Claude Certified Associate - Foundations (CCAO-F) Exam Questions

Domain 1: Prompting and Task Execution (14%)

A coordinator reuses a saved prompt every month to draft a route-change notice for drivers. This month’s draft is too formal and leaves out the effective date. The prompt says nothing about tone or the date. What should she do?

⬜ Add the tone and effective-date requirements to the saved prompt, then run it again.
⬜ Fix the tone and add the date to this month’s draft by hand before sending it.
⬜ Reply in the chat and ask Claude to soften the tone and add the effective date.
⬜ Ask Claude for three alternative drafts and choose the one with the best tone.

A grants officer pastes a 50-page funding agreement into a chat and needs a complete list of every reporting deadline and penalty clause. Which prompt structure works best?

⬜ Paste the full agreement first, then give the task and output format after it.
⬜ Give the task and output format first, then paste the full agreement below it.
⬜ Ask for a general summary first, then request the deadlines in a follow-up.
⬜ Paste only the sections that look relevant to deadlines.

An administrator is building an accreditation submission with Claude. She has defined the audience and required sections, and Claude has just extracted compliance findings from three inspection reports. What should she do next?

⬜ Check the extracted findings against the three inspection reports.
⬜ Have Claude draft the full submission from the extracted findings.
⬜ Define the submission’s audience and required sections.
⬜ Ask Claude to rewrite the findings in the accreditor’s preferred terminology.

A coordinator needs conference theme ideas before the committee has formed any preferences, and a comparison of two catering bids against a fixed budget and set dietary rules. Which two prompt choices fit these tasks? (select two)

⬜ For the bids, state the budget and dietary rules Claude should judge against.
⬜ For the themes, ask for many varied ideas the committee can narrow down later.
⬜ For the themes, ask Claude to rank the ideas against the committee’s priorities.
⬜ For the bids, ask Claude to propose its own comparison criteria after reading both.

A writer’s property listing prompt already states the audience, length, tone and format. The drafts meet all of these but still miss the agency’s house style, which the writer can’t describe in words. What should the writer add?

⬜ A few of the agency’s published listings, pasted into the prompt as examples
⬜ More adjectives describing the desired tone
⬜ A role statement saying Claude is an expert real-estate copywriter
⬜ An instruction to check each draft against the stated requirements and revise

An event planner sends Claude one prompt: check her connected calendar for Thursday’s walkthrough, find the venue manager’s latest email, and draft a reply confirming the times. Each tool and target is named precisely. The draft quotes times that aren’t in the email. What should she change?

⬜ Run the three steps as separate requests, one after another.
⬜ Turn on thinking and resend the same prompt.
⬜ Rewrite the prompt to name the calendar and email tools more exactly.
⬜ Switch to a more capable model and resend the prompt.

A building manager asks Claude for a one-time elevator outage notice. His prompt specifies under 100 words and plain language, but the draft is 300 words and full of technical terms. What should he do next?

⬜ Give targeted feedback in the current chat and ask for a revised draft.
⬜ Shorten and simplify the draft himself before sending it.
⬜ Start a new conversation and rewrite the original prompt.
⬜ Create a Project and add the two requirements to its instructions.

A museum staff member drafts a welcome letter for new members. Her prompt dictates the exact wording of several sentences, and the result reads stiff and disjointed. What should she change next time?

⬜ Specify the audience, tone and length, and let Claude choose the phrasing.
⬜ Give only the goal and ask for a wide range of possible letters.
⬜ Ask Claude to critique the draft against the prompt and then redraft it.
⬜ Keep the required sentences and ask Claude to smooth the transitions.

An event organizer asks Claude to “draft a sponsorship proposal for our conference.” The two-page draft reads well but could describe almost any event. The organizer’s planning file already has the audience profile and the sponsor package details. Which two prompt changes will most improve the next draft? (select two)

⬜ Add details about the event, its audience and the sponsor packages.
⬜ Ask for each section of the proposal in a separate request.
⬜ State the required sections, length and tone of the proposal.
⬜ Turn on web search so Claude can research sponsorship rates at similar events.

A benefits coordinator’s one-paragraph prompt asks Claude to summarize a policy change, list the affected employee groups, state the effective date and draft an FAQ entry. Responses often leave out one of these. Which prompt change most directly fixes this?

⬜ Rewrite the request as a numbered list of the required items.
⬜ Paste a past benefits announcement as an example of the expected structure.
⬜ Add a closing line telling Claude not to leave out any part of the request.
⬜ Add a role line telling Claude to act as an HR communications specialist.

A grants officer must summarize the funding priorities that three target foundations announced in recent weeks. The director will check each claim. Which two prompt choices fit? (select two)

⬜ Turn on web search and ask for a citation for each claim.
⬜ State the question and name the three foundations in scope.
⬜ Ask first for a broad scan of the wider foundation funding landscape.
⬜ Have Claude work only from the office’s past grant application files.

A manager is drafting a four-part training guide with Claude, one request per part: role overview, procedures, troubleshooting FAQ and quick reference. The FAQ and the quick reference must match the procedures. Which part should she draft first?

⬜ The procedures, then build the FAQ and quick reference from them.
⬜ The quick reference, then use it as an outline for the other parts.
⬜ The role overview, then continue in the guide’s reading order.
⬜ All four parts at once in separate chats, then merge the drafts.

A procurement officer reuses one prompt every month to summarize vendor performance. She changed its context, tone and format in a single edit. The new draft has the right tone but new formatting problems. What should she do next?

⬜ Go back to the original prompt and test one change per round.
⬜ Ask Claude which of her edits helped the draft and which hurt it.
⬜ Change the context, tone and format again in another single edit.
⬜ Keep the new draft and fix the formatting problems with follow-up feedback.

A program coordinator must turn 40 pages of visitor survey comments into a one-page recommendations memo for the director, working through the task with Claude in several steps. Which step should come first?

⬜ Define the memo’s audience, purpose and the questions it must answer.
⬜ Ask Claude to draft the complete one-page memo in one request.
⬜ Upload the comments and have Claude pull out the recurring themes.
⬜ Check Claude’s themes against the original survey comments.

A manager asks Claude to score vendor proposals against four criteria. For proposals with no pricing, Claude guessed the numbers. The manager plans to ask those vendors for their pricing. Which rule should the next prompt include?

⬜ Flag proposals with missing pricing and leave them unscored.
⬜ Fill in missing pricing with the average price of the other proposals.
⬜ Score only the non-price criteria when pricing is missing.
⬜ Cite the page of the proposal where each price appears.

An associate plans a one-off task with Claude: summarize three reports, compare the summaries, draft a recommendation from the comparison, then refine its tone. Each step uses the previous step’s output. How should she run the steps?

⬜ One at a time in a single conversation.
⬜ Each in its own chat, so every request stays focused.
⬜ All together as one request with thinking turned on.
⬜ In a Project, with the four steps in its instructions and each step in its own chat.

A museum curator asks Claude to summarize visitor comments. The draft is accurate but written as paragraphs, and she needs a table for a staff meeting. Her prompt included only the comments and the task. What should she change?

⬜ Add a format requirement asking for a table with named columns.
⬜ Add background about the museum, its exhibits and its visitor programs.
⬜ Split the request into separate prompts, one per comment theme.
⬜ State the staff-meeting audience and the tone she wants.

A hotel manager asks Claude to turn a 12-page staff handbook into plain language. He pastes one section per message, repeating the same detailed tone and terminology instructions each time, but the edited sections still vary in style. What should he change?

⬜ Send the whole handbook in one message with the instructions.
⬜ Use a more capable Claude model for each section.
⬜ Ask for three tone versions of each section and pick the most consistent.
⬜ Write a plain-language style guide and keep editing one section per message.

An associate’s prompt combines three parts: background notes about a customer account, the task instructions and a pasted customer email to answer. Claude sometimes replies to questions in the background notes instead of the email. Which change most directly fixes this?

⬜ Separate the three parts with clear, descriptive labels or tags.
⬜ Add a sample reply as an example of the tone and format.
⬜ Give Claude the role of an experienced support agent.
⬜ Move the customer email to the top, before the instructions.

A food bank coordinator pastes a volunteer orientation guide and asks Claude to cut it to one page. The result drops the food-safety steps but keeps the welcome message. Which prompt change most directly fixes this?

⬜ Explain the purpose of the page and what new volunteers need most from it.
⬜ Move the pasted guide above the instructions.
⬜ Attach a one-page guide from another program as a format example.
⬜ Replace the one-page limit with an exact word count.

