# CCAO-F Practice Questions (Free, 2026) | Your AI Department > 60 free scenario-based CCAO-F practice questions for Claude Certified Associate – Foundations, with explanations, weighted by the published exam blueprint. Source: https://youraidept.com/network/ccao-f-practice-questions Practice questions # CCAO-F practice questions, free. 60 original scenario questions for Claude Certified Associate – Foundations, written from the published blueprint and weighted the way the exam is. Pick an answer, check it, read why. None of these are exam questions; the real exam tests the same judgment on scenarios you have not seen. Want the real thing? [Take the timed CCAO-F practice exam](https://youraidept.com/network/ccao-f-practice-exam): 60 questions in 120 minutes, scored against the 720 pass mark, with a domain breakdown. ## How to use these Filter by domain to drill a weak area, or work through the whole bank. Answer before you reveal. Note the domain of every miss. When you are done, the score box tells you whether you are above the equivalent of the 720 pass mark; if you are not, the domains you missed are your reading list. The full domain-by-domain notes are on [the CCAO-F exam page](https://youraidept.com/network/ccao-f). ## The questions All domains (60) Output evaluation and validation (13) Workflow integration and solution design (10) Governance, risk, and responsible use (9) Prompting and task execution (8) Product and model selection (7) Configuration and knowledge management (7) Troubleshooting and optimization (6) Output evaluation and validation Claude gives you a precise-sounding market statistic for a board deck. What do you do before using it? - Ask the same question in a fresh conversation and keep the figure if both answers agree - Ask Claude to restate the figure with a confidence rating and keep it if the rating is high - Label the slide as an AI-generated estimate so the board knows to treat the figure with caution - Ask for the source, then confirm the number against that original before it goes in the deck Check answer ### Show answer Answer: Ask for the source, then confirm the number against that original before it goes in the deck. Models can be wrong while sounding certain, so the figure has to be checked against a source you can open yourself. A confidence rating or a second run from the same model is still the model grading itself, and a caution label puts an unverified number in front of the board anyway. Output evaluation and validation You asked Claude to summarize a 30-page report. How do you check the summary is faithful? - Ask Claude to grade its own summary for accuracy and revise anything it rates below confident - Compare it with a summary from a different model and keep only the points the two share - Read the report's executive summary and confirm the two versions cover the same themes - Ask for the supporting passage behind each key claim and spot-check a few against the report Check answer ### Show answer Answer: Ask for the supporting passage behind each key claim and spot-check a few against the report. Grounding each claim in a quoted passage makes the summary checkable in minutes. Matching themes against the executive summary only shows the topics line up, not that the specific claims are accurate, and two models agreeing is not evidence either was faithful. Output evaluation and validation Which output is safe to send without a human review step? - A termination letter for an employee, since the template language is standard - A short reply telling a customer whether their refund is required under your policy - A press statement about a service outage, if the facts in it came from the incident log - A first draft of an internal meeting agenda that you will edit before circulating it Check answer ### Show answer Answer: A first draft of an internal meeting agenda that you will edit before circulating it. A low-stakes internal draft you will edit anyway can go straight to use. The press statement is tempting because the facts are sourced, but public statements carry consequences that need a named person to sign off, and the HR and customer-policy items carry legal weight. Workflow integration and solution design Where does Claude fit best in a weekly reporting process? - Pulling the figures from source systems and deciding which ones are worth reporting this week - Signing off on the final report, since it can check the numbers faster than a person can - Writing and sending the report directly to stakeholders once the numbers have been approved - Drafting the narrative from the approved numbers, with a person reviewing before it is sent Check answer ### Show answer Answer: Drafting the narrative from the approved numbers, with a person reviewing before it is sent. Drafting and formatting from approved inputs is the sweet spot, with a person reviewing before it goes out. Sending directly once numbers are approved skips the review of the narrative itself, and choosing what to report or signing off are judgment calls that stay with people. Workflow integration and solution design You connect Claude to your company's document store. What should you check first? - Whether the connector supports write access so Claude can file its outputs back into the store - Whether search results come back quickly enough for the team to use it in daily work - Whether the document store vendor has published guidance on AI assistant integrations - Which folders the connector can read and whether that matches the data's classification Check answer ### Show answer Answer: Which folders the connector can read and whether that matches the data's classification. Integrations expose data, so the first question is what is reachable and whether policy allows it. Vendor guidance is useful background but does not tell you what your own connector can see, and write access or speed are secondary to scope. Governance, risk, and responsible use A colleague wants to paste a customer's full medical file into a consumer chat account to summarize it. What is the right response? - Paste it, get the summary, and delete the conversation from the account history afterwards - Remove the customer's name and date of birth, then paste the rest into the consumer account - Ask the customer for written consent by email first, then paste the file into the consumer account - Check the data policy first and use an approved workspace with the right settings, if one exists Check answer ### Show answer Answer: Check the data policy first and use an approved workspace with the right settings, if one exists. Sensitive personal data is governed by policy and by the workspace's data handling, not by speed. Stripping the name still leaves a medical