下一批伟大的AI公司可能看起来像服务公司

Broadly speaking, software is credited for streamlining messy industries by making them vastly more legible and manageable. And in many cases, this reputation is well deserved. After all, software holds the power to systemize record-keeping, digitize workflows, and scale processes far more efficiently than paper, spreadsheets, and phone calls ever could.

But until recently, software’s reach stopped short at the edge of the work itself. While a bookkeeping platform may have handily organized the books, someone still had to close them. Sure, insurance systems could manage policies, but someone still had to marshal data, negotiate coverage, and walk customers through the process. And although logistics software tracked shipments, humans still had to coordinate carriers, resolve exceptions, and troubleshoot problems.

That all began changing with the next generation of AI, whose most consequential innovation isn’t merely improving software in the old sense, but its ability to compress that coordination layer and build better service models on top of it. Within this paradigm shift, the pertinent question isn’t whether AI hastens existing workflows; it’s whether an AI-native operator can outperform the incumbent that bolts AI onto the old workflow framework.

An AI-native services company is an enterprise that sells a completed service or outcome to the customer, while commandeering AI as a core part of the machinery that delivers it. The customer is not primarily buying a software seat. They’re buying insurance, bookkeeping, benefits access, home maintenance, compliance, and other finished jobs. Underneath that experience--software, models, agents, and human operators all collaborate to deliver desired outcomes.

This distinction matters. These businesses may seem more like brokers, administrators, agencies, or managed service providers than traditional SaaS companies. But their economics can fundamentally differ from the services businesses that came before them because AI sits inside the production function.

Consequently, the question is no longer simply whether AI can make an existing workflow 20 or 30 percent faster. It’s whether a company built around AI from day one can deliver the service faster, cheaper, and more reliably than an incumbent attempting to layer AI onto a labor model and organizational structure built decades ago.

中小型企业（SMBs）是这种模式极具吸引力的客户。能够轻松负担专业团队的大型企业有能力购买软件并配备人员来操作，从而支持采购部门、福利部门、财务团队、合规专家和其他运营人员。相反，大多数中小企业没有足够雄厚的财力来设立这些职能。

一家拥有30名员工的公司老板不一定想要更好的保险软件；他们想要的是公司获得保险。他们不想要另一个记账仪表盘；他们想要月底准确的账目。他们不想要采购工作流程；他们希望有人来寻找供应商并谈判价格。

这就是为什么中小企业已经在服务上花费巨大，将那些缺乏专业知识或人力来建立内部团队的职能外包出去。这使得中小企业成为AI原生服务公司异常有吸引力的市场切入点。初创公司通常不必说服客户创建新的软件预算并改变工作方式，而是可以针对现有的服务预算，提供客户已经付费购买的东西的更好版本。

推销话术从“这是一款能让你的员工更有效率的软件”转变为“把工作交给我们”。

这比在SaaS中添加AI功能是一个更为重大的转变。

获得最大吸引力的初创公司并不会试图全盘重建运营。相反，它们解决一个足够痛苦、紧急或昂贵的单一断裂切入点，使客户愿意接受行为改变。

Corgi就是一个典型例子。它们的切入点是初创企业保险，这一类别中，保险覆盖直接位于融资、董事会组建或关闭客户的关键路径上。Corgi自称AI原生，是一家为初创企业提供全栈保险平台和承保方，而不仅仅是“保险软件提供商”。因此，Corgi将软件和AI包裹在一个繁琐且时间紧迫的服务周围，使客户体验更快、更可靠。

这种切入点优先的方法很重要。成功的公司不会在第一天就一举取代整个类别，而是挑选足够痛苦、足以要求行为改变的狭窄任务，从而为更周到的宏观方法铺平道路。

最具吸引力的机会具有几个特征：庞大的现有服务支出、分散的供应商、重复的协调工作、昂贵的人力劳动，以及更关心结果而非工作如何完成的最终客户。

中小企业后台服务可能是最明显的例子。保险、记账、税务、工资、合规、采购和其他行政职能在小型企业内部消耗大量时间和金钱，尽管很少成为战略差异化因素。在这些类别中，客户往往希望完全将责任从自己身上卸下。

医疗保健管理具有类似特征。医疗保健系统的大部分制约并非来自缺乏软件，而是来自 navigating 机构、文书工作、资格要求、付款方和政府项目所需的巨大协调工作。

Mindset Care是一个帮助用户获取财务资源的技术平台，体现了医疗保健管理的消费者一侧。通过以无前期成本的方式巧妙处理SSI和SSDI申请，它在传统上缓慢、手动且难以导航的文书密集型工作流程周围，包裹了个性化、端到端的支持。

