a16z报告：AI native商业模式的范式转移 a16z Report: The Paradigm Shift of AI-Native Business Models
where's next aha moment?
a16z最近的报告非常值得一看，我整理了一些核心的观点。
1、软件即劳动力是最大增量
SaaS 行业的逻辑正在发生根本性转变，从卖工具进化为直接交付工作结果。过去企业按人头购买软件（如 Office）来辅助员工工作，而 AI 时代，软件将直接取代劳动力完成任务。
SaaS行业的黄金法则一直是：将原本由人完成的工作流程化、工具化，然后按人头向企业收费。但是在a16z看来，这个逻辑正在失效，AI应用正在进入新的阶段，软件即劳动力。「每个人都想省钱，但每个人更想赚钱」当软件不再只是工具，而是直接交付结果时，客户意愿支付的就不再是每个人每月几美元的订阅费，而成为了结果的分成。
2、专有数据是唯一的围墙花园
随着 OpenAI、Google 等巨头将大模型能力不断推高，模型本身的稀缺性正在下降。在模型日益商品化的今天，专有数据（Proprietary Data）成为了唯一的围墙花园。
a16z 认为人性底层的逻辑非常简单：每个人都想要两样东西——变得更富和变得更懒。对于企业来说，采用 AI 不仅仅是为了降本（更懒），更是为了直接创造营收（更富）。从 Ramp 的数据来看，企业在 AI 上的支出在 2025 年 1 月出现了一次巨大的跃升。
不同于移动互联网时代仅仅是把电脑装进口袋，AI 时代的变革是叠加在过去五十年 PC、互联网、云和移动技术之上的。它面对的是全球 80 亿已经联网的用户，扩散速度前所未有。
虽然外界对 AI 泡沫的担忧从未停止，但在 a16z 的投资版图中，那些能在数据上构建围墙、在业务上直接交付结果的公司，正以 0 到 1 亿美元营收的最快纪录，证明着这个时代的真实性。对于创业者而言，现在不是担心巨头的时候，而是去寻找那些还未被数字化的角落，用 AI 把苦力活变成印钞机的最佳时机。
3、商业模式变革:从卖原材料转向卖成品
在传统的数据商业模式中,公司像 PitchBook、Crunchbase 这样售卖原始数据订阅,客户需要自己花时间分析处理。但在 AI 时代,这种模式的价值被大幅压缩——因为 AI 可以快速处理海量数据,原材料本身不再稀缺。
真正的价值转移发生在”成品”层面。与其让客户购买数据后自己分析,不如直接基于独家数据源用 AI 生成完整的投资备忘录、行业分析报告或尽职调查文档。这就像从”卖菜市场的新鲜蔬菜”升级到”卖米其林餐厅的精致套餐”——后者的定价权和利润空间可以是前者的 10-100 倍。
关键在于:你的数据优势要体现在最终交付物中,而不是让客户自己去挖掘价值。这种转变需要公司重新思考产品形态、定价策略和客户获取方式。
4、巨头的防御：拥有人而非客户
传统软件巨头如 Salesforce、Workday、Oracle 看似臃肿低效,但它们有一个致命优势:它们控制着企业的”记录系统”(System of Record)——所有客户数据、销售历史、员工信息都存储在他们的系统中。
这种锁定效应造就了人质经济：迁移成本极高(数据迁移、员工再培训、业务流程重构),企业即使不满意也难以离开。巨头们深知这一点,因此可以采取”捆绑销售”策略——在原有系统基础上增加 AI Copilot 功能,然后涨价 20-30%。客户虽然抱怨,但算上迁移成本和风险,最终还是会接受。
这就是为什么颠覆巨头比想象中困难得多：不仅要提供更好的产品,还要说服客户承担巨大的迁移成本。
5、垂直整合服务：AI 时代的门口野蛮人
传统 SaaS 创业的困境在于:你开发了一个很棒的工具,但说服律师、会计师、医生这些专业人士改变工作习惯极其困难。销售周期长、客单价低、客户教育成本高。
AI 时代提供了新思路:与其卖工具给专业服务机构,不如直接收购一家小型事务所,用 AI 重构其服务流程。比如收购一家 20 人的会计师事务所,用 AI 将人效提升 5-10 倍,然后以更低价格服务 10 倍的客户量。
这种模式的优势在于:
即时验证:真实业务环境中测试 AI 能力
跳过销售:不需要说服外部客户采用新工具
掌握专业知识:收购团队带来的行业 know-how 是 AI 训练的最佳素材
规模化路径清晰:证明模式后可以快速复制到更多地区或细分领域
6、AI 重构劳动力价值方程：增强而非单纯替代
当前的 AI 变革常被误读为”大规模失业浪潮”,但实际情况更复杂。AI 的价值主要体现在三个维度:
成本优势：AI 可以 7×24 小时工作,不需要休假、保险、情绪管理,边际成本接近零。但这不意味着它会完全替代人类,而是接管那些”人类不想做”的重复性、低价值工作——比如初步筛选简历、回复常见客服问题、生成初稿报告。
能力增强：一个会计师用上 AI 工具后,可能从服务 50 个客户提升到 200 个客户,但她的角色从”做账”变成”审核 AI 做账结果+提供战略建议”。这是增强,不是替代。
市场扩容：AI 降低服务成本后,原本买不起专业服务的小企业也能负担得起,市场规模反而扩大。这意味着需要更多人来服务这个扩大的市场,只是工作内容发生了变化。
未来的工作形态更像是人机协作：AI 处理可标准化的部分,人类专注于创造性、战略性和人际关系部分。那些能善用 AI 的人,生产力会呈指数级增长。
7、消费者 AI 的机会在于聚合与新类别
在消费者 AI 市场,存在两种主要机会：
原生新类别:创造此前不存在的全新体验。比如 ElevenLabs 开创的实时语音克隆市场、Character.AI 的 AI 陪伴市场。这些不是在改进现有产品,而是创造全新需求。难点在于市场教育成本高,但一旦成功就能成为品类定义者。
模型聚合平台:OpenAI、Anthropic、Google 这些大厂受限于自家模型,他们不会(也不方便)提供”跨模型比价与选择”服务。这就给第三方聚合者留下巨大空间。比如 Perplexity（搜索聚合）、Poe（对话聚合）这样的产品，让用户可以:
在一个界面访问多个模型
根据任务选择性价比最优的模型
享受更好的用户体验和额外功能层
a16z’s recent report is well worth reading. I’ve compiled some core insights.