record that is often re-identifiable, consent from the customer does not change what the consumer account does with the data, and deleting afterwards is too late. Governance, risk, and responsible use Claude drafts screening criteria for job applicants. What must happen before they are used? - A review for bias and legality by the people who are accountable for the hiring decision - A check that each criterion appears in the job posting so applicants know how they are assessed - A comparison against the criteria a competitor lists publicly to make sure nothing is missing - A trial run against last year's applicant pool to confirm the criteria produce a sensible shortlist Check answer ### Show answer Answer: A review for bias and legality by the people who are accountable for the hiring decision. Decisions about people carry bias and legal risk, so accountable people review before use. A trial run on past applicants sounds rigorous but only shows the criteria produce a shortlist, not that the shortlist is fair or lawful, and it processes real candidates' data for no decision. Prompting and task execution The prompt "write something about our product" produced a vague result. What improves it most? - Pasting the full product spec sheet and asking for a write-up that covers everything in it - Telling Claude the result was vague and asking it to try again with more concrete detail - Adding the audience, purpose, format, length, and an example of the tone you want - Asking Claude to write three versions so you can pick the one closest to what you had in mind Check answer ### Show answer Answer: Adding the audience, purpose, format, length, and an example of the tone you want. Context, constraints, and an example are what turn a vague ask into a usable draft. Pasting the spec sheet adds material but no purpose, so the result is a long version of the same vague piece; asking for concrete detail without saying which detail leaves Claude guessing. Prompting and task execution The first draft is close but not right. What is the efficient next step? - Start a fresh conversation and rewrite the prompt with the lessons from the first attempt - Edit the draft by hand, since explaining the changes takes longer than making them - Ask for ten alternative drafts and combine the best paragraphs from each yourself - Give specific feedback on what to change and iterate on the same draft Check answer ### Show answer Answer: Give specific feedback on what to change and iterate on the same draft. Iterating with specific feedback converges faster than restarting, because the draft already carries most of what you want. Rewriting the prompt from scratch throws that away, and hand-editing works once but teaches Claude nothing for the next draft. Product and model selection You need to rename a field across 200 files in a code repository. Which tool fits? - A Project with the repository files attached so Claude can see all of them at once - A chat conversation, pasting each file in turn and copying the edited version back - A spreadsheet listing every file path so Claude can generate the edits row by row - Claude Code, or a developer using it, working in the repository directly Check answer ### Show answer Answer: Claude Code, or a developer using it, working in the repository directly. Repository-wide changes are what Claude Code is for: it reads and edits the files in place. A Project with the files attached lets Claude read them but not change them in the repository, so every edit still has to be copied back by hand. Product and model selection For quick drafting of routine emails, what is the sensible model choice? - A faster, cheaper model, saving the most capable one for deep analysis - Rotate between models on each email so you can compare their tone over time - The most capable model for every email, since consistency of quality matters most - Whichever model was released most recently, since newer models replace older ones Check answer ### Show answer Answer: A faster, cheaper model, saving the most capable one for deep analysis. Match the model to the job: routine drafting does not need the heaviest model. Using the most capable model everywhere is not wrong on quality, but it spends time and cost on work that a faster model handles just as well. Configuration and knowledge management Your team runs the same weekly task with the same standing instructions and reference files. What setup avoids re-explaining every time? - A saved prompt template that each person pastes as the first message every week - A Project with the standing instructions and reference files attached - A single long-running conversation that the whole team continues week after week - A shared document of past outputs that Claude is asked to imitate each week Check answer ### Show answer Answer: A Project with the standing instructions and reference files attached. Projects hold standing instructions and knowledge so every conversation starts with the right context. A pasted template works but still re-sends everything each week and drifts as people edit their copies, and one endless conversation degrades as it grows. Troubleshooting and optimization Halfway through a long conversation Claude stops following an instruction you gave at the start. Most likely cause and fix? - The instruction conflicted with a later request; shorten your messages from here on - The instruction is buried in a long context; restate it, or move it into project instructions - A model update changed its behavior mid-conversation; report the bug and switch to a different model - The conversation hit a length limit; keep going and the instruction will apply again shortly Check answer ### Show answer Answer: The instruction is buried in a long context; restate it, or move it into project instructions. Long contexts dilute early instructions, so restating them or moving them into standing instructions fixes it. A conflict with a later request is possible, but shortening messages does not resolve it; you would restate the rule either way. Troubleshooting and optimization Claude declines a legitimate request about a sensitive topic. What is the productive move? - Tell Claude that you accept full responsibility for how the answer will be used and retry - Move the request to a different model that is known to refuse fewer sensitive topics - Add context about who you are and the legitimate purpose, then rephrase the request - Split the request into smaller pieces that each look harmless and combine the answers yourself Check answer ### Show answer Answer: Add context about who you are and the legitimate purpose, then rephrase the request. Context about purpose resolves most false refusals. Splitting the request into