附录：行业 AI Agent 市场分析 Appendix: Market Analysis of Industry AI Agents

从功能附加走向行业执行系统 From Feature Add-On to Industry Execution System

如果把 2026 年的市场状态说得直接一点，那么行业 AI agent 这件事已经不再停留在概念验证阶段，但也还远没有进入大规模稳定成熟期。更准确的判断是，市场已经确认了方向，产品形态正在迅速分化，而真正能够长期成立的，不会是会聊天的行业助手，而是能够进入真实工作流、承担一部分执行责任的行业系统。 Put directly, the 2026 market is no longer at the proof-of-concept stage, but it has not yet reached stable large-scale maturity. The more accurate reading is that the direction is now clear and product forms are diverging quickly. What can last is not a chatty industry assistant, but an industry system that enters real workflows and takes on part of the execution responsibility.

今天真正有价值的，不再是让模型回答行业问题，而是让它在行业语境下理解对象、遵守规则、推动流程、保留记录，并在必要时把工作交还给人。也正因为如此，行业 AI agent 的核心竞争，不再只是模型能力本身，而是模型与真实工作流之间的结合深度。 What matters now is not having a model answer industry questions, but making it understand objects in industry context, follow rules, move processes forward, keep records, and hand work back to people when needed. That is why the core competition is no longer only model capability, but the depth of integration between the model and the real workflow.

为什么垂直软件会先长出 Agent Why Vertical Software Grows Agents First

从供给侧看，行业 AI agent 之所以会先在垂直软件里长出来，并不难理解。垂直软件天然掌握领域工作流、历史数据和日常操作入口，也就是同时拥有场景、上下文和分发能力。换句话说，行业 AI agent 不是先有模型再去寻找行业，而是先有行业流程和工作结构，再把模型压进去。 From the supply side, it is easy to see why industry AI agents appear first inside vertical software. Vertical software naturally owns the domain workflow, historical data, and daily interaction entry points, which means it already has the scenario, the context, and the distribution channel. In other words, the industry AI agent does not start with a model looking for an industry; it starts with the industry process and work structure, and then the model is inserted into it.

最先占优势的玩家是谁 Who Has the Early Advantage

所以现在市场上真正占优势的第一类玩家，并不是新冒出来的纯 AI 创业公司，而是原来就占据行业入口的软件平台。法律、会计、地产、保险这些行业中，最先成型的 AI 能力，大多都是先嵌入原有行业软件，再逐步向更强的执行能力推进。这些平台的优势在于分发快、客户教育成本低、历史数据现成、切入日常流程自然。 The first real winners are not the new pure-AI startups, but the software platforms that already sit at the industry entry point. In law, accounting, real estate, and insurance, the earliest AI capabilities are usually embedded into existing vertical software and then gradually extended toward stronger execution. Their advantage is fast distribution, low customer education cost, existing historical data, and a natural fit with daily workflows.

三种产品形态 Three Product Forms

如果进一步区分，可以把当前市场上的所谓行业 AI 产品大致分成三类。第一类是嵌入式 AI 功能。它们依附于原有行业软件存在，主要承担问答、草拟、摘要、数据提取或局部流程辅助。第二类是工作流型 agent。它们开始围绕一个相对完整的业务流程承担更多执行责任，例如线索跟进、文档整理、日程推进、状态更新、任务分流、审核前准备等。第三类才是真正接近独立行业执行系统的产品。 If we distinguish further, current industry AI products can roughly be grouped into three types. The first is embedded AI functionality: it lives inside the existing vertical software and mainly handles Q&A, drafting, summaries, extraction, or partial workflow help. The second is workflow-based agents: they begin to take on responsibility for a more complete business process such as lead follow-up, document preparation, schedule progress, status updates, task routing, and prep before review. The third is the product type closest to an independent industry execution system.

这类产品不只是原系统中的一个 AI 按钮，而是围绕行业工作对象、默认角色、默认流程和默认界面形成一套独立成立的工作空间。到了这一层，AI 不再是外挂功能，而是产品运行逻辑的一部分。 At this level, the product is not just an AI button inside the old system, but a separate workspace built around industry objects, default roles, default processes, and a default interface. Here AI is no longer an add-on feature; it is part of the product's operating logic.

行业分布并不平均 Industry Adoption Is Not Evenly Distributed

从行业分布来看，这三类形态并不是均匀分布的。法律和地产目前最容易出现更完整的独立行业 agent，原因在于这两个方向同时具备几个条件：数字入口足够多，重复流程足够强，结果相对容易核验，单次工作价值也足够高。相比之下，会计和保险虽然同样适合 agent 化，但更深地绑定既有系统、表单规范和责任链条，因此目前更常见的仍然是嵌入式或半独立形态，而不是完全独立的新产品体系。 By industry, these three forms are not distributed evenly. Law and real estate are currently the easiest places for a more complete independent industry agent to appear because they share several traits: many digital entry points, strong repetition, outcomes that are relatively easy to verify, and high value per task. Accounting and insurance are also suitable for agentification, but because they are more deeply tied to existing systems, form rules, and responsibility chains, embedded or semi-independent forms are still more common than fully independent new product systems.