A recruiter uses Claude’s Google Drive and email connectors to prepare a summary for each candidate. She already splits the work into separate requests, but each one is vague, such as “find the candidate’s file,” and Claude keeps retrieving the wrong documents. What should she change?

⬜ Name the connector and the exact file or folder in each request.
⬜ Turn on thinking before running the retrieval requests.
⬜ Ask Claude to list the files it retrieved before drafting.
⬜ Combine the steps into one detailed prompt that states the full goal.

A policy officer at a regional utility asks Claude to “review this rate-change proposal” and gets generic, mildly positive feedback. What she actually needs is to anticipate objections from a skeptical review board. Which prompt revision helps most?

⬜ Ask Claude to review the proposal from the skeptical board’s perspective and list the objections its members would raise.
⬜ Tell Claude it’s the world’s most experienced utility policy expert and ask it to review again.
⬜ Send the same review request in a new conversation for a fresh look.
⬜ Ask Claude to be much more critical and thorough this time.

After learning about task decomposition, an associate now splits every request into several prompts, even a two-sentence internal announcement, and the team’s turnaround has slowed. Which rule should the associate use instead?

⬜ Break a task into steps when it has several parts and intermediate results need checking before later steps build on them.
⬜ Break a task into steps whenever at least three prompts can be written.
⬜ Break a task into steps only after a single prompt has already failed.
⬜ Break a task into steps whenever the finished deliverable is longer than about a page.

A training coordinator’s prompt for a workshop recap explicitly asks for a five-bullet action list, but Claude’s draft buries the action items in long paragraphs. What’s the most effective way to iterate?

⬜ Point out in the same conversation that the actions must be a five-bullet list and ask for a revision.
⬜ Rewrite the whole prompt from scratch, since an ignored instruction shows the prompt was flawed.
⬜ Start a new conversation and send the same prompt again.
⬜ Reply that the draft should be better overall and let Claude decide what to improve.

An associate asks Claude to compare three vendor quotes and gets back a bland restatement of what each quote says. Which two prompt additions most improve the result? (select two)

⬜ State the decision the comparison will inform, such as which vendor to shortlist.
⬜ Tell Claude to act as a procurement expert with decades of experience.
⬜ List the criteria to compare, such as total cost, support terms and contract length.
⬜ Ask for a much longer, more thorough version of the same comparison.

An analyst revising a Claude-drafted policy memo finds that each revision fixes the issue named in the feedback but drops a formatting requirement that earlier rounds had met. How should the analyst iterate?

⬜ Say which requirements must stay unchanged in each revision request, then check the new draft against all requirements.
⬜ Treat the dropped formatting as normal variation and restore it by hand after each round.
⬜ Start a new conversation for each round so earlier feedback can’t interfere.
⬜ Combine all earlier feedback into one broad request to improve the draft overall.

A team used a deliberately loose prompt to get 25 wide-ranging tagline ideas from Claude. It now needs to narrow them to the three strongest. How should the prompting change for this step?

⬜ Tighten it: state the selection criteria and goal, then ask Claude to judge the ideas against them.
⬜ Ask Claude to name its three favorites and use those.
⬜ Keep the prompt as loose as before so the creative range carries over.
⬜ Ask for another 25 ideas so there’s a bigger pool to choose from.

A membership adviser’s prompt keeps producing generic renewal messages. Which revision best applies the AI Fluency competency of Description?

⬜ Ask Claude for a warmer, more persuasive renewal message than the last one.
⬜ Tell Claude it is an experienced membership adviser and ask it to write from that perspective.
⬜ Spell out the task’s scope, the intended recipients, the output format and the constraints, then judge the draft against that description.
⬜ Paste three renewal messages from other organizations and ask Claude to blend their styles.

A consultant has two days to write a due-diligence brief from finance records, customer interviews, operating procedures and current market evidence, and every finding must trace back to its source. The planning steps are: (1) review each workstream’s findings for conflicts and gaps; (2) run a focused Claude analysis on each source set; (3) define the decision questions and required sections; (4) combine the findings into the brief and check every conclusion cites its sources; (5) split the sources into workstreams with defined outputs. What’s the right order?

⬜ 3, 5, 2, 1, 4
⬜ 5, 2, 3, 4, 1
⬜ 2, 5, 1, 3, 4
⬜ 3, 2, 4, 1, 5

An HR specialist is turning 60 job descriptions into one-page summaries for a talent portal that requires five fixed headings and approved job-family terms. The first ten summaries vary in heading order, detail and terminology. Which change will most improve consistency?

⬜ Send all 60 descriptions in one long request so Claude uses the same judgment throughout.
⬜ Let Claude summarize freely, then run a second pass that maps its headings onto the portal’s five.
⬜ Add an instruction to keep each new summary consistent with the earlier ones in the chat.
⬜ State the five headings, the approved terms and a length limit, and include one approved example.

A marketing team wants original campaign concepts that stick to approved product claims and required legal wording. The last batch was creative but used two claims that legal had rejected. Which approach best balances creativity and compliance?

⬜ Ask Claude to model the concepts closely on past approved campaigns so they reuse accepted language.
⬜ Give Claude the approved claims and prohibited wording as constraints, generate several concepts, then review them for both creativity and compliance.
⬜ Leave the product claims out of the prompt so the ideas aren’t limited by existing language.
⬜ Ask for the most creative concepts with no constraints, and have legal rewrite the chosen one.

A team asked Claude to redesign a returns process covering intake, eligibility checks, warehouse routing, refunds and exceptions. The first draft mixed up responsibilities between stages and left out approval points. The steps are: (1) ask Claude to draft the procedure from the completed stage map; (2) mark which decisions at each stage need approval, and from whom; (3) list the stages from intake to exceptions; (4) have the stage owners validate the handoffs in the draft; (5) assign an owner and required inputs to each stage. What’s the right order?

⬜ 1, 3, 5, 2, 4
⬜ 3, 2, 5, 4, 1
⬜ 3, 5, 2, 1, 4
⬜ 5, 3, 1, 2, 4

A procurement team is using Claude to evaluate three translation vendors and needs a ranking it can defend. The steps are: (1) collect information from each vendor; (2) rank the vendors and explain any remaining limits; (3) define the criteria and how each will be judged; (4) score each vendor against the criteria; (5) check the vendor information for gaps or details that can’t be compared. What’s the right order?

⬜ 3, 1, 5, 4, 2
⬜ 1, 3, 4, 5, 2
⬜ 3, 1, 4, 2, 5
⬜ 1, 5, 3, 4, 2

An HR team is drafting guidance for managers who respond to workplace concerns. It must follow approved policy, acknowledge facts that are still uncertain and give managers practical ways to approach each conversation. Which two prompting choices fit? (select two)

⬜ Provide the approved policy excerpts and have Claude separate policy requirements from optional communication suggestions.
⬜ Ask for one standard script every manager uses word for word.
⬜ Ask Claude to flag missing facts and offer approaches for different versions of events.
⬜ Tell Claude to settle the disputed facts first so managers have one clear account.

A team has refined a policy brief over many rounds in one chat. Claude still follows an early instruction to write for executives, but the team has since decided the audience is frontline managers. The evidence analysis is strong and should stay. What’s the best next step?

⬜ Turn on memory so Claude can work out which audience instruction is newest.
⬜ Start a fresh chat and rebuild the brief from Claude’s general knowledge.
⬜ Say the executive-audience instruction no longer applies, describe the frontline managers and ask for a revision that keeps the evidence.
⬜ Ask Claude to align the brief more closely with what stakeholders now need.

A conference coordinator uses one unchanged prompt template for naming sessions, analyzing survey results and drafting policy summaries. What feedback best addresses this habit?

⬜ Keep the template and add a line at the top naming the task type.
⬜ Shape each prompt to the task’s purpose, evidence and output form, reusing template parts only where they fit.
⬜ Keep the template but use a more capable model for analysis and policy work.
⬜ Build one master template with instructions for every task and send it each time.

A project manager asked Claude to evaluate six logistics bids. Claude mixed mandatory eligibility failures into its preference scores, so a bid that failed a requirement still ranked second. The revised steps are: (1) record the recommendation and rationale for approval; (2) screen every bid against the mandatory requirements; (3) review trade-offs among the top eligible bids; (4) remove bids that fail any mandatory requirement; (5) score the remaining bids against the weighted preferences. What’s the right order?

⬜ 2, 5, 4, 3, 1
⬜ 2, 4, 5, 3, 1
⬜ 5, 2, 4, 3, 1
⬜ 2, 4, 3, 5, 1

A theatre director’s note just says, “Help us with next season.” You’re turning it into a structured prompt for a planning brief. The steps are: (1) add the relevant facts and constraints; (2) send the prompt; (3) clarify the specific task behind the note; (4) set the length, tone and format; (5) confirm who will read the brief and what decision it supports. What’s the right order?

