AI浪潮：软件行业的第五次范式转变
我们正站在一个激动人心的时刻，AI应用的黄金时代似乎已然来临，标志着软件行业有史以来最剧烈的范式转变。为了理解为何是现在，而非过去或未来，我们需要将当前的AI应用浪潮置于更宏大的历史视角下审视。回顾从1977年至今的纳斯达克指数走势，软件行业经历了四次重大的产品周期：PC时代（以苹果、微软为基石，催生了Lotus、Adobe等应用巨头）、互联网时代（思科搭建基础设施，eBay、亚马逊等平台涌现）、云计算时代（AWS成为核心，Workday、Shopify等公司崛起），以及移动时代（智能手机普及，重塑生活工作）。如今，我们正站在第五次浪潮——AI时代的开端。AI的迅速普及得益于前几次技术浪潮打下的基础，特别是智能手机和云计算的普及。数据显示，全球约15%的成年人每周使用ChatGPT，这一用户采用速度是前所未有的。企业支出数据也显示，成熟企业正将AI视为节省时间、降低成本的实用工具，而非仅是炫技的演示。在这样一个AI驱动的时代，构建持久的竞争优势成为关键。
Original English
We are standing at an exciting moment, where the golden age of AI applications seems to have arrived, marking the most profound paradigm shift in the software industry's history. To understand why now, rather than in the past or future, we must place the current AI application wave within a broader historical context. Looking back at the Nasdaq index from 1977 to the present, the software industry has experienced four major product cycles: the PC era (with Apple and Microsoft as cornerstones, giving rise to application giants like Lotus and Adobe), the Internet era (Cisco built the infrastructure, leading to platforms like eBay and Amazon), the Cloud Computing era (AWS became the core, with companies like Workday and Shopify rising), and the Mobile era (smartphone ubiquity reshaping life and work). Today, we stand at the dawn of the fifth wave—the AI era. The rapid adoption of AI benefits from the foundations laid by previous technological waves, particularly the widespread use of smartphones and cloud computing. Data shows that approximately 15% of adults worldwide use ChatGPT weekly, an unprecedented rate of user adoption. Enterprise spending data also indicates that mature companies are increasingly viewing AI as a practical tool for saving time and reducing costs, rather than just a demonstration of capability. In this AI-driven era, building sustainable competitive advantages has become critical.AI原生：重塑传统软件格局
第一个核心主题是传统软件正全面转向AI原生。正如过去投资云原生公司获得了丰厚回报一样，AI原生公司正重新定义各个软件类别。传统本地部署软件公司因其一次性收费模式和对订阅模式的不适应，难以应对云计算的转变；如今，同样的转变正在AI领域发生。从企业资源规划（ERP）、客户关系管理（CRM）到客户支持，AI原生公司有机会重新定义这些领域。区分绿地机会（服务全新客户）和棕地机会（争夺现有客户）至关重要。当公司发展到关键转折点时，它们会选择市场上最好的AI原生产品，而非被锁定在旧系统中。现有的大公司虽然也会因AI而增强，但它们的核心优势在于强大的护城河和作为企业记录系统的地位，这使得它们拥有“人质”而非仅仅是客户。AI原生公司要么通过AI功能切入市场，要么直接构建新的记录系统。与云计算时代不同，AI的普遍认同加速了其市场采用速度，因为几乎所有人都认为AI是一个好主意。
Original English
The first core theme is the comprehensive shift of traditional software towards **AI-native**. Just as investing in cloud-native companies yielded significant returns in the past, AI-native companies are redefining various software categories. Traditional on-premise software companies struggled to adapt to the cloud transition due to their one-time fee models and inability to embrace subscription services; today, a similar transformation is occurring in the AI domain. From Enterprise Resource Planning (ERP), Customer Relationship Management (CRM), to customer support, AI-native companies have the opportunity to redefine these sectors. Differentiating between **greenfield opportunities** (serving entirely new customers) and **brownfield opportunities** (competing for existing customers) is crucial. When companies reach critical inflection points, they opt for the best AI-native products on the market rather than being locked into legacy systems. While established large companies will also be strengthened by AI, their core advantage lies in their robust **moats** and their position as the enterprise **record system**, which gives them "hostages" rather than mere customers. AI-native companies either penetrate the market with an AI feature or directly build new record systems. Unlike the cloud era, the widespread acceptance of AI is accelerating its market adoption, as almost everyone considers AI a good idea.劳动力替代：开辟全新软件市场
