CCAO-F Study Guide

Seven domains, 60 questions, 120 minutes, and a pass mark of 720 on a scale running from 100 to 1000. The weights are what should shape your revision, and they are uneven: Output Evaluation and Validation takes 21% of the paper, Workflow Integration and Solution Design 16%, Governance, Risk and Responsible Use 15%, Prompting and Task Execution 14%, Product and Model Selection 12%, Configuration and Knowledge Management 12%, and Troubleshooting and Optimization 10%. Read that list twice before planning anything, because the gap between the largest and the smallest domain is about seven questions.

CCAO-F is the associate credential for people who use Claude as a working tool rather than building on top of it. The official guide (version 1.0, effective July 2026) names operations, marketing, project management, education and communications as its audience, and there is no code anywhere on the paper: no API parameters, no agent architectures, no model internals. What you get instead is a long run of workplace situations in which something has gone slightly wrong, or is about to, and you have to name the real problem.

That makes it harder to revise than a technical exam, because there is no syntax table to memorise and be safe. What you can learn is the shape of each domain's questions and the family of wrong answers each one favours. This page walks the seven areas in weight order, says what each is genuinely testing, and closes with a revision sequence that spends your first and best hours on the largest block of marks.

How the 60 questions are distributed

Convert the percentages into question counts and the paper stops being abstract. At 21%, Output Evaluation and Validation is worth roughly 13 questions. Workflow Integration and Solution Design at 16% is around 10. Governance, Risk and Responsible Use at 15% is 9. Prompting and Task Execution at 14% is about 8. The two 12% areas, Product and Model Selection plus Configuration and Knowledge Management, are around 7 apiece. Troubleshooting and Optimization at 10% is 6.

Two things follow. The top three areas together account for 52% of the paper, so slightly over half your marks come from evaluation, implementation and governance. Someone strong across those three and merely adequate elsewhere is in decent shape. Someone fluent with the product surfaces but weak on evaluation is not, however many hours they have spent inside Claude.

The second point is that no one domain can carry you and no one domain can sink you outright. Dropping every question in Troubleshooting costs 10% of the paper. Dropping every question in the evaluation domain costs more than twice that and would be very hard to recover from. Weight your preparation accordingly, and resist the pull toward the domain you already enjoy.

Treat the percentages as the blueprint's declared distribution rather than a guarantee about your particular sitting. What is stable and useful is the ordering.

Output Evaluation and Validation, 21%

The largest domain, and arguably the reason the credential exists. This exam certifies people who put generated work in front of customers, boards, regulators and students, so the competence under test is knowing what to check before your name goes on it.

The questions are built so that nothing looks wrong. A figure is specific and well proportioned. A survey is named and plausible. A summary of thirty case files concludes that none of them involved an external contractor. Your job is to identify which element cannot be taken on trust and what would settle it. The correct answer almost always sends you to something that was not itself generated: the underlying report, the signed document, the published guidance, the person quoted.

Learn the four evasions the distractors keep offering. Asking Claude whether it is confident produces more text, not provenance. Regenerating and seeing the same figure twice produces a second draft from the same process. Softening a claim until nobody could falsify it hides the problem rather than resolving it. And assuming that uploaded source documents make an answer audited confuses grounding with verification.

Two subtler ideas earn marks here. Assertions about a whole set (nothing, all, the most common theme) are the hardest to check and the easiest to believe. And you are expected to triage by consequence, not to pretend you will verify a forty-page draft line by line.

Workflow Integration, 16%, and Governance, 15%

Together these are 31% of the paper, and they are the two areas where organisational experience beats product knowledge.

Workflow Integration asks what you do when a director watches a demonstration and wants a team moved across by month end. The reliable opening move is to map the existing process before changing any part of it, because the constraint is rarely the step you were asked to speed up. Pilots are chosen on volume and reversibility rather than visibility. The human checkpoint and the exception path are designed in from the start instead of bolted on. And a workflow producing excellent drafts while taking longer end to end has relocated effort into review rather than removed it, which you are expected to spot and to report at its true size.

Governance is not a values quiz. It hands you a specific task, ranking staff for redundancy selection, cutting an applicant pool before a panel sits, drafting an assessment that determines someone's care, and asks where the line falls. One distinction resolves most of it: whether the output informs a decision or becomes the decision. A comparable summary of forty applications against published criteria is defensible work. A ranked list adopted as the provisional outcome is not, because nobody can then explain any individual result. Around that sit personal data handling, honesty about who is speaking, and your organisation's own policy as the binding constraint.

