CCAO-F questions and answers — Domain 4: Workflow Integration and Solution Design

10 practice questions on Workflow Integration and Solution Design, the domain weighted at 16% of the CCAO-F exam. Every answer explains why the distractors fail — that is the part that transfers to questions you have not seen before.

Question 1

Which task is the best candidate for a Claude-assisted workflow?

1. Approving expense claims against company policy
2. Producing first drafts of weekly customer-facing release notes from engineering changelogs
3. Deciding which of two suppliers to terminate
4. Signing off the year-end accounts

Answer: B. It is language-heavy, recurring, works from material you can supply, and tolerates a draft-then-review cycle. The other three are decisions or approvals that a person must own, and where Claude's role can only ever be preparation.

Question 2

You are designing a workflow where Claude drafts responses to customer complaints. Which element is most often left implicit and therefore skipped?

1. The trigger that starts the task
2. The review step: who checks what, and to what standard
3. The format of the output
4. Which model to use

Answer: B. Review is the step that gets designed as "someone will look at it" and then quietly stops happening once the novelty wears off. Naming the reviewer and the standard is what makes it survive contact with a busy week.

Question 3

A manager asks whether Claude can "handle" the team's monthly reporting. What is the best answer?

1. Yes — it will save the team about 70% of the time they currently spend
2. It can produce the first draft from the source data, which typically cuts drafting time substantially; someone still needs to verify the figures and approve the final version
3. No — reporting is too important to automate
4. Yes — and once it is set up nobody will need to review the output

Answer: B. Specific about the value, explicit about the limit, honest about who is still accountable. The invented percentage oversells, the flat refusal underuses a good fit, and removing review is both wrong and the fastest route to an embarrassing report.

Question 4

Your team already has a documented process for producing case studies. Where should Claude be introduced?

1. Replace the process with a new AI-first workflow
2. Insert it at the drafting step, leaving the rest of the process intact
3. Introduce it across every step at once so the benefit is maximised
4. Run it in parallel as a separate optional process

Answer: B. Integration beats replacement. A process people already follow, with the drafting step accelerated, gets adopted; a new parallel or wholesale-replacement process asks everyone to change how they work and usually gets abandoned regardless of merit.

Question 5

You want to roll a Claude workflow out to a department of thirty. What is the best sequence?

1. Train everyone at once, then let each person find their own use
2. Prove it on one well-understood task, document the setup so it is reproducible, then expand
3. Give everyone access and see what emerges
4. Wait until a formal policy exists before starting

Answer: B. One proven task, documented so it can be repeated, gives you evidence and a template. Mass rollout without a proven case produces thirty inconsistent experiments, and indefinite waiting delivers nothing — though checking existing policy before you start is separately required.

Question 6

Which measure best supports a case for continuing an AI-assisted workflow?

1. Team members say it feels helpful
2. Time to first usable draft, and number of revision rounds, before and after
3. The number of prompts sent per week
4. How advanced the setup is

Answer: B. Measures tied to the task itself survive a budget conversation. Sentiment is easily dismissed, prompt volume measures activity rather than outcome, and sophistication of setup is not a benefit to anyone.

Question 7

A workflow depends on someone remembering to attach the right data file each time, and it keeps failing. What is the design problem?

1. The prompt is not detailed enough
2. The inputs step depends on human memory rather than being defined and reliable
3. The wrong model is being used
4. The review step is too heavy

Answer: B. Inputs are one of the five things a workflow has to define explicitly — what material, from where, supplied how. When that is left to memory it fails intermittently and looks like a tool problem. A Project holding the standing material, or a connector reaching the system directly, addresses it.

Question 8

Which is a legitimate limitation to communicate to stakeholders?

1. Claude cannot write in a professional tone
2. Claude cannot access systems it has not been connected to, and cannot guarantee factual accuracy
3. Claude cannot summarise long documents
4. Claude cannot produce structured output

Answer: B. Reach and accuracy guarantees are the two real limits that shape workflow design. The other three describe things Claude does well, and stating them as limitations misleads stakeholders in the opposite direction — which the exam treats as an error too.

Question 9

A colleague proposes using Claude to generate personalised outreach messages to a purchased list of contacts. What is the correct response?

1. Proceed, since the messages will be reviewed before sending
2. Raise it as a compliance question first — bulk unsolicited outreach and the provenance of the contact data are the issues, not the drafting
3. Proceed, but disclose that the messages were AI-assisted
4. Proceed if the volume is modest

Answer: B. The drafting is not the problem; the activity is. Bulk unsolicited outreach and purchased contact data raise policy and data-protection questions that review, disclosure and low volume do not resolve. Recognising which part of a proposal is the risk is a core Domain 4 and 6 skill.

Question 10

Which workflow is most likely to be abandoned within a month, even though it produces good output?

1. One that inserts a drafting step into an existing process
2. One that requires three people to review every low-stakes internal summary
3. One that uses a shared Project with documented instructions
4. One that starts with a single well-understood task

Answer: B. Disproportionate process is a real failure mode and an examinable wrong answer. Triple review of low-stakes internal output costs more than the task is worth, so people route around it — and then the checks that genuinely mattered get skipped too.

← Previous: Questions & answers — Domain 3  |  Next: Questions & answers — Domain 5 →