在物流和实体运营中，基础工作仍然分散在电话、电子邮件、电子表格、供应商和操作人员之间。AI不需要物理移动托盘或修理机器来改变这些业务。在报价、路线规划、沟通、文档、异常管理和跟进等物理行动周围处理一切事务，具有巨大价值。

这一论点的更具逆向思维的延伸出现在消费者一侧，即在供应和客户体验出现问题的市场。家庭服务是一个明显例子，挑战在于寻找可信赖、高度响应的人员，协调工作，并管理跟进协议。这不是关于是否有人出现，而是关于是否有人能够建立一家AI原生的家庭服务公司，让房主从平台购买可靠性和便利性，而不是每次都要审查不同的个体提供商。

Casa是最近的一个例子。Forerunner没有将其框定为另一个点解决方案，而是将Casa宣传为围绕家庭的协调服务层，建立在数字家庭记录、礼宾层和底层管理的供应商网络之上。在这种模式中，公司位于分散的供应之上，并围绕最终客户重新组织体验。

一个含义是，当用传统SaaS的视角来评估这些业务时，它们可能显得很奇怪。收入最初可能类似于服务收入。毛利率可能更低。人类可能仍深度参与交付过程。关键问题在于，随着公司规模扩大，其经济特征会发生什么变化。

在传统服务业务中，收入增长通常需要按完成的工作量成比例地增加劳动力。在AI原生服务业务中，目标就是打破这种关系。随着系统改进，单个运营人员可以管理更多客户。更多交互可以实现自动化。异常可以更早被标记出来。定价和路由可以改善。响应时间可以缩短。曾经需要专业员工完成的工作，可能越来越多地由软件处理，而人类则保留用于最高价值或最模糊的决策。

因此，风险投资级别的规模化机会，不一定是第一天就具备SaaS利润率的服务业务。它是一种随着更多生产过程转移到软件中，交付服务的边际成本持续下降的业务。

That is why the best AI-native services companies may not look like classic SaaS businesses from the outside. Instead, they resemble operators, brokers, administrators, or managed service providers. But the software is still there, moving inside the company, behind the scenes. The customer is buying the finished work, while software, models, agents, and human labor are orchestrated underneath.

That also means the moat may not initially look like classic SaaS. It may show up in routing, pricing, utilization, service quality, response times, trust, and supply-side density. In some categories, this comes from owning the relationship with demand. In others, it comes from owning more of the workflow itself. Either way, the defensibility compounds as the system learns.

Underneath these startups are software companies in a different form. The customer will not be buying the software. They will be buying the work.

The views expressed are those of the author at the time of writing. Other teams may hold different views and make different investment decisions. The value of your investment may become worth more or less than at the time of original investment. While any third-party data used is considered reliable, its accuracy is not guaranteed. For professional, institutional or accredited investors only.

Get our latest market insights straight to your inbox.

You will shortly receive an email with your unique link to our preference center

Our three-part video series brings together experts from across our venture capital platform to explore opportunities outside of AI, how to identify quality among today’s noise, and what investors may be underestimating for the year ahead.

Collaboration in practice: Early-stage venture

Jackson Cummings, head of Wellington Access Ventures, outlines how collaborating with our later-stage private teams, public market industry analysts, and Value Creation Team can benefit his early-stage venture team.

The transformative power of vertical AI agents

Olivia Hurley and Rie Kimoto explore Japan's banking sector, highlighting regional banks' undervaluation and M&A opportunities driven by regulatory changes and demographic shifts.

Building the future: Advances in AI infrastructure for autonomous agents

Van Jones highlights AI infrastructure advancements, emphasizing emerging protocols and data solutions for autonomous agents driving innovation and connectivity.

Early-stage venture: Innovation versus hype

Wellington Access Ventures' Jackson Cummings discusses promising innovations in edge AI and embedded fintech. In addition, he highlights the potential hype surrounding consumer AI and quantum computing.

Early-stage venture: Deployment, valuations, exits

Head of Wellington Access Ventures Jackson Cummings discusses his latest views on early-stage venture capital and why he’s “cautiously optimistic” on deployments, valuations, and exits.

Venture capital outlook for 2025: 5 key trends

Our venture capital experts are encouraged by the healthier environment for venture capital in 2025. In this year's outlook, they explore key themes for the year ahead, including: higher deployment, AI, rebounding IPOs, and impacts of the 2024 election.