CCA 备考课 · 第 1.2 课CCA COURSE · LESSON 1.2

智能体与工作流：界线划在哪里Agents and Workflows: Where to Draw the Line

第一域 · 智能体架构与编排 · 编排层 · 第 3 关 · 编排 · 时长 11:21 · 英文授课Domain 1 · Agentic Architecture and Orchestration · Orchestration layer · Gate 3, Orchestration · 11:21 · Taught in English

上一课把循环本身拆透，这一课往上走一层做切分：先有这一刀，再谈任务分解（1.3）、拓扑连接（1.4）与护栏（1.6）。Lesson 1.1 fitted a running loop with signals and brakes; this lesson makes the decision that comes before the loop exists: which segments get a fixed path, and which earn a model in charge.

YouTube

小红书

数据源是固定的，格式是固定的，口径是领导定的，几年来没变过。团队要做的每周经营数据报表，还是被交给了全自主 Agent，理由是「显得智能」。结果每周一早上，报表格式先漂移：这周小计放前面，下周放后面；数字口径它还时不时自己「优化」一下。财务每次用之前都得先验一遍，比原来手工做还累。（开场为教学示意场景。）

本课内容

自主谱：从纯固定流程到全自主 Agent 的四个位置，为什么是给每个段落定位，而不是给整个系统选一个位置
下刀的三个属性：步骤能不能枚举、结果要不要可复现、出错的代价有多大、能不能撤回
切错的两个方向：把判断存成几百条脆规则的过度编排，和送出成本、延迟、不可复现、不可审计四笔账单的过度自主
一张客服系统的切分表：什么写死、什么放权，为什么退款可以让 Agent 提方案、执行必须收回固定流程
五个坑，从「全系统一刀切」开始

核心图

段落	定位	理由
身份核验、权限校验	固定流程	规则明确，错了代价高，必须可复现
理解问题、检索资料、生成答复	受约束的 Agent	情况千变万化，正是模型的主场
执行退款	方案放权，执行收回	Agent 提方案，执行走固定流程，加金额校验和人工审批

图：自主谱的四个位置，与一张客服系统的切分表。这一刀按段落切，不按系统切。

五个坑

全系统一刀切：要么全自主图省事，要么全写死图放心，不做段落切分。
拿「智能程度」当选型标准：觉得越自主越先进，这是最常见的误判来源。
放权不设界：段落交给了 Agent，却不给工具白名单、步数上限和预算上限。
确定性的活让模型干：金额、日期、校验这类必须复现的结果，交给概率输出。
规则写到几百条还不回头：明明是判断型任务，硬撑着决策树不放手。

分界不在技术，在任务：能枚举的写死，要判断的放权。

学完就用

这一课落在编排层 · 第 3 关 · 编排。把这一关放到你自己的系统上量一遍：免费 17 项快速自检，答完当场出报告，看你最弱的三关。

独立学习材料，与 Anthropic 无隶属关系。Claude 与 Claude Certified Architect 为 Anthropic 的商标；本课为独立备考课程，不授予、也不暗示任何认证。Independent study material, not affiliated with Anthropic. Claude and Claude Certified Architect are trademarks of Anthropic; this course is an independent preparation course and does not grant or imply certification.

The data sources never changed. The format never changed. The definitions had not changed in years. The team still handed the weekly report to a fully autonomous agent, because it looked smarter. By Monday the subtotals had moved, the definitions had been "optimized", and finance was checking every figure by hand. (Illustrative example.)

What this lesson covers

The autonomy spectrum: four positions, from a pure fixed workflow to a fully autonomous agent, and why you place segments on it, never whole systems
The three attributes that make the cut: can the steps be enumerated, must the result be reproducible, and what does an error cost, and can it be undone
The two ways to cut wrong: over-orchestration, where judgment is stored as hundreds of brittle rules, and over-autonomy, which sends four bills
A customer service system split on the table: what stays fixed, what is delegated, and why a refund is planned by the agent but executed by the workflow
The five traps, starting with one cut for the whole system

Core diagram

Segment	Placement	Why
Identity verification, permission checks	Fixed workflow	Rules explicit, error cost high, must be reproducible
Understand the question, retrieve material, generate the answer	Constrained agent	Situations vary endlessly; the model's home ground
Execute the refund	Plan delegated, execution taken back	Agent proposes; execution runs the fixed workflow, with an amount check and human approval

The autonomy spectrum, and the split on a table. The cut runs by segment, not by system.

The five traps

One cut for the whole system: either all autonomous, because it saves effort, or all hard-coded, because it feels safe, with no segment splitting at all.
Taking the level of intelligence as the selection standard: believing that more autonomous means more advanced. This is the most common source of a wrong call.
Delegating without setting boundaries: a segment handed to an agent, with no tool whitelist, no step limit, and no budget limit.
Letting the model do deterministic work: amounts, dates, checks, results that must be reproducible, handed over to a probabilistic output.
Rules grown into the hundreds, and still no turning back: a judgment task propped up on a decision tree that should have been let go long ago.

The boundary is not in the technology; it is in the task. What can be enumerated gets hard-coded; what needs judgment gets delegated.

Related cases

Put it to work

This lesson sits in the Orchestration layer, at Gate 3, Orchestration. Run that gate against your own system with the free 17-point self-check: it reports on the spot and names your three weakest gates.