⬜ 5, 3, 1, 4, 2
⬜ 3, 1, 5, 4, 2
⬜ 3, 5, 1, 4, 2
⬜ 3, 5, 4, 2, 1

An associate asked Claude to “summarize the attached policy changes for executives.” The draft is accurate but technical, twice as long as executives will read and doesn’t show which passages support each point. Which revised prompt is best?

⬜ Ask for a shorter, simpler summary and let Claude pick what matters.
⬜ State the audience, a length limit and the decisions needed, put the documents in tags and ask for supporting quotes for each point.
⬜ Ask for more detail on every change so executives can skip what they don’t need.
⬜ Start a new chat and let Claude infer the audience from the document title.

A charity adviser split a funding recommendation into eighteen tiny prompts. The combined draft feels fragmented because closely related evidence and implications ended up in separate sections. What’s the best adjustment?

⬜ Keep the eighteen prompts and add a final one asking Claude to improve the flow.
⬜ Keep the structure and give each prompt the full text of all earlier sections.
⬜ Combine related tasks into fewer, cohesive sections that keep connected evidence and implications together.
⬜ Write a strong introduction and conclusion to tie the sections together.

An analyst asks Claude for three market scenarios from uncertain, incomplete evidence. Executives want alternatives for contingency planning, not one forecast, and need to see what could push the market toward each outcome. Which approach fits best?

⬜ Ask for one definitive forecast without the assumptions.
⬜ Ask for optimistic, baseline and adverse scenarios, each with its assumptions, evidence, triggers and confidence limits.
⬜ Encourage imaginative scenarios free of any tie to the evidence.
⬜ Copy current conditions into three scenarios and change only the headings.

An associate needs a 220-word quarterly update for branch managers in five districts. The approved key messages and district details are ready, and an older newsletter is useful only as a tone example. Which prompt gives the strongest first-draft specification?

⬜ State the audience, five-district coverage, the 220-word limit, the tone and the current messages, and label the old newsletter as a tone example only.
⬜ Attach the last three newsletters and ask Claude to follow their typical length and structure.
⬜ Attach the old newsletter with the new messages and ask for something similar.
⬜ Use the old newsletter’s tone and structure as the overall specification.

A Claude-drafted visitor guide is accurate but uses several unofficial terms instead of the museum’s approved glossary. Which refinement fixes this most directly?

⬜ Tell Claude to follow the approved glossary, naming it and its version.
⬜ Switch to a more capable model that knows specialist museum terms.
⬜ Provide the current glossary and instruct Claude to use its terms throughout.
⬜ Ask Claude for more formal, professional wording.

A vague request to “look at coworking membership prices” turns out to mean the team needs a structured, sourced summary of five competing providers, not ideas for its own pricing. Which prompting approach fits?

⬜ Research the five providers using the same pricing fields and source references, and mark any detail that can’t be verified.
⬜ Discuss prices in an open-ended chat and let the structure emerge.
⬜ Ask Claude to recommend a price for the team’s own space.
⬜ Ask Claude to list the providers’ prices from what it already knows.

In one request, a team asked Claude to analyze 240 customer comments, build an issue taxonomy, count each issue and recommend process changes. The categories overlap and the counts don’t add up. Which two actions should the team take? (select two)

⬜ Define and review the taxonomy first, before Claude classifies all 240 comments against it.
⬜ Repeat the same request, adding that every count must be accurate.
⬜ Check the classifications and totals against the comments before asking for recommendations.
⬜ Analyze a convenient sample and treat its counts as representative.

An analyst must explain a drop in customer renewals by listing possible causes, testing them against eighteen months of account data and twelve interview transcripts, and drafting a short recommendation. Which prompting approach fits best?

⬜ Ask for the single most likely cause first, then organize the evidence around it.
⬜ Brainstorm causes broadly, test each against the data and interviews, then draft from the ones that hold up.
⬜ Draft the recommendation first, then find evidence for it.
⬜ Ask for a customer story about renewals and base the recommendation on it.

You’re using Claude to map a first-time subscriber’s experience with a community meal-delivery service. The steps are: (1) capture pain points and improvement ideas; (2) identify the touchpoints the subscriber meets; (3) define the subscriber persona and the boundaries of the journey; (4) combine the findings into the final journey map; (5) identify the major stages of the journey. What’s the right order?

⬜ 5, 3, 2, 1, 4
⬜ 3, 5, 2, 1, 4
⬜ 3, 2, 5, 1, 4
⬜ 3, 5, 1, 2, 4

---

Domain 2: Output Evaluation and Validation (21%)

A communications manager uses Claude with web search to research a competitor’s product recall for a press statement. The citations link to news articles and the competitor’s announcement, and a legal review is already scheduled. What should the manager do first?

⬜ Open each cited source and check that the claims match it.
⬜ Send the gathered facts to the legal team for review now.
⬜ Check every claim against only the competitor’s official announcement.
⬜ Ask Claude to quote the sentence from each source that supports each claim.

An event planner using Claude with no connectors enabled asks for a reminder email to attendees. Claude writes the draft and then says it has sent the email to everyone. What should the planner do?

⬜ Treat the “sent” claim as false and email the attendees from her own account.
⬜ Ask Claude to resend the reminder and confirm the delivery time.
⬜ Ask IT to investigate why attendees never received Claude’s email.
⬜ Ask several attendees to confirm the reminder reached their inbox.

A safety analyst’s incident report for engineers shows a 12% rise in minor injuries and notes that night-shift data is missing. A plain-language version is now needed for frontline staff, including night-shift workers. Which change is appropriate?

⬜ Simplify the wording and length, but keep the 12% figure and the missing-data note.
⬜ Replace the 12% figure with “a noticeable rise” to suit the plain style.
⬜ Keep the 12% figure but remove the missing-data note from the staff version.
⬜ Open the staff version with the data collection method so readers can judge the figure.

A grants coordinator asks Claude to list every deadline in a 25-page grant agreement, and Claude returns eight. Before adding them to the team calendar, which check confirms the list is complete?

⬜ Search the agreement for every date or deadline and confirm each one appears in Claude’s list.
⬜ Ask a colleague to read the eight deadlines and flag any that look wrong.
⬜ Run the same extraction in a new chat and confirm the lists match.
⬜ Check that each listed deadline matches the agreement word for word.

A long, methods-heavy analysis written for data specialists must now go to executives deciding whether to fund the project. Which two changes adapt it for that audience? (select two)

⬜ Move the recommendation, the key number and the main risk to the top.
⬜ Cut the step-by-step description of the analysis method.
⬜ Add the complete list of modeling assumptions behind the projections.
⬜ Rewrite every step of the method in plain, non-technical language.

A procurement officer asks Claude to compare five software vendors on cost, support and security. Claude answers in long paragraphs. The officer won’t share or reuse the text but struggles to compare vendors on each criterion. What should she request next?

⬜ Reformat the comparison as a table of vendors and criteria.
⬜ Redo the comparison one vendor at a time in separate messages.
⬜ Shorten each vendor’s section to three sentences.
⬜ Move the comparison into an artifact to refine and share.

An associate drafted, with Claude, an email to 40,000 customers announcing updated subscription terms. It quotes a cancellation deadline from the terms of service and can’t be recalled once sent. The tone already matches the company voice and there are no pricing figures. Which two steps are required before it goes out? (select two)

⬜ Have qualified legal counsel review the terms-change wording.
⬜ Verify the quoted cancellation deadline against the terms of service.
⬜ Rewrite the tone to sound warmer and more personal.
⬜ Send the draft to the finance team for a pre-send review.

A communications officer attaches a nonprofit’s final published annual report and asks Claude for three financial figures for a press release. How should the officer validate the figures?

⬜ Check each figure against the page of the report it came from.
⬜ Turn on web search and compare each figure with recent news coverage.
⬜ Use Research to check each figure across external sources.
⬜ Ask the finance team to confirm the figures from internal records.

A clinic office manager with no clinical training uses Claude to draft a staff parking note and to adapt a physician’s aftercare handout for patients. Claude simplified the wording and reordered care steps. What review does each draft need?

⬜ She checks the parking note herself; the physician reviews the aftercare instructions.
⬜ She checks the parking note; she asks Claude to list every change made to the handout.
⬜ The physician reviews both drafts before either is sent.
⬜ She checks the parking note and compares the aftercare draft with the original herself.