1. Software as Labor is the Biggest Incremental Value
The logic of the SaaS industry is undergoing a fundamental shift, evolving from selling tools to directly delivering work outcomes. In the past, companies purchased software per seat (like Office) to assist employees, but in the AI era, software will directly replace labor to complete tasks.
The golden rule of the SaaS industry has always been: systematize and tool-ify work previously done by humans, then charge companies per seat. But according to a16z, this logic is becoming obsolete. AI applications are entering a new phase: software as labor. “Everyone wants to save money, but everyone wants to make money even more.” When software is no longer just a tool but directly delivers results, customers are willing to pay not a few dollars per person per month in subscription fees, but a share of the outcomes.
2. Proprietary Data is the Only Walled Garden
As giants like OpenAI and Google continue to push the capabilities of large models higher, the scarcity of models themselves is declining. In today’s increasingly commoditized model landscape, proprietary data has become the only walled garden.
Open Evidence is a typical case. While ChatGPT can also answer medical questions, Open Evidence has exclusive licenses to core medical literature like the New England Journal of Medicine. Answers built on this closed data cannot be obtained by general large models through public crawling. When AI gains understanding and reasoning capabilities, dormant data becomes a gold mine.
a16z believes the underlying logic of human nature is very simple: everyone wants two things—to become richer and to become lazier. For enterprises, adopting AI is not just about cost reduction (being lazier), but more about directly creating revenue (becoming richer). According to Ramp’s data, enterprise spending on AI saw a massive jump in January 2025.
This is real productivity landing.
Unlike the mobile internet era which simply put computers in pockets, the AI era’s transformation is layered on top of fifty years of PC, internet, cloud, and mobile technology. It faces 8 billion already-connected users globally, with an unprecedented diffusion speed.
While concerns about the AI bubble never cease, in a16z’s investment portfolio, companies that can build walls around data and directly deliver results in business are setting records for the fastest time from 0 to $100 million in revenue, proving the reality of this era. For entrepreneurs, now is not the time to worry about giants, but the best moment to find corners that haven’t been digitized yet and use AI to turn grunt work into money-printing machines.