harmless-looking pieces is a workaround that violates usage policies even when the underlying purpose is fine, and a blanket acceptance of responsibility gives Claude no information about the purpose. Output evaluation and validation You ask for a summary of a document you attached, and the answer includes facts not in the document. What do you do? - Ask Claude to use only the attached document and cite where each point comes from, then verify - Re-upload the document in a new chat, since the extra facts suggest the upload was corrupted - Remove the extra facts yourself and send the rest, since the document content is presumably fine - Keep the extra facts if they are accurate, since they add useful context for the reader Check answer ### Show answer Answer: Ask Claude to use only the attached document and cite where each point comes from, then verify. Constrain the source and demand citations, then check them. Removing the additions by hand assumes the rest of the summary is grounded, which is exactly what you have not verified; the additions are a sign the whole summary needs checking. Output evaluation and validation Claude produced a legal-sounding clause for a contract. What is the right next step? - Compare it against a similar clause in a signed contract and use it if the wording matches - Ask Claude to confirm the clause is enforceable in your jurisdiction before adding it - Have counsel review it, treating the draft as a starting point rather than as advice - Add it to the contract with a note that it was AI-drafted so the other party can review it Check answer ### Show answer Answer: Have counsel review it, treating the draft as a starting point rather than as advice. Drafts with legal consequences need a qualified reviewer. Matching the wording of a signed contract feels like a check, but that clause was reviewed for a different deal and a different context; only counsel can say whether this one fits yours. Output evaluation and validation A calculation in Claude's answer looks off. What do you do? - Ask Claude to show its working step by step and accept the result if the steps look sound - Ask Claude to double-check its own calculation and flag anything it is not sure about - Ask the same question again in a new conversation and use whichever figure appears twice - Re-run the arithmetic yourself or in a spreadsheet and correct the answer if needed Check answer ### Show answer Answer: Re-run the arithmetic yourself or in a spreadsheet and correct the answer if needed. Verify numbers independently. Step-by-step working is easier to read but the model can present a wrong step just as confidently as a right one, and asking it to double-check or re-run is still the model grading itself. Output evaluation and validation Which is the best signal that a summary is faithful to the source? - It reuses the source's own terminology and section headings throughout - Every key claim can be traced back to a specific passage in the source - It was produced by the most capable model with the source attached to the chat - It reads as balanced and does not add opinions that the source did not state Check answer ### Show answer Answer: Every key claim can be traced back to a specific passage in the source. Traceability is what faithfulness means: each claim has a passage behind it. Reusing the source's vocabulary makes a summary sound faithful without proving any claim, and the absence of added opinion says nothing about whether the facts were kept straight. Output evaluation and validation Claude answers a question about your company's Q3 revenue with a specific figure you never provided. What happened? - It inferred the figure from earlier numbers in the conversation, so the math is checkable - It pulled the figure from a connected finance system that an admin set up earlier - It guessed; models fill gaps with plausible text unless grounded in provided data - It retrieved the figure from a public filing, so it is probably close to correct Check answer ### Show answer Answer: It guessed; models fill gaps with plausible text unless grounded in provided data. Ungrounded specifics are guesses. A public filing is a tempting explanation, but unless you attached or connected one, the model had no such source and produced a plausible-looking number; treat it as unverified. Output evaluation and validation What is a practical way to check a translated customer email before sending? - Ask Claude to rate the translation's accuracy on a scale of one to ten before you send it - Run the email through a second translation tool and send it if the two outputs match closely enough - Ask Claude to confirm the tone is polite and appropriate for a customer in that market - Ask for a back-translation into your language and compare meaning, or have a native speaker skim it Check answer ### Show answer Answer: Ask for a back-translation into your language and compare meaning, or have a native speaker skim it. Back-translation exposes meaning drift in a language you can read. A second tool agreeing is weaker than it looks, because two tools can share the same misreading, and a tone check does not catch a sentence that says the wrong thing politely. Output evaluation and validation Which output needs the most scrutiny? - A list of medication interactions for a patient handout - A draft tweet announcing a new feature to your followers - A meeting agenda for the weekly product sync - A brainstorm of twenty blog titles for the marketing team Check answer ### Show answer Answer: A list of medication interactions for a patient handout. Consequence sets the review bar, and medical information that reaches patients sits at the top. The tweet is public and deserves a look, but a wrong feature announcement is embarrassing while a wrong interaction list can hurt someone. Output evaluation and validation Claude gives two different answers to the same factual question on two runs. What does that tell you? - The first answer is more likely right, since the context was cleaner at the start - Both answers are probably partly right, so combine them into a single response - The second answer is more likely right, since the model refined its response - The model is uncertain or the question is ambiguous; verify against a source Check answer ### Show answer Answer: The model is uncertain or the question is ambiguous; verify against a source. Inconsistency is a verification flag, not a tie-break puzzle. Neither run has a claim to being the refined one; the two runs are independent samples, so the disagreement tells you to go to a source rather than pick or merge. Output evaluation and validation Your team wants a rule of thumb for when to trust Claude's output without checking. Best