第二个核心主题是全新软件类别的出现，这类软件开始直接替代劳动力，从而开辟了一个比传统软件市场大得多的新市场。这不再是软件与软件的竞争，而是与人力成本的竞争。AI创造了一个介于传统软件和劳动力之间的全新市场，其定价区间远超传统软件，但低于全职雇佣的劳动力成本。例如，一个年薪4.7万美元的前台接待岗位，若能被一个软件产品替代其部分职责，企业可能愿意支付约2万美元。这种可能性解锁了人们“更加富有和更加懒惰”的愿望，即用更少的工作获得更多经济价值。Eve公司为原告方律师提供AI解决方案，通过提高效率，显著增加了律师的收入潜力，而非像企业律所那样削减工时费。Eve通过端到端的工作流和专有数据，构建了强大的护城河。同样，Salient公司通过AI优化汽车贷款催收和保险理赔，不仅节省成本，更大幅增加了客户收入，其价值主张几乎无法拒绝。AI在这些场景中提供了比人力更稳定、合规且高效的服务。
Original English
The second core theme is the emergence of entirely new software categories that directly replace labor, thereby opening up a market significantly larger than traditional software. This is no longer a competition between software and software, but between software and **labor costs**. AI creates a new market situated between traditional software and labor, with pricing tiers far exceeding traditional software but falling below the cost of full-time employment. For instance, a receptionist role costing $47,000 annually could potentially be replaced by a software product handling some of its duties, for which a company might be willing to pay around $20,000. This possibility unlocks the desire for people to be "richer and lazier," achieving greater economic value with less work. **Eve**, a company providing AI solutions for plaintiff lawyers, significantly increases their earning potential by enhancing efficiency, unlike corporate law firms that might reduce billable hours. Eve builds a strong moat through its end-to-end workflow and proprietary data. Similarly, **Salient**, through AI optimization of auto loan collections and insurance claims, not only saves costs but also substantially increases client revenue, making its value proposition almost irresistible. In these scenarios, AI offers more stable, compliant, and efficient services than human labor.围墙花园：专有数据构筑的竞争壁垒
第三个核心主题是围墙花园，即基于专有数据建立的护城河。OpenAI等大模型公司如同“蔬菜农场”，提供基础算力，但当它们开始涉足应用层时，便与依赖其原材料的“餐厅”（应用开发者）产生竞争。在AI时代，拥有大模型无法获取的专有数据源至关重要。FlightAware通过部署天线收集公开的飞机ADS-B数据，并将其转化为有价值的产品。PitchBook（融资数据）、LexisNexis（法律数据）、CoStar（房地产数据）、Bloomberg（金融数据）等公司，通过聚合和数字化公开信息，构建了数据壁垒。Ancestry.com甚至通过购买家谱记录来强化其护城河。关键在于，这些数据是ChatGPT或Anthropic等模型无法直接获得的。AI极大地放大了数据的价值，使得原本的原始数据能够创造十倍甚至百倍的价值，因为现在可以提供高价值的成品而非原材料。OpenEvidence独家授权顶级医学期刊内容，为循证医疗提供权威数据；VLex通过聚合西班牙法律记录并结合AI，实现了收入五倍增长，提供高价值的法律成品；AskLeo利用企业采购部门的专有合同数据，为企业提供比通用AI更优的采购决策支持。
Original English
