Sequential Chaining vs. Structured Single-Prompt Decomposition

Why Decomposition Matters

Complex requests fail when they ask for too much at once. Decomposition breaks a goal into ordered, checkable steps. This mirrors the *Delegation* and *Description* ideas in Anthropic's AI Fluency Framework: describe the work clearly and hand off well-scoped pieces rather than one sprawling ask.

Two Practical Patterns

• Sequential (prompt chaining). Do step 1, review it, then feed it into step 2. Example: (1) extract the key complaints from 40 survey responses, then (2) group them into themes, then (3) draft an action memo from the themes. Reviewing between steps catches errors before they compound into the final output.
• Structured single prompt. For a moderately complex task, one prompt can still spell out sub-steps: "First summarize the contract. Then list the three riskiest clauses. Finally recommend one change per clause." This works when the task doesn't need a human check between sub-steps.

Verification Benefit

Decomposition also makes verification easier: you can check each intermediate result (the extracted complaints, the themes) instead of trying to judge one giant, undifferentiated output at the end. If a downstream step goes wrong, it's traceable to a specific, reviewable intermediate step.

Common exam traps

• Trying to solve a multi-part problem in a single vague prompt when the pieces should be sequenced and checked -- e.g., asking "what should we do?" over 50 raw interview notes instead of extracting points, grouping themes, and drafting recommendations in order.
• Confusing decomposition with just "making the prompt longer." Decomposition is about *structure and order* -- ordered, checkable sub-steps -- not about word count. A long but unstructured prompt is not decomposed.