Recognising When a Request Needs Decomposition (CCAO-F)

Some requests are too large for one instruction

A prompt can be perfectly specified against the five components and still underperform, because the problem is not how the request is worded but how much it is trying to do at once. The Claude Certified Associate - Foundations (CCAO-F) exam introduces task decomposition here as an understand-level skill: recognising when a single request is actually several distinct tasks wearing one sentence, and would be done better as a sequence.

The classic example is a vendor evaluation. "Evaluate these three vendors and tell me which to pick" reads like one request, but it hides four: derive the criteria that matter, score each vendor against them, weigh the trade-offs, and recommend one. Asked all at once, Claude has to do all four in a single pass, so it does each of them shallowly and you never see the reasoning behind the recommendation. The wording is fine; the problem is the bundling.

How to spot a bundled request

The tell of a request that needs decomposition is that it contains several distinct actions, often joined by "and" or implied by the goal. "Analyse this data and recommend a strategy" is two actions. "Read these documents, extract the key changes, and write three announcements" is at least three. When you can list more than one verb-driven task inside a single request, you are probably looking at a sequence collapsed into one sentence.

The reason bundling hurts is mechanical. In a single pass, Claude cannot pause between stages, so it produces one flat output that touches each stage lightly rather than a considered result for each. You lose depth on every stage, and you lose the ability to inspect the intermediate work - the criteria, the scores, the trade-offs - because none of it is surfaced separately. Decomposition fixes both by turning the one pass into an ordered sequence, where each step gets its own focused attention and produces a result you can see.

Decomposition has a cost too

The skill is not "always decompose." Splitting a request that is genuinely simple adds steps, overhead, and back-and-forth for no benefit. A single, well-specified question that Claude can answer well in one pass should stay a single prompt; wrapping it in a five-step ceremony wastes effort and can even fragment a coherent answer.

So recognition cuts both ways. The judgement is whether the request truly bundles multiple distinct sub-tasks that each deserve their own focus, or whether it is one task that merely looks involved. Multi-stage work - evaluate-and-recommend, extract-then-draft-several - benefits from decomposition. A rich but single-action request does not. The exam rewards candidates who can tell the difference rather than reaching for decomposition or the single prompt by reflex.

What the CCAO-F exam trips candidates on

Two traps recur, one in each direction.

The first is assuming a single well-worded prompt can substitute for decomposition on genuinely multi-stage work. A scenario shows shallow output from a bundled request and offers "reword the prompt more carefully" as the fix. The credited reading is that no amount of wording collapses four distinct tasks into a good one-pass answer - the structural fix is to split them into a sequence.

The second is the opposite: decomposing a request that is actually simple, adding unnecessary steps and overhead. A scenario takes a single-action task and breaks it into an elaborate multi-step flow. The reliable reading is that decomposition is for genuinely multi-stage work; imposing it on a simple request wastes effort. Recognising which case you are in is the whole skill.

Common misreadings to avoid

How this shows up on the exam

Domain 1 questions on this knowledge point show either shallow output from a bundled request or an over-engineered breakdown of a simple one, and ask for the correct judgement. The reliable frame is to count the distinct sub-tasks: several genuinely different actions call for decomposition, while a single action wrapped in complexity does not.

This knowledge point opens task statement 1.2 and leads directly into ordering steps for checkable intermediate results, which explains how to sequence the steps once you have decided to decompose. It also anticipates decomposition failure modes, where getting the recognition or the ordering wrong produces shallow or contradictory results.