市场是否适合独立 Agent，取决于行业结构 Whether the Market Fits an Independent Agent Depends on Industry Structure

这说明一个关键事实：行业 AI agent 的市场并不是平均展开的，而是高度依赖行业结构本身。一个行业是否适合出现独立 AI agent，通常取决于四个因素。第一，是否存在高频、强重复、可结构化的流程。第二，是否有足够多的数字工作入口，例如网页后台、文档、邮件、消息、表单。第三，工作结果是否相对容易验证，而不是完全依赖不可解释的专业判断。第四，客户是否愿意为节省机械劳动和提高流程推进效率而付费。 This points to a key fact: the market for industry AI agents does not unfold evenly. It depends heavily on industry structure itself. Whether an industry is suitable for an independent AI agent usually comes down to four factors. First, are there high-frequency, highly repetitive, structured workflows. Second, are there enough digital work entry points such as web back offices, documents, email, messages, or forms. Third, are the outcomes relatively easy to verify rather than fully dependent on opaque expert judgment. Fourth, are customers willing to pay for reduced manual labor and improved process throughput.

真正有市场空间的行业 AI agent，通常不是从最宏大的全行业全能平台切入，而是从某个足够窄、但足够高频且价值清楚的工作闭环进入。先在一个局部流程里承担真实责任，逐步扩大覆盖范围，这比一开始就声称重构整个行业，更符合当前市场的接受方式和产品现实。 The industry AI agents with the real market space usually do not start from a giant all-purpose platform. They begin with a narrow but frequent and clearly valuable workflow loop. Taking on real responsibility in one local process first, then expanding coverage gradually, fits market acceptance and product reality far better than claiming to rebuild an entire industry from the start.

定价逻辑正在变化 Pricing Logic Is Changing

从商业模式看，行业 AI agent 也正在逐步脱离普通聊天工具的定价逻辑。客户愿意为它付费，不是因为它能说得更像人，而是因为它能替代一部分原本需要人重复完成的机械劳动，或者提高一部分原本依赖人工维持的流程效率。因此，定价真正对应的不是 token 成本，而是被系统接管的工作量、责任密度和业务价值。 From a business model perspective, industry AI agents are also moving away from the pricing logic of normal chat tools. Customers are willing to pay not because the system sounds more human, but because it can replace repetitive manual labor or improve the efficiency of workflows that previously depended on humans. So the real pricing basis is not token cost, but the amount of work, responsibility density, and business value the system takes over.

这意味着行业 AI agent 的价格通常会高于通用 AI 助手，也会更接近行业软件、专业工具或工作系统的价格区间。行业越专业、流程责任越高、错误成本越大，客户能够接受的价格也通常越高。相反，如果一个产品只是把通用大模型加上一些行业术语包装，它很难长期获得行业软件级别的定价能力。 That means industry AI agents are usually priced above general-purpose AI assistants and closer to vertical software, professional tools, or work-system pricing. The more specialized the industry, the higher the workflow responsibility, and the larger the cost of mistakes, the higher the acceptable price. By contrast, if a product only wraps a general model in some industry jargon, it is hard to sustain industry-software-level pricing.

SmallClaw 的位置 Where SmallClaw Fits

因此，对整个市场可以给出一个比较明确的判断。行业 AI agent 已经从概念走向现实，但市场尚未定型；嵌入式 AI 仍然会在相当长一段时间内占据主流，但独立行业执行系统会逐步出现，并成为下一阶段更有壁垒的方向；真正有前景的产品，不是把 AI 放进行业软件，而是让 AI 成为行业工作系统的一部分。 So we can make a fairly clear judgment about the market. Industry AI agents have moved from concept into reality, but the market is not yet fully formed. Embedded AI will remain dominant for a long time, but independent industry execution systems will gradually appear and become the more defensible direction in the next stage. The most promising products are not the ones that simply put AI into industry software, but the ones that make AI part of the industry work system.

在这个判断之下，SmallClaw 的定位是清晰而成立的。SmallClaw 的价值，不在于把通用大模型换一个行业名字重新包装，而在于把共享内核、组织化运行结构与行业壳层结合起来，形成可以在真实工作流中承担执行责任的行业系统。它面对的不是通用问答市场，而是行业工作系统正在被重新定义的这一轮结构性机会。 Under this judgment, SmallClaw's positioning is clear and valid. Its value is not in repackaging a general model under an industry label, but in combining a shared core, an organized runtime structure, and an industry shell into a system that can take execution responsibility inside real workflows. It is not aimed at the general Q&A market; it is aimed at the structural opportunity created as industry work systems are being redefined.

也正因为如此，SmallClaw 的价值不只是技术层面的 agent 能力，更在于它具备走向行业执行系统的产品基础。它不是原有行业软件里附带的一块 AI 功能，而是一种更接近下一代行业工作平台的形态。这种定位本身，就是 SmallClaw 在未来行业 AI agent 市场中的核心意义。 That is also why SmallClaw's value is not just its technical agent capability, but the product foundation to become an industry execution system. It is not a sidecar AI feature inside legacy software; it is a form closer to the next-generation industry work platform. That positioning itself is the core meaning of SmallClaw in the future industry AI agent market.