rule? - Trust output from the most capable model; verify output from faster, cheaper models - Trust drafts you will edit anyway; verify anything factual, numeric, legal, medical, or public-facing - Trust answers under 100 words; verify anything longer, since longer answers have more room for errors - Trust answers that cite a source; verify only the answers that come with no citation at all Check answer ### Show answer Answer: Trust drafts you will edit anyway; verify anything factual, numeric, legal, medical, or public-facing. Stakes and checkability decide, not model size or length. Citations are the strongest-looking distractor, but a cited answer still needs the citation opened, because models can produce plausible references that do not say what is claimed. Output evaluation and validation Claude cites a study by name and year. What is the check? - Ask Claude for the DOI and journal name, which it can only supply if the study is real - Confirm the study exists and says what is claimed, using the journal or a database - Ask Claude whether it is confident the citation is accurate and keep it if it says yes - Check that the year is recent enough for the findings to still apply to your situation Check answer ### Show answer Answer: Confirm the study exists and says what is claimed, using the journal or a database. Citations can be fabricated, so confirm both existence and content outside the model. Asking for a DOI does not help, because a model can produce a well-formed DOI for a study that does not exist; the check has to happen in a database or the journal itself. Workflow integration and solution design A support team wants Claude to draft replies to tickets. Which design keeps quality high? - Claude answers tickets it rates as simple automatically and routes the rest to agents, with no review of the automated ones - Claude drafts from the ticket; agents send the draft unchanged and log any customer complaints for review each quarter - Claude drafts from the ticket and knowledge base; an agent reviews and sends; wrong drafts feed back into instructions - Claude drafts from the ticket; a manager samples a fixed 5% of sent replies each week and updates instructions from the sample Check answer ### Show answer Answer: Claude drafts from the ticket and knowledge base; an agent reviews and sends; wrong drafts feed back into instructions. Draft, review, learn: every reply is checked before it goes out and mistakes improve the instructions. Sampling 5% after sending has the learning loop but not the review, so 95% of errors reach customers unchecked. Workflow integration and solution design Which task is a poor fit for Claude in a finance workflow? - Making the final approval decision on employee expense reports - Drafting a memo that explains the monthly results to leadership - Extracting figures from scanned invoices for a person to review - Summarizing variance commentary from the business unit leads Check answer ### Show answer Answer: Making the final approval decision on employee expense reports. Approval decisions stay with accountable people. Invoice extraction is the one candidates sometimes pick because the numbers matter, but extraction feeding a human review is exactly the pattern that works; it is the decision that is the poor fit. Workflow integration and solution design What should you map before adding Claude to a process? - The team's existing prompts and templates, so you can standardize them before rolling out anything new - The steps, inputs, outputs, and where judgment happens, so you know which steps are drafting and which are decisions - The list of available models and their costs, so you pick the cheapest one that works - The budget and timeline, so you know how much automation you can afford to build Check answer ### Show answer Answer: The steps, inputs, outputs, and where judgment happens, so you know which steps are drafting and which are decisions. Process clarity precedes automation: you need to know which steps are drafting or extraction and which are decisions. Standardizing prompts is worthwhile later, but it assumes you already know which steps Claude should touch. Workflow integration and solution design A marketing team wants weekly competitor summaries. Which setup? - A fresh chat each week where the analyst pastes the competitor list and the latest news links - An automated agent that gathers competitor news and posts the summary straight to the team channel each Monday - A shared conversation that the team keeps open all quarter so Claude retains the full history - A Project with standing instructions and the competitor list, run weekly with a person reviewing before circulation Check answer ### Show answer Answer: A Project with standing instructions and the competitor list, run weekly with a person reviewing before circulation. Standing setup plus review: the Project holds the context and a person checks before circulation. The automated agent saves the most effort but drops the review, so a misread headline or a fabricated detail goes to the whole team. Workflow integration and solution design Which is a sign that a task should stay manual for now? - It involves judgment calls that the team currently makes based on experience - Errors are costly, the inputs are inconsistent, and there is no quick way to check outputs - It runs weekly and the inputs arrive in slightly different formats each time - It takes under an hour and the team already has a template that works well Check answer ### Show answer Answer: Errors are costly, the inputs are inconsistent, and there is no quick way to check outputs. Hard-to-verify, high-cost tasks wait. Judgment calls sound like a stop sign, but many judgment-heavy tasks still benefit from Claude drafting options for the person deciding; it is the combination of costly errors and no quick check that rules a task out. Workflow integration and solution design How do you measure whether Claude improved a workflow? - Survey the team after a month on whether the work feels faster and less tedious - Compare the tool's monthly cost against the hours the team estimates it saved - Compare time, error rate, and rework before and after on the same kind of work - Track how many prompts the team writes per week as a proxy for adoption Check answer ### Show answer Answer: Compare time, error rate, and rework before and after on the same kind of work. Measure outcomes on comparable work: time, errors, and rework, before and after. The cost-versus-estimated-hours comparison looks quantitative, but estimated savings are an opinion with a number on it and it ignores whether quality held. Workflow integration and solution design When connecting Claude to a shared drive, which