Prompting and Task Execution, 14%

Around 8 questions, and the area candidates most often skip on the grounds that they write prompts every day. The confidence is misplaced, because the exam is not marking your phrasing. It is checking whether you can look at a piece of work and see what the request failed to supply.

The recurring setup is a fluent answer that is useless. Newsletter copy that would fit any organisation on earth. A summary weighting everything evenly when only three changes mattered. Nothing is incorrect and nothing is specific. The right option nearly always adds what only the human knew: the reader, the decision the work feeds, the source material, the length, and what a good version looks like.

Four habits come up repeatedly. Show a worked example rather than describing a standard in the abstract, because a sample of the voice you want beats three adjectives about it. Break a large deliverable into steps whose output can be checked before the next one runs. State what you do want instead of piling up things to avoid. And notice when a long conversation has drifted far enough that starting clean beats another round of corrections.

Distractors in this domain tend to offer a second lap of the same vague request, or a way to make Claude explain its reasoning instead of redoing the work with more to go on.

Product and Model Selection, 12%, and Configuration, 12%

Around 7 questions each, and the pair most amenable to deliberate study, because both reward precise boundaries rather than situational judgement.

Product and Model Selection punishes the instinct to reach for the heaviest option. The right choice is usually the lightest thing that does the job, and the items are written so the expensive alternative is plainly available and plainly wasteful. The distinction tested hardest is between web search, extended thinking and research: a quick factual lookup, a hard reasoning problem over material you already hold, and a multi-source briefing that comes back with citations. They answer different questions and are not ranked against one another. Model tier follows the same logic, matched to what the work demands rather than to what sounds most capable.

Configuration and Knowledge Management is three containers and the rule that sorts between them. A Project's instructions hold what is constant across all work in that Project. Its knowledge holds the documents the work draws on. The individual request holds whatever changes each time. Most items present something filed in the wrong place, or someone assuming a container reaches further than it does. Learn how far a Project's knowledge extends, why a superseded document left in place is worse than no document at all, whose permissions a connector uses, and what sharing a Project does and does not share.

Troubleshooting and Optimization, 10%

The smallest area at roughly 6 questions, and the one where real usage pays off most directly, because the items are diagnostic rather than declarative. You are handed a symptom and must name the cause before you can choose a remedy.

The structure that unlocks it is that a bad result has four possible origins. The request may not have said enough. The supporting material may be absent, out of date, or present twice in conflicting versions. The conversation may have drifted, carrying forward decisions abandoned an hour earlier. Or the expectation was never stated, in which case nothing failed at all and the output is simply not the thing that was wanted.

Distractors apply a sound fix to the wrong layer. Rewriting a prompt when a Project is answering from a superseded policy. Refreshing documents when nobody ever said what the summary was for. Repeating a correction more firmly when the durable answer is structural.

Two supporting points are worth committing to memory. Change one variable at a time, or you will not know which change did the work. And variation between two runs of the same request is ordinary rather than evidence of a fault, which matters because at least one option will invite you to treat it as a defect and escalate.

Revise this domain last. It costs very little once the other six are solid.

A revision order that respects the weighting

Begin with Output Evaluation and Validation. It is the biggest block of marks, its habits transfer directly into governance and troubleshooting scenarios, and it is where confident daily users of Claude most often score below their own expectations.

Take Workflow Integration second, then Governance. Those carry another 31% between them and are the areas where reading alone helps least. Work through situations and force yourself to say out loud where the constraint sits, or who remains accountable, before you look at any options.

Put Prompting and Task Execution fourth. Daily habits will get you part of the way, and the shortfall is usually about briefing rather than technique.

Do Product and Model Selection alongside Configuration and Knowledge Management, fifth, as one block. They share subject matter, they are the most memorisable content on the paper, and covering them together gives you one coherent picture of the product surfaces instead of two half-formed ones.

Finish with Troubleshooting and Optimization. Its four diagnostic layers map onto the request (Prompting), the material (Configuration), the conversation, and the choice of surface (Product and Model Selection), so by that point most of the thinking is already done.

If you have less than a week, invert the opening step: run a full timed sitting first and let your two weakest domains, weighted by their share of the paper, pick the order for you.

Common questions