Without searching the web, Claude summarizes an import rule that changed last month, for a client advisory. The regulator has published the updated rule text online. What should the associate do before sending the advisory?

⬜ Compare the summary against the regulator’s published rule text.
⬜ Break the request into one question per section and ask again.
⬜ Ask Claude to search the web and re-check its summary with citations.
⬜ Read the summary closely for internal contradictions.

A team drafts two kinds of documents with Claude. Its internal recaps are easy to correct after they go out. Its customer notices are audited by regulators and can’t be retracted once sent. Which two rules should decide when a draft goes to a qualified reviewer? (select two)

⬜ Require reviewer sign-off when the draft covers regulated or legally binding content.
⬜ Require reviewer sign-off when the draft supports a personnel or hiring decision.
⬜ Require reviewer sign-off when a mistake in the draft would be hard to reverse.
⬜ Require reviewer sign-off when the draft will be issued as a public press release.

A housing agency employee attaches a rental assistance handbook and asks Claude to answer applicant questions from it. Several answers include rules that aren’t in the handbook. Which prompt change best addresses this?

⬜ Tell Claude to answer only from the attached handbook and to say when it doesn’t know.
⬜ Turn on web search so Claude can check each answer against current online information.
⬜ Ask Claude to rate its confidence in each answer and flag the low ones.
⬜ Use Research to gather more sources on assistance rules before answering.

The policy owner approved a fact-checked summary of an updated travel expense policy and asked an associate to turn it into plain language for all staff. Claude’s rewrite reads well. What should the associate do before sending it?

⬜ Compare the rewrite with the approved summary to check that no detail changed.
⬜ Send the rewrite to the policy owner for another round of content approval.
⬜ Ask Claude to confirm that the rewrite kept every detail of the approved policy.
⬜ Test the rewrite on a few staff readers to confirm the language is clear.

An associate asks Claude for a quick definition to read once, and for exact totals from an attached expense sheet. Which two steps fit these needs? (select two)

⬜ Ask Claude to show the totals in a table within its reply.
⬜ Save the definition as an artifact so it can be refined and shared later.
⬜ Have Claude calculate the totals with code execution instead of writing them as text.
⬜ Keep the definition in the chat instead of creating an artifact.

An event planner has finished a long Claude conversation about venues, catering and budget, in which many options were raised and rejected. Two coworkers need only the final choices so they can start booking next week. What should the planner share?

⬜ A short summary of the confirmed choices that Claude pulls together.
⬜ The conversation link, so the coworkers can read the whole discussion.
⬜ A comparison table of every venue and catering option discussed.
⬜ A decision log of each option considered and why it was rejected.

An associate uploads a 350-page inspection report as one PDF and asks Claude to summarize the findings. Several key findings appear only in charts, and the summary leaves them out. What should the associate do?

⬜ Split the PDF into parts of 100 pages or fewer and ask for the summary again.
⬜ Upload the same file in a new chat and request the summary again.
⬜ Reply with feedback asking Claude to add the chart findings to the summary.
⬜ Turn on web search so Claude can find the figures shown in the charts.

A manager has two Claude-drafted versions of a memo announcing a shift-schedule change that staff haven’t been told about yet. Both drafts read well. How should the manager choose which one to send?

⬜ List what the memo must achieve and judge both drafts against that list.
⬜ Send both drafts to a few staff members and use the one they prefer.
⬜ Ask Claude to review both drafts and send the one it rates higher.
⬜ Combine the strongest paragraphs from both drafts into one longer memo.

A grant writer’s prompt left out the funder’s two-page limit. Claude’s draft has verified figures, runs to four pages, and doesn’t state the funding request until the last page. Which two actions best refine the draft? (select two)

⬜ Run the same prompt again with a more capable Claude model.
⬜ Give Claude the two-page limit and ask for a rewrite within it.
⬜ Check every figure in the draft against the approved budget before editing.
⬜ Tell Claude to move the funding request into the opening paragraph.

A travel agency associate asks Claude to draft website copy about the typical customers for each tour. The draft presents adventure tours as for young men and river cruises as for retired women. What should the associate do?

⬜ Revise the copy to describe travelers’ interests and activities instead of age and gender.
⬜ Check whether the age and gender patterns match the agency’s booking records first.
⬜ Send the draft to legal counsel for review before it’s published.
⬜ Ask Claude to rewrite the copy in a livelier tone.

An associate drafts a briefing with Claude that contains a dozen factual claims about developments from the past few months, across several topics. The associate’s paid plan has web search turned on. Which approach fits this fact-checking task best?

⬜ Use Research on the claims, then review the cited sources.
⬜ Ask Claude to search each claim one at a time and read each source.
⬜ Give Claude one recent news article to compare against.
⬜ Check each claim against last year’s industry report on file.

An office manager drafts a notice about the city’s recycling rules with Claude. Web search is on, and the manager has the link to the city’s current official rules page. How should she check the draft’s claims?

⬜ Paste the link into the chat and ask Claude to check the draft against that page.
⬜ Ask Claude to review the draft using its own knowledge of local recycling rules.
⬜ Ask Claude to search the web for the rules without sharing the link.
⬜ Ask Claude whether the draft is accurate and keep it if Claude agrees.

An associate used Claude to analyze three relocation options and recommend one. The finished analysis goes to managers who must approve the recommendation but weren’t involved in the work. How should the document be organized?

⬜ Recommendation and key reasons first, then supporting detail under clear headings.
⬜ As an artifact the managers can revise together.
⬜ A brief note with only the recommendation, and the full analysis on request.
⬜ A comparison table of the three options for the managers to weigh.

An analyst used Claude to turn a forecasting study into a one-page executive summary that leaves out the method. A panel of statisticians must now review whether the study’s approach is sound. Which adaptation is appropriate?

⬜ Ask Claude to expand the summary with the method and the assumptions behind the findings.
⬜ Send the panel the executive summary unchanged.
⬜ Ask Claude to rewrite the summary in plainer language with every term defined.
⬜ Ask Claude to shorten it to the key figures and the final forecast.

A dispatcher drafts a customer reply template with Claude. It states what customers can claim under the company’s damage policy, and the team will send it many times a week. What should happen before it’s adopted?

⬜ The owner of the damage policy reviews the template once before adoption.
⬜ Each team member checks every reply they send against the policy.
⬜ It’s sent to a few customers first and refined from their replies.
⬜ The dispatcher reviews it for tone and clarity, then shares it.

An office manager asks Claude to summarize an expense spreadsheet for the owner’s quarterly report. The summary states total spending of $84,000, but its five category figures add up to $79,500. What should the manager do?

⬜ Check every figure against the source spreadsheet and correct the summary.
⬜ Ask the bookkeeper to confirm the quarterly total before sending.
⬜ Ask Claude to recheck its arithmetic and use the numbers it confirms.
⬜ Replace the total with the sum of the five category figures.

Claude drafts a competitor briefing with web search off. It cites revenue from last year’s annual report, which the associate has, and describes an acquisition from two months ago. Which two checks should happen before sharing? (select two)

⬜ Send the briefing to the legal team for review.
⬜ Check the acquisition claim against current news coverage.
⬜ Compare the revenue figure with the annual report the associate has.
⬜ Confirm the acquisition claim using the same annual report.

With web search off, a grants writer asks Claude for studies linking museum visits to learning. Claude lists five references with links, and two links don’t load. Grant rules allow only citations confirmed in an academic database. What should the writer do?

⬜ Look up all five references in the academic database.
⬜ Look up the two broken ones and keep the three whose links load.
⬜ Cite only the studies whose links load and match their references.
⬜ Turn on web search and ask Claude to check each reference online.

An associate checks every factual claim in a Claude-drafted vendor risk assessment against the source documents and finds each one accurate. What still has to be checked before the assessment is relied on?

⬜ That no material risks were left out, by comparing the draft with the sources and the original requirements.
⬜ That a fresh conversation produces the same assessment again.
⬜ That Claude confirms the assessment covers everything important.
⬜ Nothing more, since verifying every claim in the draft makes it reliable.

Which two statements show sound judgment when reviewing Claude’s responses for invented content, inconsistencies or bias? (select two)

⬜ Biased wording can only appear if the prompt itself contained biased instructions.
⬜ A response with formal citations is dependable, because Claude only cites sources it actually consulted.
⬜ An authoritative-sounding quote from a named report stays unverified until it’s found in the actual source.
⬜ A contradiction between a response’s conclusion and one of its own details means at least one claim is wrong.

An associate drafting a grant proposal with Claude includes national charitable-giving statistics. To fact-check them, she pastes the draft into a new Claude conversation, which confirms every figure. What should she conclude?