question comes first? - Which team members should get access to the connected assistant first - Whether Claude can write files back to the drive or only read them - How quickly it returns results when searching across a large folder tree - Which folders it may read and whether the data classification allows it Check answer ### Show answer Answer: Which folders it may read and whether the data classification allows it. Access scope and policy come first, because a connector exposes whatever it can reach. Deciding who gets access is a real question, but it comes after you know what data the connector can see; scoping the folders is what limits the exposure. Workflow integration and solution design A colleague builds a workflow where Claude's output goes straight into a customer-facing form. What is missing? - A review or validation step before anything reaches the customer - A more capable model so the output is accurate enough to send directly - A log of every output so problems can be traced after customers report them - A more detailed prompt so the output matches the form's fields exactly Check answer ### Show answer Answer: A review or validation step before anything reaches the customer. Customer-facing output needs a check before it goes out. Logging is good practice, but it only finds problems after a customer has seen them; a better prompt or model lowers the error rate without catching the errors that remain. Governance, risk, and responsible use Which content should not be pasted into a tool your organization has not approved for it? - Product descriptions that are already on the public website - Your own meeting notes from an internal planning session - Customer personal data, health records, or confidential financials - Press releases the company published in the last quarter Check answer ### Show answer Answer: Customer personal data, health records, or confidential financials. Sensitive classes need approved tools and settings. Internal meeting notes are the closest call, since they are not public, but ordinary planning notes are a policy question for your organization; personal, health, and confidential financial data are the classes that are governed everywhere. Governance, risk, and responsible use Why does responsible use require a named person accountable for AI-assisted decisions? - Because it gives the organization a single person to hold responsible when the model makes a costly error - Because the vendor's terms of service transfer liability for outputs to the customer - Because the model cannot be accountable; a person must be able to explain and stand behind the outcome - Because regulators require a named AI officer to be listed for every automated decision Check answer ### Show answer Answer: Because the model cannot be accountable; a person must be able to explain and stand behind the outcome. Accountability stays human because a model cannot explain or defend an outcome. Having someone to blame is a distortion of the same idea: the point is that a person can understand and stand behind the decision, not that there is a name to point at afterwards. Governance, risk, and responsible use Claude drafts performance review language for an employee. Which risk is highest? - Formatting that does not match the HR system's review template and has to be redone by hand - Language so generic that it gives the employee nothing concrete to act on in the coming year - Bias in the language or reasoning affecting a person's career, so it needs careful human review - Length that runs well past what managers are expected to write for an annual review Check answer ### Show answer Answer: Bias in the language or reasoning affecting a person's career, so it needs careful human review. People decisions carry bias risk, and a review shapes someone's career. Generic language is a real quality problem, but it is visible and easy to fix; biased framing can look polished and still be unfair, which is why a person has to read it carefully. Governance, risk, and responsible use What does an acceptable use policy typically prohibit? - Using the tool for harmful, deceptive, or otherwise prohibited purposes - Asking questions about topics that are unrelated to your job role - Sharing outputs with anyone outside the team that produced them - Using the tool for any task that involves customer names or contact details Check answer ### Show answer Answer: Using the tool for harmful, deceptive, or otherwise prohibited purposes. Acceptable use policies define forbidden purposes: harm, deception, and listed prohibited uses. Customer data handling is governed by data policy and workspace settings rather than a blanket ban in the AUP, and sharing outputs or off-topic questions are not what it covers. Governance, risk, and responsible use A vendor claims its AI tool 'never makes mistakes'. Your response? - Ask the vendor for the accuracy benchmarks behind the claim and accept it if they look strong - Include the claim in the contract so the vendor bears the cost of any mistakes - Treat the claim as false and design review steps into the workflow anyway - Run a two-week pilot and drop the review steps once the pilot shows no errors Check answer ### Show answer Answer: Treat the claim as false and design review steps into the workflow anyway. All models err, so design for it. A pilot is sensible but two clean weeks do not prove the tool is error-free, only that the errors did not show up yet; removing review on that basis is the mistake the claim was inviting you to make. Governance, risk, and responsible use What should you do if Claude produces content that seems to reproduce a copyrighted text at length? - Shorten the passage so that it falls under what fair use typically allows - Use it, since content produced by an AI tool is not subject to copyright - Cite Claude as the author so the original rights holder is not misrepresented - Do not use it verbatim; rewrite or source it properly, and flag the case Check answer ### Show answer Answer: Do not use it verbatim; rewrite or source it properly, and flag the case. Avoid reproducing protected text. Trimming to a length you believe is fair use is the strongest distractor, but fair use depends on purpose and context, not word count, and you would still be publishing someone else's text; rewrite or source it and flag the case. Governance, risk, and responsible use Which practice supports transparency when AI helped produce a document? - Keeping AI-assisted drafts out of the document history so reviewers see only the final - Disclosing AI assistance where your organization's policy or the audience expects it - Publishing under a team name rather than an individual so no one is misattributed - Adding