Agentic Workflow Patterns

Most "agents" in production aren't agents — they're workflows. Anthropic's December 2024 post "Building Effective Agents" drew the line: workflows have predetermined structure, agents decide structure at runtime. Workflows are cheaper, more debuggable, and more reliable. Reach for an autonomous loop only when the structure genuinely can't be known in advance.

Learning Objectives

After this lesson, you will be able to:

Walk through Anthropic's 5 canonical workflow patterns — prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer — and pick the right one per task
Distinguish 'workflows' (deterministic LLM compositions) from 'agents' (autonomous loops) — and prefer workflows whenever they suffice
Implement an orchestrator-workers pattern in 50 lines that splits a task across parallel sub-agents and aggregates their outputs
Recognize when to escalate from workflow patterns to autonomous agents — and the cost of doing it prematurely

Prerequisites

Before You Start

Review

Why

Why This Matters

Why Anthropic's "Building Effective Agents" became required reading in 2025 -- their Dec 2024 blog post defined a vocabulary the entire industry now uses; "workflow vs agent" + the 5 patterns are the new standard taxonomy
Why most production agentic systems use workflows, not autonomous agents -- workflows are simpler, more predictable, easier to debug; only escalate to fully autonomous agents when the task structure genuinely demands it
Why "agent" became meaningless without these patterns -- before this taxonomy, "agent" meant any LLM-with-tools; now teams can speak precisely about prompt-chaining vs orchestrator-workers vs evaluator-optimizer
Build this --> Implement all 5 patterns on the same task (e.g., "write a marketing email for product X"); compare quality, cost, latency; understand the tradeoffs viscerally

Task pattern	Recommended workflow
Linear pipeline (research → draft → translate → format)	Prompt chaining
Distinct categories needing specialized handling	Routing
Independent subtasks that can run in parallel	Parallelization
Subtasks discovered at runtime	Orchestrator-workers
Iterative quality improvement with clear criteria	Evaluator-optimizer
None of the above	Autonomous agent (escalate carefully)

Agentic Workflow Patterns

After this lesson, you will be able to:

Before You Start

#Workflows vs Agents

Your Reflection

Continue Learning

Discussion

#The Agent-vs-Workflow Decision Tree

#The Five Patterns

#1. Prompt Chaining

#2. Routing

#3. Parallelization

Two flavors

#4. Orchestrator-Workers

#5. Evaluator-Optimizer

#Implementation: Orchestrator-Workers

#Decision Rubric

#Key Takeaways

#Quick Check

Deep Dive