⬜ The confirmation proves nothing; check the figures against the original published reports or another authoritative source.
⬜ Run the same check in a third conversation and accept the figures if they agree.
⬜ The figures are verified, because the new conversation reviewed them independently.
⬜ Ask the new conversation to rate its confidence and check only the uncertain figures.

An associate has a Claude-drafted response to a regulator’s information request that is due today. The designated compliance reviewer is away until tomorrow, and the draft reads well. What should the associate do?

⬜ Find another qualified reviewer or ask for a deadline extension.
⬜ Ask Claude to check the draft against the request and submit if it confirms.
⬜ Submit today and have the compliance reviewer check it tomorrow, correcting anything afterward.
⬜ Submit after a careful personal read-through, given the time pressure and draft quality.

An associate attaches a supplier’s 12-page audit report and asks Claude to summarize it. The summary includes a specific compliance score that appears nowhere in the report. What should the associate conclude?

⬜ The score isn’t supported by the source; remove it or verify it independently before using the summary.
⬜ Keep the score with a note that it’s AI-generated so readers can judge it.
⬜ Assume the score was worked out from other data in the report.
⬜ Ask Claude whether the score is in the report and keep it if Claude stands by it.

Claude drafted a school’s parent newsletter, event checklist and volunteer FAQ as three artifacts, each updated by a different staff member on its own schedule. A colleague suggests merging them into one master document. What’s the sound decision?

⬜ Keep them as three separate artifacts, since each is a self-contained document revised on its own schedule.
⬜ Combine them into one document with a table of contents.
⬜ Move all three into the chat so edits happen in one thread.
⬜ Merge them into one master artifact so all the material is in one place.

A Claude-drafted troubleshooting guide had a light review and has been used internally for months without problems. The team now wants to post it on the public customer help site. What decides the review it needs?

⬜ It needs full verification first, because public release is a higher-stakes use than the one it was reviewed for.
⬜ The same light review again, since the content hasn’t changed.
⬜ None, since months of internal use show it’s accurate.
⬜ Claude’s re-examination, publishing once Claude confirms the steps.

In a new chat with no Project, files, connectors or relevant memory, an associate asks Claude how the company’s private inventory system calculates reorder alerts. Claude gives confident, step-by-step details. Which two considerations should guide validation? (select two)

⬜ Ask Claude to explain again and accept it if the details stay the same.
⬜ Confirm the behavior against the system’s own documentation or the team that maintains it.
⬜ The specific, confident answer shows Claude really knows the system.
⬜ Claude has no access to the private system’s details, so the specifics are likely invented.

An associate attaches an interview transcript and asks Claude to pull out notable quotes for the company newsletter. Claude returns several quotes, each attributed by name. What must happen before they’re published?

⬜ Check each quote word for word against the transcript.
⬜ Ask Claude to confirm each quote is exact, since the transcript is in its context.
⬜ Shorten each quote to a short phrase to limit any mismatch.
⬜ Publish them as returned, since they came from the attached transcript.

An analyst is about to share a Claude-drafted tourism brief with a planning committee. Which action best applies the AI Fluency competency of Discernment?

⬜ Confirm that every citation points to a real, reputable source and that the links open.
⬜ Compare the brief’s conclusions with the analyst’s own sense of local tourism trends.
⬜ Review the whole brief against the task’s requirements, the source evidence and professional standards before deciding whether it’s fit to use.
⬜ Read the executive summary and recommendations closely, since those are what the committee acts on.

Claude summarizes 140 promotion comments for a calibration meeting and concludes that remote employees show weaker leadership. Six managers used different rubrics, and several remote staff got strong examples. Which two problems should the HR analyst raise? (select two)

⬜ The comments don’t establish that remote status causes weaker leadership.
⬜ The summary is dependable because it covers all 140 comments and cites one for each finding.
⬜ Ratings given against six different rubrics can’t be pooled until they’re normalized or separated.
⬜ The draft only needs regenerating with an instruction to stay neutral about work location.

Claude drafts a performance-improvement recommendation from manager notes, attendance records and employee feedback. It could lead to the employee’s dismissal, and two managers’ notes disagree about the employee’s conduct. Which review approach is most appropriate?

⬜ Have a second Claude chat check the recommendation for bias and approve it if it finds none.
⬜ Have an accountable HR reviewer check the evidence, the conflicting notes, policy alignment and fairness before any action.
⬜ Approve it once Claude cites every record it relied on.
⬜ Ask Claude to soften it to a formal warning and use that as the outcome.

Claude drafts a monthly status report for a museum renovation. The project charter defines scope and milestones, and the director has listed mandatory report fields and the decisions readers need to make. Which evaluation best shows whether the draft is ready?

⬜ Accept it when each section has at least one citation.
⬜ Check its claims against the charter, confirm each required field has real content, and judge whether it supports the director’s decisions.
⬜ Approve it if it uses the same template as last month’s report.
⬜ Ask Claude to check it against the charter and fields, and approve it if Claude finds no issues.

A Claude summary of a quarterly report says revenue rose, but the source table seems to show a fall, and other figures may be wrong too. The steps are: (1) correct every wrong statement; (2) recalculate the revenue change from the table; (3) send the corrected summary with a note of what changed; (4) find the revenue statement and the table cells it draws on; (5) check the other key figures against the report. What’s the right order?

⬜ 2, 4, 1, 5, 3
⬜ 4, 2, 1, 3, 5
⬜ 4, 2, 5, 1, 3
⬜ 5, 4, 2, 3, 1

Claude recommends merging four local packing stations into one central site. It cites the documents’ evidence on operating savings, but leaves out what the same documents say about equipment-transfer costs and longer collection journeys. What should the cooperative do?

⬜ Accept it, since the passages Claude used all support centralization accurately.
⬜ Approve it and add an appendix with the transfer costs and journey times.
⬜ Ask for a revised recommendation that weighs the transfer costs and longer journeys before relying on it.
⬜ Ask Claude for more evidence on savings and a more confident conclusion.

A colleague has questioned several statistics and references in a Claude-drafted comparison of delivery services. The review steps are: (1) record each unsupported claim, where it is and how much it matters; (2) read the whole draft and flag passages that need a closer look; (3) decide whether to correct, re-prompt or escalate; (4) list every statistic, percentage and reference the draft relies on; (5) cross-check those claims against the supplied documents or an authoritative source. What’s the right order?

⬜ 4, 5, 2, 1, 3
⬜ 2, 4, 5, 1, 3
⬜ 2, 5, 3, 4, 1
⬜ 5, 4, 1, 3, 2

A program office has 80 meeting notes full of decisions, owners, deadlines, open questions and repeated background discussion. Leads need a summary they can act on at Monday’s planning session. Which instruction works best?

⬜ Group the verified decisions, owners, deadlines and open questions into separate sections, and drop the repeated background.
⬜ Summarize each of the 80 notes in its original order so nothing is lost.
⬜ Write one narrative in date order that weaves decisions and deadlines into the background.
⬜ List only the final decisions, since owners and deadlines are still changing.

A program office gets three Claude-generated portfolio summaries: one concise but missing dependencies, one complete but disorganized, and one polished but with conclusions the data doesn’t support. Which two actions should guide the choice? (select two)

⬜ Pick the polished summary, since its confident style is most likely to be accepted.
⬜ Score each summary against explicit criteria for accuracy, completeness, organization and usefulness for decisions.
⬜ Merge the three sentence by sentence so each contributes equally.
⬜ Check the key claims in each against the portfolio data, then combine the strongest verified parts.

A team uses Research to judge whether a supplier market is consolidating. The draft relies on one trade association forecast, several recent acquisition announcements and an undated vendor presentation predicting more mergers. Which two checks most strengthen the conclusion? (select two)

⬜ Count every announced deal as equal evidence of consolidation.
⬜ Confirm each deal’s status and market scope from regulatory filings or other primary records.
⬜ Give the trade association forecast the most weight, since it speaks for the whole industry.
⬜ Compare the forecast with independent analyses and note differences in definitions, methods and assumptions.

Claude explains a newly revised safety standard for site staff training, but cites an old training manual and a forum post rather than the authority that issued it. The steps are: (1) compare Claude’s explanation with the current documents and note each difference; (2) have a qualified safety reviewer approve the revised explanation; (3) identify the authority that issued the revised standard; (4) revise the explanation to match; (5) obtain the current standard and the authority’s guidance. What’s the right order?