a footnote naming the model version used on every document you produce Check answer ### Show answer Answer: Disclosing AI assistance where your organization's policy or the audience expects it. Disclose according to policy and audience expectation. A footnote on every document sounds maximally transparent, but blanket labels are noise on routine drafts and do not reflect where disclosure matters; policy and audience decide where it belongs. Prompting and task execution Which prompt is most likely to produce a usable customer apology email? - 'Draft a customer apology email. Use our standard tone. Include a credit. Do not make it too long.' - 'Write an apology to a customer for a late order. Keep it professional and make sure they feel heard.' - 'Write a 120-word apology to a customer whose order arrived a week late, warm but not groveling, offering a 15% credit; here is our tone guide.' - 'Write a sincere apology for a shipping delay from a customer service manager, with a goodwill gesture.' Check answer ### Show answer Answer: 'Write a 120-word apology to a customer whose order arrived a week late, warm but not groveling, offering a 15% credit; here is our tone guide.'. Specifics win: audience, length, tone, the exact offer, and a tone reference. The second option mentions tone and a credit but names neither, so Claude has to invent both; 'our standard tone' means nothing unless the guide is attached. Prompting and task execution You want three distinct options rather than three variations of one idea. What do you say? - 'Give me three options and make sure they are genuinely different from each other, not variations' - 'Give me three options that differ in approach, for example one conservative, one bold, one unconventional' - 'Give me three options. Be as creative as possible and avoid repeating yourself across them.' - 'Give me three options, then rank them by how creative each one is so I can see the range' Check answer ### Show answer Answer: 'Give me three options that differ in approach, for example one conservative, one bold, one unconventional'. Define the axis of difference. Saying 'genuinely different, not variations' states the goal but not what different means, so you often get the same idea in three tones; naming the spread (conservative, bold, unconventional) gives Claude something to differ on. Prompting and task execution Claude's draft uses jargon your audience will not understand. Best fix? - Tell it the audience and ask for plain language with an example of the level you want - Replace the jargon yourself before sending, since you know the audience best - Ask it to add a glossary at the end so readers can look up the terms they do not know - Ask for a shorter draft, since cutting length usually forces out the technical terms Check answer ### Show answer Answer: Tell it the audience and ask for plain language with an example of the level you want. Name the audience and show the level. A glossary keeps the jargon and pushes the work onto the reader, and replacing terms by hand fixes this draft without changing the next one. Prompting and task execution What is the value of giving Claude the raw material (notes, data, previous versions) with your request? - Grounded requests produce specific, accurate drafts instead of generic ones - It lets the model learn your company's style for future conversations - It shortens the conversation, since Claude can skip the clarifying questions - It lets Claude verify its own claims against your documents before answering Check answer ### Show answer Answer: Grounded requests produce specific, accurate drafts instead of generic ones. Material grounds the output. The verification framing overstates it: attached material makes the draft specific and accurate, but it does not make the model check itself, and you still verify the result. Prompting and task execution When a task is large, what approach works better? - Start with the hardest section so the remaining ones can follow its pattern - Break it into steps: outline first, then sections, reviewing between steps - Ask for the full deliverable twice and merge the stronger parts of each version - Write one detailed prompt that covers every section so the output is consistent Check answer ### Show answer Answer: Break it into steps: outline first, then sections, reviewing between steps. Stepwise work is easier to steer, because each review catches drift before it spreads. One giant prompt gives consistency in theory but a mistake in the framing shows up everywhere at once; starting with the hardest section is a sequencing choice, not a structure. Prompting and task execution Claude keeps adding a closing paragraph you do not want. What is the efficient instruction? - 'Rewrite this without the fluff at the end' - 'Keep it brief and skip anything that is not essential' - 'End after the final bullet; do not add a closing paragraph' - 'Match the length of the previous version exactly' Check answer ### Show answer Answer: 'End after the final bullet; do not add a closing paragraph'. State the exact boundary. 'Without the fluff at the end' points at the problem but leaves Claude to decide what counts as fluff, so the closing often comes back shorter rather than gone. Product and model selection For a long, careful analysis of a 60-page report, which is the better choice? - The most capable model, accepting slower responses for the depth of analysis - Whichever model the team used last time, so results are comparable across reports - The fastest model, since it can be run several times and the answers compared - A smaller model with the report split into chunks so each fits in a short prompt Check answer ### Show answer Answer: The most capable model, accepting slower responses for the depth of analysis. Depth favors capability. Chunking the report for a smaller model is the strongest alternative, but it loses the cross-references between sections that a careful analysis depends on. Product and model selection Which feature lets a whole team share standing instructions and files? - A shared Project with the instructions and files attached - A shared browser bookmark to a chat with the instructions pinned at the top - A saved conversation that each team member opens and continues - A team email thread that holds the latest version of the instructions Check answer ### Show answer Answer: A shared Project with the instructions and files attached. Projects share context. A saved conversation looks similar but mixes everyone's messages into one growing thread, and it has no place for instructions that apply to every new conversation. Product and model selection A colleague asks whether to use the consumer chat app or an organization workspace