3. Business Model Transformation: From Selling Raw Materials to Selling Finished Products
In traditional data business models, companies like PitchBook and Crunchbase sell raw data subscriptions, requiring customers to spend time analyzing and processing themselves. But in the AI era, this model’s value is greatly compressed—because AI can quickly process massive amounts of data, raw materials themselves are no longer scarce.
The real value shift occurs at the “finished product” level. Rather than having customers purchase data and analyze it themselves, it’s better to directly use AI to generate complete investment memos, industry analysis reports, or due diligence documents based on exclusive data sources. This is like upgrading from “selling fresh vegetables at the farmer’s market” to “selling refined set menus at Michelin restaurants”—the latter’s pricing power and profit margins can be 10-100 times the former.
The key is: your data advantage must be reflected in the final deliverables, not letting customers dig out the value themselves. This transformation requires companies to rethink product form, pricing strategy, and customer acquisition methods.
4. The Giants’ Defense: Holding Hostages, Not Customers
Traditional software giants like Salesforce, Workday, and Oracle seem bloated and inefficient, but they have one fatal advantage: they control enterprises’ “System of Record”—all customer data, sales history, and employee information are stored in their systems.
This lock-in effect creates a “hostage economy”: migration costs are extremely high (data migration, employee retraining, business process restructuring), and even if enterprises are dissatisfied, they find it hard to leave. The giants know this well, so they can adopt “bundling sales” strategies—adding AI Copilot features on top of existing systems, then raising prices 20-30%. Customers complain, but factoring in migration costs and risks, they ultimately accept.
This is why disrupting giants is much harder than imagined: you not only need to provide a better product, but also convince customers to bear enormous migration costs. Unless your product has a 10x advantage, it’s difficult to shake these players with “moats.”
5. Vertical Service Integration: Barbarians at the Gate in the AI Era
The dilemma of traditional SaaS startups is: you’ve developed a great tool, but convincing professionals like lawyers, accountants, and doctors to change their work habits is extremely difficult. Long sales cycles, low customer unit prices, high customer education costs.
The AI era provides new thinking: rather than selling tools to professional service firms, why not directly acquire a small firm and use AI to restructure its service processes? For example, acquire a 20-person accounting firm, use AI to increase per-person efficiency 5-10x, then serve 10x the customer volume at lower prices.
The advantages of this model include:
Instant validation: Test AI capabilities in real business environments
Skip sales: No need to convince external customers to adopt new tools
Master domain expertise: The acquired team’s industry know-how is the best material for AI training
Clear scalability path: After proving the model, quickly replicate to more regions or segments
6. AI Reconstructs the Labor Value Equation: Augmentation, Not Simple Replacement
The current AI transformation is often misread as a “massive unemployment wave,” but the reality is more complex. AI’s value is mainly reflected in three dimensions:
Cost advantage: AI can work 24/7, doesn’t need vacation, insurance, emotional management, with marginal costs approaching zero. But this doesn’t mean it will completely replace humans; rather, it takes over repetitive, low-value work that “humans don’t want to do”—like preliminary resume screening, responding to common customer service questions, generating draft reports.
Capability augmentation: An accountant using AI tools might go from serving 50 clients to 200 clients, but their role shifts from “bookkeeping” to “auditing AI bookkeeping results + providing strategic advice.” This is augmentation, not replacement.
Market expansion: After AI lowers service costs, small businesses that previously couldn’t afford professional services can now afford them, and the market size actually expands. This means more people are needed to serve this expanded market, just with changed work content.
The future work format is more like human-machine collaboration: AI handles the standardizable parts, humans focus on creative, strategic, and interpersonal aspects. Those who can leverage AI well will see exponential productivity growth.
7. Consumer AI Opportunities Lie in Aggregation and New Categories
In the consumer AI market, two main opportunities exist:
Native new categories: Create entirely new experiences that didn’t exist before. For example, ElevenLabs’ real-time voice cloning market, Character.AI’s AI companion market. These aren’t improving existing products, but creating entirely new demands. The difficulty lies in high market education costs, but once successful, you become the category definer.
Model aggregation platforms: Giants like OpenAI, Anthropic, and Google are limited to their own models; they won’t (and can’t conveniently) provide “cross-model comparison and selection” services. This leaves huge space for third-party aggregators. Products like Perplexity (search aggregation) and Poe (conversation aggregation) allow users to:
Access multiple models in one interface
Choose the most cost-effective model based on task
Enjoy better user experience and additional feature layers