⬜ 5, 3, 1, 4, 2
⬜ 3, 5, 1, 4, 2
⬜ 3, 5, 4, 2, 1
⬜ 1, 3, 5, 2, 4

A Claude-drafted article for a heritage publication reports a 16% rise in volunteer participation. Three trade-news sites repeat the figure, each crediting a survey the editor hasn’t read. What should the editor do before publishing it?

⬜ Accept it, since three sites agree on it.
⬜ Find the original survey and check the percentage, period, population and wording against it.
⬜ Ask Claude to search the web and confirm the 16% in its reply.
⬜ Cite the most authoritative of the three news sites as the source.

A team must share a decision log with dates, owners, alternatives considered, final choices and evidence links. People in three departments need to filter it by owner and update each decision’s status for months. Which output fits best?

⬜ A memo describing each decision in date order.
⬜ A structured spreadsheet with a column for each decision field and evidence link.
⬜ A short message listing final choices, with the rest available on request.
⬜ A brainstorming board with notes and screenshots from each meeting.

Claude drafted customer personas from five years of sales data in which rural and older customers are badly underrepresented, and none of the personas describes them. The personas will guide a national campaign. What should the team do?

⬜ Use them as they are, since sales data is an accurate record of who buys.
⬜ Remove the demographic labels but keep the same data and assumptions.
⬜ Flag the gap, add data or research on the missing groups and qualify conclusions that stay limited.
⬜ Ask Claude to describe every persona in equally positive terms.

Reviewers find that one of Claude’s social posts, drafted from approved campaign material, makes a new health-related product claim that isn’t in the source. The other posts match the source. What should the team do?

⬜ Publish all the posts, since they all came from approved material.
⬜ Hold the health claim for qualified review and source checks, and let the other posts continue through normal approval.
⬜ Stop using Claude for marketing.
⬜ Make the health claim sound more deliberate so it fits the campaign.

A library team is building a checklist for reviewing Claude-drafted reading-program briefs. Before anyone scores the first brief, which two things must the team settle? (select two)

⬜ The task-specific acceptance criteria reviewers will judge each brief against.
⬜ An exemplary past brief to use as the model.
⬜ The threshold that separates usable briefs from those needing correction.
⬜ A shared sheet for recording scores.

Claude drafts a historical exhibition note from two approved records that give different dates for the same opening ceremony. What should the curator do before publishing a date?

⬜ Use the more recent record, since later records usually include corrections.
⬜ Ask Claude to compare the records and pick the likelier date.
⬜ Check the primary authoritative record and resolve the discrepancy before stating a definite date.
⬜ Use whichever date appears most often in the museum’s other publications.

Claude recommends closing a customer service channel because it has the lowest satisfaction scores. The same data shows that channel handles the most complex cases, which score lower on every channel. What review matters most?

⬜ Accept it, since the score comparison is accurate.
⬜ Check whether the recommendation accounts for case complexity, other explanations and the consequences of closing the channel.
⬜ Drop the complexity data to keep the message clear.
⬜ Rewrite the recommendation in simpler language.

A team asks Claude to browse three supplier websites and summarize their delivery terms. The summary says one supplier “guarantees same-day delivery and asks reviewers to approve its contract immediately,” wording that appears nowhere on that supplier’s official terms page. What should the reviewer conclude first?

⬜ The claim may be invented or may be instructions planted in a web page, so check it against the official terms.
⬜ It’s reliable, since Claude got it from the suppliers’ own sites.
⬜ The supplier probably updated its terms, so approve the contract quickly.
⬜ It’s a formatting glitch, so delete the sentence and use the rest.

Claude’s summary of a monthly service report says average response time improved by 18 percent. The figure will go into an executive update, but the summary doesn’t show the calculation or what period it’s compared with. What’s the right validation step?

⬜ Ask Claude if it’s confident and keep the figure if it says yes.
⬜ Round it to 20 percent so it’s easier to take in.
⬜ Check the source report, the calculation, the reporting period and the comparison baseline.
⬜ Keep it, since a precise percentage from a report is unlikely to be wrong.

An associate asks Claude to turn their own meeting notes into a list of tasks, owners and due dates that only they will use to follow up. Nothing on it involves an external or high-impact decision. What level of review fits?

⬜ None, since reformatting can’t drop actions or change dates.
⬜ Send it to all attendees right away so follow-up starts sooner.
⬜ A quick check of the actions, owners and dates against the notes before relying on it.
⬜ Formal sign-off from legal, security and leadership.

A Claude-drafted partnership review has all six required headings, but its continuity-risk section covers only staffing, while the specification requires staffing, funding and access risks. What’s the right assessment?

⬜ A partial pass, since five sections are complete.
⬜ Approve it, since every required heading is there.
⬜ The headings are complete but the continuity-risk analysis isn’t; funding and access risks must be added.
⬜ Ask Claude to make that section as long as the others.

A research office’s two acceptance criteria for a Claude-drafted evidence note are that factual claims match the supplied sources and that the note uses the requested format and length. No house style or executive summary is required. Which two checks test those criteria directly? (select two)

⬜ Check the structure and length against the agreed format and limit.
⬜ Check for an executive summary and the house style.
⬜ Compare the claims with the supplied sources and investigate anything unsupported.
⬜ Ask Claude whether the note meets both criteria.

Claude wrote a detailed outage analysis for engineers, including internal system names and a security weakness that’s still being fixed. A communications specialist needs a customer update on impact, restoration status and next steps. What should they request?

⬜ A short general apology with no outage details.
⬜ The engineering analysis as written.
⬜ A plain-language rewrite with the confirmed impact, current status and approved next steps, leaving out internal and security details.
⬜ A version describing the outage as minor to keep customers calm.

Claude produced two drafts of a community-impact report for residents. Draft 1 is concise and readable but drops the uncertainty around two key estimates. Draft 2 keeps every caveat and source but reads like a technical memo. What should the communications lead do?

⬜ Publish Draft 1, since residents prefer short, confident reports.
⬜ Publish Draft 2, since keeping every caveat is always more transparent.
⬜ Revise Draft 1 to restore the key uncertainty in plain language, using Draft 2 to check its sources.
⬜ Merge the drafts by alternating paragraphs.

Claude’s market-size estimate for a client relies on one vendor report. A regulator dataset and a trade association release give figures about 40 percent lower, and the three sources use different market definitions and years. What should the consultant do?

⬜ Average the three figures.
⬜ Use the vendor figure, since the client wants an ambitious plan.
⬜ Ask Claude to pick the most credible source.
⬜ Compare the sources’ definitions, years and methods, then explain which figure fits the client’s market and why.

Claude summarizes 300 employee comments on promotion fairness. Its main conclusion echoes a few long, strongly worded comments, while many short neutral or positive ones barely appear. Leaders will use it to decide whether to redesign the promotion process. What’s the best response?

⬜ Keep it, since strong comments usually signal the most serious problems.
⬜ Ask for more neutral adjectives but keep the conclusion.
⬜ Ask for a revision that shows how often each view appears and represents the short and positive comments fairly.
⬜ Remove comment counts so leaders focus on themes.

Claude wrote a detailed risk analysis for a steering committee covering approved controls, escalation conditions and governance roles. Frontline teams in three roles now need a practical version that keeps every safeguard and line of accountability. Which adaptation is best?

⬜ Send the committee version unchanged so everyone sees the same wording.
⬜ Summarize only the benefits and leave out risks the teams can’t resolve.
⬜ Turn it into role-specific actions, checkpoints and escalation steps that keep every approved control and its owner.
⬜ Replace the formal controls with general encouragement.

A membership-services associate is deciding which Claude drafts need escalation before use. Which one carries the highest error impact?

⬜ An unreviewed reply to a paying member that promises a specific service deadline.
⬜ An internal FAQ summarizing the volunteer-leave policy.
⬜ A social post announcing next month’s open day.
⬜ A thank-you email repeating figures from the approved annual report.

An HR team needs three outputs from Claude this week: a quick answer to one manager’s leave question, a ten-page handbook section that will go through revisions with legal and a list of 200 policy updates to import into a tracking spreadsheet. Which formats fit best?

⬜ An artifact for all three, for consistency.
⬜ Inline text for all three.
⬜ An inline reply for the quick answer, an artifact for the handbook section and structured data such as CSV for the tracker.
⬜ Structured data for the answer and handbook, and an inline list for the 200 updates.

Claude drafted a public success story about a service improvement for the annual report. It credits senior leaders throughout but never mentions the frontline team whose records supplied the evidence and whose work produced the result. What should the editor do?

⬜ Publish it, since the figures are accurate.
⬜ Add more praise for the leaders.
⬜ Leave it, since recognition is the organization’s call, not Claude’s.
⬜ Revise it to credit the frontline team accurately, and confirm the attribution with the people involved.