for client work. Deciding factor? - Which one has access to the newest model, since client deliverables should use the best output - Cost per seat, since the consumer app is cheaper and client budgets are tight - Data handling and admin controls: the organization workspace governs what happens to client data - Response speed and rate limits, since client work is time-sensitive and volume can be high Check answer ### Show answer Answer: Data handling and admin controls: the organization workspace governs what happens to client data. Data handling and admin controls decide; client work belongs in the governed workspace. Model access is a real difference between plans, but a better model in an ungoverned account does not fix where the client's data goes. Product and model selection Which task fits Claude Code rather than the chat app? - Drafting a newsletter that goes out to the engineering team each month - Making the same change across dozens of files in a code repository - Writing a job post for a senior developer role from the hiring manager's notes - Summarizing a recorded architecture review call for the people who missed it Check answer ### Show answer Answer: Making the same change across dozens of files in a code repository. Repository work is Claude Code's job. The other three are about engineers but are still drafting and summarizing tasks, which the chat app handles well; what matters is whether the work happens inside a codebase. Product and model selection When would a smaller, faster model be the right pick? - For high-volume, simple tasks like tagging or short drafts where speed and cost matter more than depth - When the conversation is going to run long, since smaller models hold context more reliably over time - For legal or medical analysis, where a concise answer is safer than a long one - When the team is new to AI tools, since simpler models are easier to prompt well Check answer ### Show answer Answer: For high-volume, simple tasks like tagging or short drafts where speed and cost matter more than depth. Match model to task: volume and simplicity favor speed and cost. Legal and medical analysis is the trap, because concise is not the same as careful, and those are the tasks where depth matters most. Configuration and knowledge management What belongs in a Project's instructions? - Every document the team has produced on the topic, so Claude has the fullest possible picture - A log of past conversations that went well, so Claude can reproduce the pattern - Standing context: who the audience is, tone, formats, and rules that apply to every conversation - The specific request for this week, updated each time the task changes Check answer ### Show answer Answer: Standing context: who the audience is, tone, formats, and rules that apply to every conversation. Standing rules, not everything. This week's request belongs in the conversation, not the instructions, because it changes; documents belong in knowledge files rather than instructions, and even then only the ones the project regularly needs. Configuration and knowledge management Your Project's knowledge files are out of date and answers reflect old policy. Fix? - Add a line to the instructions telling Claude to prefer the most recent file when they conflict - Replace the outdated files and tell the team that the source of truth has changed - Add the new policy document alongside the old ones so Claude can see both versions - Start a new Project from scratch and move the team over once it is set up Check answer ### Show answer Answer: Replace the outdated files and tell the team that the source of truth has changed. Maintain the knowledge base: remove what is wrong and tell people. Keeping both versions with a 'prefer the newest' rule leaves the old policy available to be quoted, and Claude cannot reliably tell which file is newer from the content. Configuration and knowledge management What is a good way to keep a team's prompts consistent? - A weekly review where the team compares the prompts each person wrote that week - A rule that everyone uses the same model so the outputs come out in a similar style - Shared templates or skills with the agreed structure, stored where the team can find them - One designated person who writes every prompt and sends it to the rest of the team Check answer ### Show answer Answer: Shared templates or skills with the agreed structure, stored where the team can find them. Shared, findable templates keep structure consistent without a bottleneck. A single prompt writer gives consistency but does not scale and stops the moment that person is away; a weekly review compares prompts after the inconsistent work is already done. Configuration and knowledge management Which is a reason to split work into several Projects? - The knowledge files are large and splitting them spreads the storage across Projects - Projects perform better when each one holds fewer than ten conversations - Each team member should own a Project so their conversations stay private - Different audiences, tones, or confidentiality levels that should not mix Check answer ### Show answer Answer: Different audiences, tones, or confidentiality levels that should not mix. Separate contexts that should not blend. One Project per person is the common misuse: it splits by owner instead of by context, so the shared instructions and files that make a Project useful end up duplicated or missing. Configuration and knowledge management How much should a knowledge file contain? - A single summary page, since longer files make answers slower and less focused - What conversations in that Project regularly need, kept current - Only finalized documents; drafts and working notes should stay out of the Project - Everything the team has written on the topic, so nothing relevant is left out Check answer ### Show answer Answer: What conversations in that Project regularly need, kept current. Relevant and current is the rule. 