. Kalavade said many businesses already employ capable engineers but need help learning how to move AI projects into production. AWS's FDEs work alongside those teams so customers can eventually maintain and expand the systems themselves.
Corfield表示，挑战不仅限于技术专长。“目前稀缺的一项技能，是既能够理解企业运作方式，又能够构建改变企业格局的系统的工程师，”他说。
The challenge goes beyond technical expertise, said Corfield. "A skill in short supply is an engineer who can understand how the business works and also build the system that changes it," he said.
Anthropic认为，部分这类专业能力可以通过结构化的工程训练来培养。其“Frontier部署工程师驻留计划”借鉴了医学教育模式，将经验丰富的工程师与行业从业者配对，并通过模拟真实企业部署场景对他们进行评估。
Anthropic believes some of that expertise can be developed through structured engineering. Its Frontier Deployed Engineer Residency borrows from medical education, pairing experienced engineers with practitioners and evaluating them on realistic enterprise deployments.
“我们将该驻留计划的设计模式参考了医学领域培训医生的方式，”Corfield表示。据Anthropic介绍，参与者会向在职工程师学习，处理真实案例，并在独立为各自公司实践之前接受评估。随后，他们将在参与Claude项目的同时，获得Anthropic工程师为期12周的指导。首批完整的认证预计将于2027年初颁发。
"We modeled the residency on how medicine trains doctors," Corfield said. Participants learn from practicing engineers, work on realistic cases, and are assessed before they practice on their own companies, according to Anthropic. They then receive 12 weeks of mentorship from Anthropic engineers while working on Claude projects. The first full credentials are expected in early 2027.
现在分析部署成果还为时过早。克里斯蒂安表示，经验丰富的FDE可以通过领导经验较少的工程师团队，帮助扩大工作队伍。例如，他说，那些成功完成10到15次部署的人员，可能具备指导五到六人团队的判断力，并将自己在处理复杂客户项目过程中积累的专业知识传授给他人。
Too soon to analyze results from deployments According to Christian, experienced FDEs can help expand the workforce by leading teams of less-experienced engineers. For instance, he said, those who have completed 10 to 15 successful deployments may have the judgment to supervise teams of five or six, passing along expertise gained from working through complicated customer projects.
克里斯蒂安认为，这种导师制模式是一种扩大部署团队的方式，而无需在每个岗位上都配备顶尖工程师。
Christian sees that mentorship model as a way to expand deployment teams without requiring an elite engineer in every position.
但亨德森表示，复杂AI部署所需的面向客户的判断力尤其难以传授。虽然技术技能可以通过培训获得，但工程师还需要具备应对现实世界中客户挑战的经验。“经验是无法压缩的算法，”亨德森说。
But Henderson said the customer-facing judgment required for complex AI deployments is particularly difficult to teach. While technical skills can be developed through training, engineers also need experience navigating real-world customer challenges. "There is no compression algorithm for experience," Henderson said.
对克里斯蒂安而言，挑战最终仍源于投资者要求产生业务成果的压力。他认为，公司需要能够直接与客户合作、将日益强大的AI系统转化为实际应用能力的工程师。“让AI真正发挥作用，唯一的方法就是人与人之间的互动，”他说。
For Christian, the challenge comes back to the pressure from investors to produce business results. He argued that companies need engineers who can work directly with customers to translate increasingly powerful AI systems into practical applications. "The only way you can get AI to work is by human-to-human interaction," he said.
最终的检验在于，这支不断扩大的团队是否能够帮助公司从AI中获得足以抵消初期投入的投资回报。迄今为止，来自FDE项目的现有证据尚不足以证实这一点。根据企业报告中的案例研究，一些早期部署虽然显示出生产力提升，但尚未带来财务回报。
The ultimate test is whether this expanding workforce can help companies generate the return on investment from AI that justifies the upfront costs. To date, the available evidence from FDE projects falls short of confirmation. Some early deployments show productivity improvements but not financial returns, according to company-reported case studies.
在抵押贷款机构PennyMac，Caylent的FDE与产品经理和工程师共同开发了一个基于AI的抵押贷款服务平台。根据AWS提供的案例研究，某些工程任务的完成速度最高可达原来的六倍，具体取决于任务类型。
At mortgage lender PennyMac, Caylent's FDEs worked alongside product managers and engineers on an AI-enabled mortgage servicing platform. Some engineering tasks were completed as much as six times faster, depending on the task, according to a case study provided by AWS.
Caylent承接的另一项项目是为通信合规软件供应商Smarsh提供服务。据该公司发布的案例研究显示，该项目成功释放了价值500万至1000万美元的工程与产品开发资源，这些资源随后被调配至其他业务领域。Caylent证实确实有FDEs参与了该项目，不过该数据并未单独体现他们的贡献程度，也不代表相应的资金节约额。
Another Caylent engagement involved Smarsh, a communications compliance software provider. A company case study estimated that the project freed $5 million to $10 million in engineering and product capacity that could be redirected to other work. Caylent confirmed that FDEs participated in the project, although the figure does not isolate their contribution or represent equivalent cash savings.
然而，这些工程效率与产能方面的提升尚无法作为人工智能投资成效的独立实证依据。
These improvements in engineering productivity and capacity are not independently verified returns on AI investment.
Anthropic推行的这项大规模培训计划目前尚未取得可量化的成果。Corfield坦言目前获取成效数据仍为时过早，并称公司计划持续监测参与者所开发的项目何时能投入实际应用，以及这些系统的使用频率与范围。
Anthropic's ambitious training initiative has yet to demonstrate measurable results. Corfield acknowledged that it is still too early for outcome data and said the company plans to track how quickly participants' projects reach production and how extensively those systems are used.