---

Domain 3: Product and Model Selection (12%)

Which two kinds of work are the best fit for Claude artifacts rather than another feature? (select two)

⬜ Building a small interactive calculator to show at a team meeting
⬜ Drafting a proposal that will go through several rounds of revision
⬜ Investigating a market question across many current web sources
⬜ Uploading reference files once so every future chat can use them

Which two tasks are best handled in a plain chat rather than a Project, Research or an artifact? (select two)

⬜ Answering a one-off vendor question using a policy document the user already has
⬜ Drafting one quick email reply with no plan to reuse the work
⬜ Producing a monthly report that always uses the same reference files and rules
⬜ Building a cited overview of recent industry changes from many current sources

A planning lead must recommend one of three hotel expansion options. Market forecasts conflict, financing and staffing constraints interact, and a subtle mistake would be costly. The analysis is one-time. Choosing among Haiku, Sonnet and Opus, which fits?

⬜ Run the analysis on Opus.
⬜ Run the analysis on Sonnet.
⬜ Run the analysis on Haiku.
⬜ Start on Haiku and move up a tier if the draft seems weak.

A grants coordinator finalized exact approved wording for five reporting definitions in one chat inside a Project. Every future chat in the Project must use this wording word for word, with no special first message. What should she do?

⬜ Add the finalized definitions to the Project’s knowledge base.
⬜ Rely on the Project to carry the earlier chat’s content into new chats.
⬜ Start each new chat by asking Claude to search the earlier conversation.
⬜ Rely on the Project’s memory to carry the definitions into new chats.

A utility company uses Claude to sort thousands of customer messages a day into three clearly defined categories. Agents wait on each label, and a wrong label is easy to fix. Which choice fits?

⬜ Use Haiku for fast handling of high-volume, well-defined work.
⬜ Use Opus for the deepest reasoning on each message.
⬜ Keep the current model and raise effort to Max.
⬜ Use Sonnet for balanced quality on everyday tasks.

A logistics coordinator has finished a long Claude conversation reworking delivery routes. Her next task is drafting a job posting, which needs nothing from the route work. Where should she do it?

⬜ In a new conversation.
⬜ In the same route-planning conversation.
⬜ In a Project, after saving the route conversation to its knowledge base.
⬜ In a new conversation that starts with Claude’s summary of the route work.

A planner used a fast Haiku-class model in one chat to clean up a delivery schedule. She now has to compare three conflicting routing options, a judgment task that depends on the cleaned schedule. What’s the right next step?

⬜ Start a new chat with a more capable model and paste in the cleaned schedule.
⬜ Switch to a more capable model in the same chat and continue there.
⬜ Keep the fast model and turn on thinking for the comparison.
⬜ Move the schedule into a new Project and ask there with the same model.

A benefits specialist is starting a Claude chat about a confidential restructuring. Nothing from it should be saved to Claude’s memory, but the chat must stay in her chat history. What should she do?

⬜ Turn memory off for that chat before sending the first message.
⬜ Ask Claude in the first message not to save anything to memory.
⬜ Turn memory off in the chat’s settings after the conversation starts.
⬜ Use an incognito chat.

A city policy analyst asks Claude to summarize local licensing rules that changed last month. The summary, written with Claude Sonnet, leaves out the recent changes. What’s the best next step?

⬜ Keep the same model and have Claude search the web for current sources.
⬜ Raise the effort level to the maximum and send it again.
⬜ Switch to Opus and ask for the summary again.
⬜ Tell Claude it missed the changes and ask for a revision.

A team lead on a Claude Team plan built a Project with supplier documents and review instructions. Teammates now run the same weekly review and re-upload the documents every time. What should the lead do?

⬜ Share the Project with the teammates with view or edit access.
⬜ Send teammates the Project instructions to paste into each chat.
⬜ Have each teammate create their own Project with copies of the documents.
⬜ Publish the supplier documents as a shared artifact.

An associate keeps Claude’s effort setting high for every request, including quick lookups and short summaries. Responses are slow, and usage limits arrive sooner than expected. What adjustment fits best?

⬜ Lower the effort for routine requests and keep high effort for tasks that need thorough answers.
⬜ Start a new conversation for each request so earlier high-effort usage resets.
⬜ Use the most capable model for every request so effort doesn’t need to be raised.
⬜ Keep effort high but write shorter prompts.

An associate helping a new teammate get started with Claude wants to explain what actually differs between the Haiku, Sonnet and Opus model tiers. Which explanation is right?

⬜ They sit at different points on a trade-off between speed and cost on one side and reasoning depth on the other.
⬜ Higher tiers remember past conversations, while lower tiers start every chat with no history.
⬜ Higher tiers unlock features such as Projects and artifacts.
⬜ Each tier is trained for a different subject area.

A planning analyst can use Claude’s long context and a calculation tool. Which two assignments genuinely need the capability named, rather than using it for a simple task? (select two)

⬜ Using the calculation tool to decide which wording sounds more welcoming.
⬜ Using long context to synthesize an annual review from many substantial approved sources.
⬜ Using long context to look up one formula in a short procedure document.
⬜ Using the calculation tool for a multi-step allocation with several rates and quantity limits.

A team’s weekly sales-and-inventory report looks plausible on both Haiku and Sonnet, but their yearly usage costs differ a lot, and leaders want the choice to be defensible. What should the team do?

⬜ Choose Sonnet, since the more capable model is the safer default for leadership reports.
⬜ Choose Haiku, since both samples look fine and cost is the only measurable difference.
⬜ Test both on representative weekly reports for accuracy, completeness, speed and cost, then choose against agreed thresholds.
⬜ Alternate models each week and let readers say which they prefer.

A consultancy must investigate this month’s competitor announcements, cite the evidence for each finding and give the client a comparison it can edit live in a workshop. Which two Claude capabilities fit? (select two)

⬜ Research, to investigate the announcements across many sources with citations.
⬜ Memory, to recall what earlier chats said about these competitors.
⬜ An artifact, so the comparison can be viewed and refined separately from the chat.
⬜ A private Project with the client’s brief only, letting Claude fill in competitor details.

A retailer produces 300 product descriptions a week. Most follow a fixed template from structured specifications, but about twenty premium items need nuanced positioning and careful comparison with detailed specs. Which model strategy best balances cost and quality?

⬜ Opus for all 300, so every description gets the strongest reasoning.
⬜ Haiku for all 300, with manual editing for the premium items.
⬜ Haiku for the templated items and a more capable model for the premium twenty, with review criteria suited to each.
⬜ Sonnet for everything, so the team manages only one model.

A coordinator is working through a long Claude conversation about a community market as it moves into new planning subtopics. Which two practices keep the work on track? (select two)

⬜ Paste the whole conversation history into each new subtopic.
⬜ Restate the relevant goals and constraints when the conversation shifts to a new subtopic.
⬜ Keep the working summary to the latest decision only.
⬜ Periodically summarize agreed decisions and open questions into a short working-state record.

A planner switches between Claude and another approved AI service and has work context they’re permitted to bring across. They want less repeated setup in future Claude chats. Which feature helps most directly?

⬜ Connect the other service as a connector so Claude reads its chats directly.
⬜ Use Claude’s memory import to bring in the approved context, then review the imported entries before relying on them.
⬜ Keep the context in incognito chats so it carries across sessions privately.
⬜ Attach a summary document at the start of every new chat.

An analyst is preparing one Claude request comparing four venue proposals. The inputs mix the relevant proposal facts with old correspondence. Which two practices improve focus and source attribution? (select two)

⬜ Paste every full email thread for background.
⬜ Remove irrelevant correspondence and keep the proposal passages the comparison needs.
⬜ Merge the four proposals into one summary first.
⬜ Separate the sources with clear venue labels and section headers.

A team is choosing a model for a time-sensitive executive risk brief. Several models can meet the deadline, but the brief may shape consequential decisions and every claim must survive review. Beyond speed, which two factors should weigh most? (select two)

⬜ Whether the model reaches the reasoning depth and evidence quality the decision needs.
⬜ Which model gave the longest answer in one informal trial.
⬜ Total workflow cost, including the review and rework needed to reach the acceptance standard.
⬜ Which model name will sound most advanced to executives.

A nonprofit coordinator will compare Claude models on representative application-summary tasks. Which two preparations should come before the comparison? (select two)

⬜ Define the required output quality, acceptable response time and expected volume.
⬜ Shortlist the models that top public leaderboards for summarization.
⬜ Choose plausible candidate models based on those requirements and their documented strengths.
⬜ Set every candidate to maximum effort so each is compared at its best.