'Only finalized documents' sounds disciplined, but a current working draft is often exactly what conversations need; the test is whether the content is used and up to date, not its approval status. Configuration and knowledge management What should you record when a prompt template works well? - The name of the person who wrote it, so colleagues know who to ask about it - The number of times it has been used, so the team can see which templates matter - The template, the inputs it expects, and an example output, so others can reuse it - The model version and the date, so the team knows when it may need to be re-tested Check answer ### Show answer Answer: The template, the inputs it expects, and an example output, so others can reuse it. Document reusable assets so others can pick them up. Model version and date are worth noting alongside, but on their own they do not let anyone reuse the template; the inputs and an example output are what make it usable. Troubleshooting and optimization Claude's answer ignores an attached document. First thing to check? - That the document uploaded correctly and that your prompt refers to it explicitly - Whether the conversation is too long and the document has dropped out of context - Whether the model you selected supports file attachments of that size and type - Whether the document contains text or only scanned images, which need conversion first Check answer ### Show answer Answer: That the document uploaded correctly and that your prompt refers to it explicitly. Confirm the input and the reference before anything else. Scanned images and context length are real causes, but they are second-line checks; most cases are a failed upload or a prompt that never mentions the document. Troubleshooting and optimization A conversation has drifted and answers are worse. What helps? - Ask Claude to summarize the conversation so far and continue from that summary in the same chat - Start a new conversation with a concise summary of the task and decisions so far - Switch to a more capable model mid-conversation so it can recover the thread - Shorten your messages so the model has less new text to process each turn Check answer ### Show answer Answer: Start a new conversation with a concise summary of the task and decisions so far. Reset with a clean brief. Summarizing inside the same chat is the near miss: the summary helps, but the drifted history is still in the context and keeps pulling answers off course. Troubleshooting and optimization Claude says it cannot access a link you pasted. Why? - The link is broken; check it in your browser and resend it once it loads - The link is too long for the model to parse; shorten it with a URL shortener and try again - It may not be able to browse, or the page may be restricted; paste the content instead - The page needs a login; sign in yourself first and paste the link again, since the session carries over Check answer ### Show answer Answer: It may not be able to browse, or the page may be restricted; paste the content instead. Provide the content directly. Signing in yourself does nothing, because your browser session does not carry over to the model; if the page is behind a login or the model cannot browse, pasting the text is the only reliable route. Troubleshooting and optimization Output is in the wrong format despite instructions. Fastest fix? - Switch to a more capable model, since format compliance improves with model size - Show a short example of the exact format you want - Ask Claude to explain why it chose that format before trying again - Repeat the same instruction more firmly and put it at the very end of the prompt Check answer ### Show answer Answer: Show a short example of the exact format you want. Examples fix format faster than any description. Repeating the instruction more firmly sometimes works, but an example removes the ambiguity that caused the miss in the first place. ## Drill one domain at a time Each domain with enough questions has its own page with study notes and just that domain's questions. - [CCAO-F Output evaluation and validation questions](https://youraidept.com/network/ccao-f-practice-questions/output-evaluation-and-validation) (13) - [CCAO-F Workflow integration and solution design questions](https://youraidept.com/network/ccao-f-practice-questions/workflow-integration-and-solution-design) (10) - [CCAO-F Governance, risk, and responsible use questions](https://youraidept.com/network/ccao-f-practice-questions/governance-risk-and-responsible-use) (9) - [CCAO-F Prompting and task execution questions](https://youraidept.com/network/ccao-f-practice-questions/prompting-and-task-execution) (8) - [CCAO-F Product and model selection questions](https://youraidept.com/network/ccao-f-practice-questions/product-and-model-selection) (7) - [CCAO-F Configuration and knowledge management questions](https://youraidept.com/network/ccao-f-practice-questions/configuration-and-knowledge-management) (7) - [CCAO-F Troubleshooting and optimization questions](https://youraidept.com/network/ccao-f-practice-questions/troubleshooting-and-optimization) (6) ## How the real CCAO-F exam differs - Questions: 60 items in 120 minutes - Passing score: 720 on a 100–1,000 scale - Exam fee: $99 per attempt, as reported by prep guides - Format: Proctored and closed-book, no AI assistance - Delivered by Pearson VUE, online proctored or at a test center, with no notes and no AI assistance. ## Practice question FAQ ### Are these real CCAO-F exam questions? + No. They are original scenario questions for non-developers, written from the published blueprint. ### Do I need to code for CCAO-F? + No. Nothing on the blueprint requires code. The heaviest domain is evaluating and validating Claude's output. ### How many questions are on the real CCAO-F exam? + 60 questions in 120 minutes, proctored, with a passing score of 720 on a 100 to 1,000 scale. ### Where do I register for CCAO-F? + Through the Anthropic Partner Academy, via a Claude Partner Network organization, with delivery through Pearson VUE. ## Practice questions for the other exams CCAR-F ### [Claude Certified Architect – Foundations](https://youraidept.com/network/ccar-f-practice-questions) CCDV-F ### [Claude Certified Developer – Foundations](https://youraidept.com/network/ccdv-f-practice-questions) CCAR-P ### [Claude Certified Architect – Professional](https://youraidept.com/network/ccar-p-practice-questions) ## Who wrote this ### [Matei Olaru](https://youraidept.com/about/matei-olaru) CEO, Your AI Department Former public-company CEO, corporate attorney, and World Bank consultant. Runs the YAID Network and the partner relationship with Anthropic. ### [Michael Greenberg](https://youraidept.com/about/michael-greenberg) AI Product Lead, Your AI Department AI at Google, Meta, and Samsung. Six patents. Leads member support and the certification prep guidance. Your AI Department is a registered Claude Partner Network company. Our members sit these exams under our partner standing. Facts on this page were last checked on October 15, 2026. ## Ready to sit CCAO-F for real? Registration needs a Claude Partner Network organization. [YAID Network](https://youraidept.com/network) members register under our partner standing and prepare with the official Partner Academy course. $149 per quarter, exam fee separate. [Apply to join YAID ↗](https://youraidept.com/network/apply?ref=practice)[How registration works →](https://youraidept.com/network/register-for-claude-certification)