English

2024 matters because artificial intelligence begins to move from isolated interaction toward delegated process.

Earlier phases of generative AI centered on visible outputs: write this, summarize that, generate an image, answer a question, produce a draft. Even when powerful, these systems often still looked like responsive surfaces. In 2024, another shift becomes historically significant. AI begins to be framed less as a single-response engine and more as something that can execute multi-step work through tools, memory, context, and iterative action.

This matters because workflow itself starts to change. The user is no longer only requesting content. The user increasingly sets goals, provides constraints, and lets a machine carry out partial sequences: search, compare, transform, route, call tools, keep state, revise, and continue. The machine begins to look less like a talking interface and more like a junior operator inside a process.

The significance of this year lies in delegation. Systems are not yet perfectly reliable, but reliability is no longer the only question. The deeper historical question becomes: how much structured work can be handed to machine intermediaries before direct human execution even begins?

In China, as elsewhere, this matters because once AI becomes workflow-capable, it enters a much larger set of domains: customer support, operations, research assistance, coding, office automation, process management, and internal coordination. The economic meaning of AI expands beyond expression into execution.

Historically, 2024 is important because it marks the beginning of an agentic turn. AI no longer only produces outputs on demand. It increasingly participates in ongoing chains of action.

One-sentence summary:

The key to 2024 is that AI begins moving from prompt-response interaction into delegated workflow, making agents a new form of machine participation in work.

---

中文

2024 年的重要性，在于人工智能开始从孤立交互，转向被委托的流程执行。

生成式 AI 的前几个阶段，重点主要放在“可见输出”上：写一段、总结一下、生成一张图、回答一个问题、起草一份文本。即使这些能力已经很强，它们在很多时候看上去仍然更像一个会回应的表面。到了 2024 年，另一个变化开始变得具有历史意义：AI 不再只是被理解成一个“单次响应引擎”，而开始越来越多地通过工具调用、上下文保持、状态记忆和多步动作，被放进真实工作流。

这件事重要，因为改变的不只是输出，而是流程本身。用户不再只是索取内容，而开始设定目标、提供约束，然后把一段局部序列交给机器去执行：搜索、比较、转换、分流、调用工具、保持状态、修订、继续推进。机器开始越来越像流程里的一个初级执行者，而不只是一个会说话的界面。

2024 年的意义，在于“委托”开始真正发生。系统当然还不完美，可靠性问题依然很多，但此时历史上更深的那个问题已经出现了：在人类直接动手之前，有多少结构化工作可以先交给机器中介去跑一轮？

在中国，这一点同样关键，因为一旦 AI 具备工作流能力，它就不再只属于内容生成，而开始进入更大范围的真实场景：客服、运营、研究辅助、编程、办公室自动化、流程管理、内部协调。AI 的经济意义，也因此从“表达能力”扩展到“执行能力”。

从历史上看，2024 年的重要性，在于它标记了一个“agent 转向”的开始。AI 不再只是按需吐出答案，而开始参与持续的行动链条。

如果说 2023 年写的是大模型如何成为基础设施，那么 2024 年写的就是：这层基础设施开始长出手脚。模型不再只是能力池，而开始通过代理、工具和状态管理进入工作过程本身。

一句话概括：

2024 年的关键，是 AI 开始从提示词—响应式交互，转向被委托的工作流执行，agent 因此成为机器参与工作的一个新形式。