Mock exam 1: Associate – Foundations
A timed practice exam in the official style. Unlike the practice questions, the domains are hidden and the answers sit at the end, so this measures readiness rather than teaching. Original questions written against the public blueprint; not items from the live exam, which is covered by a non-disclosure agreement.
How to take it. 15 questions, 30 minutes, closed book: no notes, no documentation, no Claude. Answer every question, since unanswered items score zero. Flag anything you are unsure of and return to it. Score yourself with the key and the readiness table afterwards.
Questions
1. A colleague shares a Claude-drafted policy summary that reads well and cites three internal documents by name. What is the appropriate review step before it circulates to the whole department?
- A. Circulate it, since the named sources demonstrate grounding
- C. Confirm each named document exists and says what the summary claims
- B. Ask Claude to double-check its own work and circulate if it agrees
- D. Add a disclaimer that the summary was AI-generated and circulate
2. You need a first draft of a stakeholder update from a long project log. Which prompt is most likely to produce a usable draft on the first attempt?
- A. "Summarize this project log."
- D. "You are writing a weekly update for non-technical executives. From the log, produce: status in one line, three progress items, two risks with owners, and next week's focus. Keep it under 200 words."
- B. "Write something good about this project log for my boss."
- C. "Summarize this project log. Be detailed and thorough and professional."
3. A workflow drafts hundreds of routine appointment confirmations each day. Latency and cost dominate; the language is formulaic. Which choice fits?
- B. The most capable model, for consistency
- A. A fast, low-cost model, reserving a more capable one for exceptions
- C. Whichever model is newest
- D. Alternate between models to spread cost
4. Your team's Project answers questions from a product manual. Users report confident answers about a feature that was removed last quarter. What is the most likely cause?
- A. The model needs retraining
- B. The Project's knowledge still contains the outdated manual
- C. Users are asking the questions wrong
- D. The context window is too small
5. Legal asks you to analyze a spreadsheet of support tickets containing customer names and email addresses. Your organization prohibits sending personal data to external tools. What do you do?
- A. Upload it, since legal requested the analysis
- C. Strip or pseudonymize the personal columns, then analyze what remains
- B. Upload it and ask Claude to disregard the personal columns
- D. Tell legal the analysis is not possible
6. Claude produces a five-paragraph answer when you needed three bullet points, twice in a row. What is the most effective adjustment?
- A. Ask again and hope for variation
- D. State the format explicitly, including the number of bullets and their maximum length
- B. Switch to a more capable model
- C. Manually reformat each response
7. Which of these is the strongest signal that a Claude-generated claim needs verification before use?
- B. The answer is long
- A. The claim is a specific number, date, or citation that would matter if wrong
- C. The answer was produced quickly
- D. The answer uses technical vocabulary
8. A manager wants to know whether Claude can "replace our weekly reporting process." What is the most useful response?
- A. Yes, automation is what AI is for
- B. A breakdown of which reporting steps Claude can draft or summarize, which need human judgment or system access, and what the review step would be
- C. No, reporting requires human oversight
- D. A recommendation to hire a consultant
9. You are configuring a Project that several colleagues will use for consistent client communications. What belongs in the Project instructions rather than in each individual prompt?
- A. The specific client's name for today's message
- C. The house tone, required disclaimers, and the standard structure of a client message
- B. Today's date
- D. The colleague's personal writing preferences
10. A summary Claude produced omits a material caveat that appears in the source document. What does this failure mode illustrate?
- A. Hallucination through fabricated detail
- D. Omission, which requires reading the source rather than only reviewing the output
- B. A context window limit
- C. A prompt injection attack
11. Your prompt worked well last month and now returns weaker results after your team added twelve documents to the Project's knowledge. What is the first thing to investigate?