The third core theme is the **walled garden**, which refers to moats built upon **proprietary data**. Large model companies like OpenAI are like "vegetable farms," providing foundational compute power, but when they venture into the application layer, they compete with "restaurants" (application developers) that rely on their raw materials. In the AI era, having proprietary data sources inaccessible to large models is crucial. **FlightAware**, by deploying antennas to collect public aircraft ADS-B data, transforms it into a valuable product. Companies like **PitchBook** (financing data), **LexisNexis** (legal data), **CoStar** (real estate data), and **Bloomberg** (financial data) build data moats by aggregating and digitizing public information. **Ancestry.com** even strengthens its moat by purchasing genealogical records. The key is that this data is not directly accessible to models like ChatGPT or Anthropic. AI significantly amplifies the value of data, enabling raw data to create ten or even a hundred times more value, as high-value finished products can now be offered instead of raw materials. **OpenEvidence**, by exclusively licensing content from top medical journals, provides authoritative data for evidence-based medicine. **VLex**, by aggregating Spanish legal records and integrating AI, achieved a fivefold revenue increase, offering high-value legal deliverables. **AskLeo**, leveraging proprietary contract data from corporate procurement departments, provides superior procurement decision support for enterprises compared to general AI.应用层崛起与消费者AI新机遇
关于大模型公司是否会吞并整个应用层，答案是不会。模型聚合器在许多类别中比单一模型更有优势，如同Kayak聚合多家航空公司信息。AI应用需要访问多种模型，因为它们各有专长且不可完全替代。大模型公司受限于自身模型和承诺（如谷歌承诺不进一步中介化互联网），且面临复杂的优先级问题（如OpenAI同时竞争消费、企业、模型、硬件领域）。因此，应用层不会被模型层吞并，而是形成繁荣的创业生态系统。消费者AI的发展模式与企业AI一致，也存在三个核心机会：传统类别AI原生化（如Krea取代Photoshop作为早期设计师的首选）、品类创造（如Eleven Labs在语音模型领域开辟新市场），以及专有数据优势（如Slingshot利用AI抄写员收集的真实治疗场景数据，构建差异化产品）。AI应用将越来越多地转向增强人类能力、创造新收入来源，而非简单替代。
Original English
Regarding whether large model companies will absorb the entire application layer, the answer is no. **Model aggregators** offer advantages over single models in many categories, akin to Kayak aggregating information from multiple airlines. AI applications need access to various models, as each has its specialization and they are not fully interchangeable. Large model companies are constrained by their own models and commitments (e.g., Google's promise not to further disintermediate the internet) and face complex prioritization issues (e.g., OpenAI competing simultaneously in consumer, enterprise, model, and hardware sectors). Therefore, the application layer will not be absorbed by the model layer but will foster a thriving startup ecosystem. Consumer AI follows a similar development pattern to enterprise AI, presenting three core opportunities: **AI-native transformation of traditional categories** (e.g., Krea becoming the preferred tool for early-career designers over Photoshop), **category creation** (e.g., Eleven Labs carving out a new market in voice models), and **proprietary data advantages** (e.g., Slingshot leveraging real therapy session data collected by AI scribes to build a differentiated product). AI applications will increasingly focus on augmenting human capabilities and creating new revenue streams, rather than simple replacement.未来展望：速度、价值与社会影响
AI驱动的产品周期速度前所未有。过去公司从主导地位到被淘汰通常需要五年，但如今软件构建速度极快，缺乏护城河的产品易被复制。价值捕获的逻辑也发生根本性改变：传统软件按功能和用户数定价，而AI替代劳动力后，定价转向与人力成本竞争，创造了更大的价值池。成功的AI应用将更多地增强人类能力并创造新收入。关于社会影响，技术进步（如拖拉机取代农民）重塑了就业市场，创造了新机会。大多数AI应用旨在扩展服务边界，让不可能变为可能，而非简单消灭工作。AI可以在凌晨两点提供服务，这是人类难以做到的。虽然部分职业可能被取代，但技术进步总是在重塑就业。正如“没有人会告诉你你正活在美好的旧时光里，直到这些时光已经过去”，AI驱动的产品周期更加软件主导且去中心化，为技术人员带来了前所未有的有趣和创造性的时代。AI应用的黄金时代已至，它不仅重塑软件行业，更深刻改变我们的工作与生活。