这期 A16Z 播客探讨了 AI 将“软件转化为劳动力”的变革性潜力，从根本上重塑了企业软件格局，并开启了超越传统软件预算的巨大新市场。在三位 A16Z 普通合伙人 Alex Rampel、Angela Strange 和 David Haber 的带领下，讨论追溯了软件在各个时代的演变历程，以及 AI 如何代表着一种范式转变。 Rampel 概述了软件的三个不同时代。第一个时代始于 20 世纪 60 年代，涉及将物理文件柜数字化，例如人力资源记录、机票和财务报表。虽然这提高了信息存储的效率和自动化程度，但并没有从根本上改变劳动力规模。第二个时代始于 1998 年左右，以 Salesforce、QuickBooks (NetSuite) 和 Zendesk 为例，看到了向基于云的软件的转变。这减轻了对本地 IT 基础设施的负担，但人力资源和客户支持等部门的核心职能和人员数量基本上没有变化。第三个时代，由金融服务推动，见证了支付处理和其他金融服务的捆绑，从而在餐饮和暖通空调等行业创造了可行的软件市场。 AI 的出现代表着第四个革命性时代。关键区别在于 AI 能够对存储在这些“基于云的文件柜”中的数据执行操作，从而有效地增强，甚至在某些情况下取代人力劳动。现在，软件代理可以接管人类已经做了几十年的任务。Rampel 用护士的例子来说明这一点，这是一个庞大的劳动力市场，但软件市场相对较小。由 AI 驱动的软件可以自动化诸如以多种语言发送结肠镜检查前提醒等任务，从而释放护士从事更关键的任务。同样，Haber 指出了 Mindbody，它运行健身工作室软件。所有这些都可以开始通过 AI 来完成。 Strange 在此基础上进一步强调了 AI 有潜力使软件公司的收入增长 10 倍。她强调，由于 Toast 收入的 80% 来自支付和保险，因此问题是，AI 使用的增加是否会使这些公司更加盈利？要使之奏效，客户必须看到他们的“软件预算开始融入到劳动力预算中”。Rampel 警告 Zendesk 和 Salesforce 等现有软件公司，AI 自动化任务的潜力可能会导致客户减少对“席位”的需求，从而可能摧毁他们现有的收入模式。与按席位定价不同，新产品可以更有效率。 为了进入记录系统，例如 Salesforce 和 Zendesk 开发的系统，专家小组提倡专注于垂直 SaaS，并从“混乱的收件箱”开始构建。这些解决方案“比人类操作好一千倍”。目前看来像魔法一样，但商品化是可能的，而且这只是“超级差异化”。防御力来自于拥有下游工作流程，深度集成到其他系统，并有效地拥有端到端的核心工作流程。最重要的是，护城河仍然重要。 专家们还讨论了一些现有的软件产品已经过时，以及一些领域，如合规官使用的领域，非常糟糕。此外，还有一些领域是新兴且增长迅速的，如美甲师，这意味着没有现有企业。这些领域拥有巨大的劳动力预算，是开发垂直 SaaS 的绝佳市场。小组以对未来可能发生什么事情的讨论结束，并表示看起来晦涩的东西是好的。他们强调这项技术尚未完全成熟，但他们也知道它可以随时到来并颠覆几乎任何事物。

This A16Z podcast episode explores the transformative potential of AI to turn "software into labor," fundamentally reshaping the landscape of enterprise software and unlocking vast new markets beyond traditional software budgets. The discussion, led by three A16Z General Partners, Alex Rampel, Angela Strange, and David Haber, traces the evolution of software through various eras and how AI represents a paradigm shift. Rampel outlines three distinct eras of software. The first era, starting in the 1960s, involved digitizing physical filing cabinets, such as HR records, airline tickets, and financial statements. While this increased efficiency and automation of information storage, it didn't fundamentally alter the size of the labor force involved. The second era, beginning around 1998, saw the shift to cloud-based software, exemplified by Salesforce, QuickBooks (NetSuite), and Zendesk. This reduced the burden of on-premise IT infrastructure, but the core functions and number of personnel in departments like HR and customer support remained largely unchanged. The third era, fueled by financial services, saw the bundling of payment processing and other financial services, creating viable software markets in industries like restaurants and HVAC. The emergence of AI represents a fourth, revolutionary era. The key difference is AI's ability to perform actions on the data stored in these "cloud-based filing cabinets," effectively augmenting and, in some cases, replacing human labor. Now software agents can take over tasks that humans have been doing for decades. Rampel illustrates this with the example of nurses, a massive labor market with a relatively small software market. AI-powered software can automate tasks like pre-colonoscopy reminders in multiple languages, freeing up nurses for more critical tasks. Similarly, Haber points to Mindbody, which runs fitness studio software. All that can start to be done with AI. Strange builds upon this by emphasizing the potential for AI to 10x revenue for software companies. It emphasizes that because 80% of Toast revenue is from payments and insurance, the question is will increased AI usage make these companies even more profitable? For this to work, the customer must see their "software budget bleed into their labor budget." Rampel cautions incumbent software companies like Zendesk and Salesforce that AI's potential to automate tasks could lead to customers needing fewer "seats," potentially decimating their existing revenue model. Instead of per seat pricing, new products can be more productive. To enter a system of record, such as those that Salesforce and Zendesk have developed, the panel promotes focusing on vertical SaaS, and building up from "messy inboxes". These solutions are "a thousand times better than humans doing." For now it seems like magic, but commoditization is possible, and this is only "super differentiated." Defensibility comes with owning downstream workflows, deeply integrating to other systems, and effectively owning that end to end core workflow. The main point is that moats still matter. The panelists also discussed how some incumbent software products are just outdated, and that some fields, like those used by compliance officers, are terrible. Also some fields are new and growing fast, like manicurists, meaning that no incumbents exist. These fields have giant labor budgets and would be great markets to develop a vertical SaaS for. The panel closed with a discussion on what might come of this, and stated that what seems obscure is good. They emphasize that the technology is not quite ready, but they also know that it can come and uproot almost anything at any time.