- B. Whether the model changed
- A. Whether the new documents are diluting or contradicting the sources the answers should use
- C. Whether colleagues are using different browsers
- D. Whether the prompt should be longer
12. For which task is a Project with persistent instructions and knowledge clearly better than a one-off chat?
- A. Asking for a synonym
- B. A monthly report that follows the same structure and draws on the same reference documents
- C. Checking a single fact
- D. Rewriting one email
13. Which practice best reduces the risk of sharing an inaccurate Claude output with a customer?
- A. Reading the output for tone before sending
- C. Verifying the claims that would cause harm if wrong, then having a second person review customer-facing material
- B. Sending it with a note that AI assisted
- D. Using the largest available model
14. You want Claude to help redesign a slow approval process rather than just speed up its current steps. Which prompt direction fits that goal?
- A. "Make our approval process faster."
- D. "Here are the current approval steps and where delays occur. Propose two redesigns: one that keeps the current structure and one that reorders or removes steps, with the tradeoffs of each."
- B. "Automate our approvals."
- C. "What do other companies do for approvals?"
15. Which situation most clearly calls for escalating to a technical colleague rather than solving it yourself?
- B. A prompt that needs better structure
- A. A request to connect Claude directly to the company's internal order database through an integration
- C. An output with the wrong tone
- D. A Project that needs its knowledge updated
Answer key
| # | Answer | Domain | Why |
|---|---|---|---|
| 1 | C | Output Evaluation and Validation | Named sources can be fabricated; verification against the actual documents is the diligence step. Self-checking (B) is not independent evidence, and a disclaimer (D) does not make content correct |
| 2 | D | Prompting and Task Execution | Role, audience, explicit structure, and length constraints are what turn a vague request into a usable draft. Intensifiers like "detailed and thorough" (C) add no structure |
| 3 | A | Product and Model Selection | Match the model tier to the work: fast and cheap for formulaic volume, capable for exceptions |
| 4 | B | Configuration and Knowledge Management | Project knowledge is maintained by you; stale documents produce confidently outdated answers |
| 5 | C | Governance, Risk, and Responsible Use | Make the task compliant, then do it. A requester's seniority does not override policy, and instructing the model to ignore data does not un-send it |
| 6 | D | Prompting and Task Execution | Format failures are fixed by stating the format, not by re-rolling or upgrading the model |
| 7 | A | Output Evaluation and Validation | Specific, consequential details are the highest-risk surface for fabrication. Length, speed, and vocabulary carry no accuracy signal |
| 8 | B | Workflow Integration and Solution Design | The valued skill is a realistic division of labor with a review step, not a yes or a no |
| 9 | C | Configuration and Knowledge Management | Durable, shared conventions belong in instructions; per-message specifics belong in the prompt |
| 10 | D | Output Evaluation and Validation | Omission is as damaging as fabrication and is only visible if you read the source, not just the output |
| 11 | A | Troubleshooting and Optimization | Diagnose the change that coincided with the regression before rewriting anything |
| 12 | B | Workflow Integration and Solution Design | Repeated work with stable context and structure is exactly what Projects are for |
| 13 | C | Output Evaluation and Validation | Verification of consequential claims plus human review for customer-facing material. Model size is not a control |
| 14 | D | Workflow Integration and Solution Design | Process redesign requires giving Claude the current steps, the pain points, and asking for alternatives with tradeoffs |
| 15 | A | Governance, Risk, and Responsible Use | System integrations belong to the Developer and Architect scope; recognizing that boundary is an Associate competency |
Interpreting your score
| Raw score | Reading |
|---|---|
| 13 to 15 | Comfortable. Review anything you missed and book the exam |
| 10 to 12 | Close. Work the domains where you lost points, then retake this and the practice questions |
| 7 to 9 | More study needed. Return to the study guide, notes, and the prep courses before testing |
| 6 or fewer | Start from the blueprint and the official exam guide; you are not yet reading questions the way the exam intends |
This scale is a study aid, not a predictor. The real exam reports a scaled score from 100 to 1,000 with 720 to pass, and it draws from a much larger item pool.
Tally your misses by domain using the key above; two or more misses in one domain marks it as your next study target, exactly as the real score report's percent-correct breakdown would.
Written by the maintainer for self-assessment. Study guide · Study notes · Mock 2 · Mock 3 · Practice questions · Repository index