AI Foundationsintermediate12 min · 10 steps
Agent Workflows I: Prompt Chaining and Routing
Anthropic's first two workflow patterns: split a task into checked sequential calls, or classify the input and send it to the right specialist.
LLM calls1
quality60%
requesterrordataresponsetool call
metrics are illustrativeProblemStep 1 of 10
One prompt that does everything does nothing well
A team asks one LLM call to read the brief, outline the message, write the copy, check brand rules and translate it into German. Each instruction competes for the model's attention, so some get skipped and nobody can tell which part failed.
The same happens with a support bot that handles FAQs, refunds and outages in one prompt. Tuning it for one kind of request makes another worse.
Watch outIf you can't point to the step that went wrong, you can't fix it.
← Previous lessonThe Augmented LLM: Retrieval, Tools and MemoryNext lesson →Agent Workflows II: Parallelize, Delegate, Refine
Lesson 29 of 34 in AI Foundations