A project manager will run weekly risk reviews for six months with the same charter, risk definitions and escalation rules. Each review produces an updated risk register that sponsors open and comment on outside the chat. Which setup fits best?

⬜ A new chat each week with the charter pasted in, and the register as a long text reply.
⬜ A Project holding the charter, definitions and rules, with each week’s register produced as a shareable artifact.
⬜ Research for every review, since registers should use the latest external sources.
⬜ Memory alone, rebuilding the register when a sponsor asks.

A team has three recurring tasks: pulling order numbers from thousands of short standardized messages, drafting weekly customer-facing summaries and analyzing a complex contract dispute with conflicting clauses. Which default models fit best?

⬜ Opus for extraction, Sonnet for summaries and Haiku for the dispute.
⬜ Sonnet for all three, to avoid routing.
⬜ Haiku for all three, since speed matters most.
⬜ Haiku for extraction, Sonnet for summaries and Opus for the contract dispute.

A policy analyst has two tasks today: confirm the effective date of one regulation announced last week, and prepare a cited overview of how twelve countries regulate the same activity. Which approach fits best?

⬜ Research for both.
⬜ A plain chat without web search for both.
⬜ Web search for the twelve-country overview and Research for the single date.
⬜ Web search for the single date and Research for the twelve-country overview.

A project manager needs a quick first-pass risk scan of a large document set today, and a defensible board recommendation from the same material next week. The board decision carries significant financial and reputational consequences. Which approach fits best?

⬜ Use the most capable model for today’s scan and reuse that scan as the board recommendation.
⬜ Use the same fast model for both, since deadlines matter most.
⬜ Use a fast model for today’s scan, then a stronger model plus accountable human review for the board recommendation.
⬜ Use whichever model has the largest context window for both.

---

Domain 4: Workflow Integration and Solution Design (16%)

A team refines a dashboard artifact, built with Claude, that only the team uses. Recent revisions came from scattered ad hoc chats and didn’t fix the problems users reported. Which two practices make the next rounds effective? (select two)

⬜ Base each change on what actually failed when the team used the dashboard.
⬜ Keep the revision work in one Project so instructions and constraints stay stable.
⬜ Hand the dashboard to a developer before making further changes.
⬜ Have Claude research dashboard design best practices and apply them each round.

A coordinator drafts donor thank-you letters with Claude, and drafting time has halved. The director asks what stops a wrong gift amount from reaching a donor. Which response answers the question?

⬜ Every letter is compared against the gift record before it is sent.
⬜ Claude can sometimes state a wrong figure in a draft.
⬜ Drafting time has halved since the team started using Claude.
⬜ Each prompt already contains the donor’s exact gift details.

A sustainability analyst is planning packaging changes for rules that five countries updated this year. The plan must reconcile the rule sets and cite sources, and the compliance manual predates the updates. Which use of Claude fits?

⬜ Use Research to investigate the rules across the five countries.
⬜ Have Claude write the comparison from its existing knowledge.
⬜ Ask one quick web search question per country in a chat.
⬜ Upload the compliance manual to a Project and answer from it.

A facilities coordinator asks Claude to turn notes from three tenant meetings into a requirements list. The list includes two requirements no tenant raised. What should the coordinator do next?

⬜ Check each requirement against the meeting notes and remove items without a source.
⬜ Ask Claude to rewrite every requirement in clear, testable language.
⬜ Ask Claude to add standard requirements the notes may have missed.
⬜ Send the list to tenants and ask them to rank the items.

Store managers use Claude only to draft emails. An associate will brief them on Claude’s value and limits for everyday store tasks. Which two points belong in the briefing? (select two)

⬜ Check important facts in Claude’s answers before acting on them.
⬜ Examples of Claude summarizing shift notes and analyzing photos of store displays
⬜ Have the developer team approve each planned use before the briefing.
⬜ Keep Claude limited to email drafting until a formal pilot finishes.

A clinic team produces patient discharge leaflets: a nurse drafts, a pharmacist checks medical accuracy, and a designer formats. Drafting is the slowest stage, and the pharmacist’s check reliably catches errors. Where should the team add Claude first?

⬜ Have Claude draft each leaflet from the nurse’s notes, and keep the pharmacist’s review.
⬜ Have Claude do the accuracy check so the pharmacist can help with drafting.
⬜ Use Claude to speed up formatting and layout.
⬜ Add Claude as an extra accuracy check after the pharmacist.

A school administrator asks whether Claude could draft weekly attendance summaries for the district office. The associate needs to decide how much human checking each summary needs. What should the associate ask?

⬜ What happens if the district acts on a wrong summary?
⬜ How many summaries are produced each week?
⬜ Which Claude model gives the most accurate summaries?
⬜ What format does the district office want?

An operations team builds a staffing plan with Claude from an uploaded year of workload data. Finance will question every figure. Which two practices make the figures defensible? (select two)

⬜ Use code execution to calculate growth and throughput from the uploaded data.
⬜ Ask Claude to state the assumptions behind each recommended figure.
⬜ Use Research to gather industry benchmark staffing figures.
⬜ Have Claude read the data and estimate the figures in its written answer.

An associate built a dashboard artifact for their own team. Three other departments now rely on it daily, and leadership wants a live database connection and guaranteed availability. What should the associate do next?

⬜ Hand the build to a developer or architect, along with the validated design.
⬜ Move the artifact into a dedicated Project with instructions and knowledge files.
⬜ Keep improving the artifact by prompt as requests arrive.
⬜ Write a maintenance plan and monthly review schedule for the artifact.

A transit planner asks Claude to synthesize four consultant reports into a memo. The reports give conflicting ridership forecasts, but the draft reads as if all four agree. What should the planner ask for next?

⬜ Ask Claude to show where the reports disagree and on which points.
⬜ Use Research to gather more outside forecasts.
⬜ Rebuild the memo from only the most recent report.
⬜ Add a source citation to each claim in the memo.

A warehouse supervisor needs a one-page summary of the safety requirements for a newly bought forklift. The manufacturer’s current manual came with it and is saved on the supervisor’s laptop. Which use of Claude fits best?

⬜ Attach the saved manual and ask Claude to extract and summarize the safety requirements.
⬜ Attach the manual and ask Claude to compare it with the company’s safety policy.
⬜ Use web search to find the latest edition of the manual on the manufacturer’s site.
⬜ Run Research to gather safety guidance for this forklift model from many sources.

A training coordinator wants Claude to build a small sign-up tracker as an artifact for their own team. The team’s feature requests are already gathered and prioritized, and the tracker won’t connect to any live system. What should the coordinator do first?

⬜ Have Claude build a simple artifact covering the top priorities, then refine it from feedback.
⬜ Write one detailed prompt asking Claude to build every requested feature at once.
⬜ Interview team members again to confirm and expand the list of needs.
⬜ Hand the prioritized list to a developer to build a production version.

Claude extracted this requirement from a museum’s check-in planning notes: “Check-in must be quick and simple during busy periods.” The notes, written by the front-desk lead, say busy periods happen when school groups of up to 60 students arrive at once. Which two actions best turn this into a testable requirement? (select two)

⬜ Ask the front-desk lead what wait time is acceptable when a school group arrives.
⬜ Have Claude research acceptable check-in wait times at comparable museums.
⬜ Ask Claude to rewrite the statement with a measurable wait-time target for a 60-student arrival.
⬜ Ask Claude to rewrite it with a target for average wait time across the whole day.

A payroll specialist runs a proven monthly Claude workflow that turns timesheet exports into reconciliation summaries, but the steps exist only in the specialist’s head. A teammate must run it during the specialist’s two-month leave. What should the specialist do first?

⬜ Write down the steps, the required inputs and what a finished result looks like.
⬜ Run a new pilot with success measures before the teammate takes over.
⬜ Ask a developer to automate the workflow so it runs on its own during the leave.
⬜ Have Claude draft a standard industry reconciliation process for the teammate.

A catering manager needs one fact: the effective date of a new allergen-labeling rule announced last month. The team’s compliance guide doesn’t cover it yet. How should the manager get the date with Claude?

⬜ Ask in a chat where Claude can search the web.
⬜ Ask Claude for the date from its own knowledge, without searching the web.
⬜ Start a Research session to reconcile many regulatory sources.
⬜ Attach the compliance guide and ask Claude to find the date in it.

An associate’s workflow uses Claude to draft weekly quality reports, and a plant supervisor reviews each draft before it’s distributed. The plant director tells other managers the reports are now fully automated. How should the associate respond?

⬜ Tell the managers that Claude drafts each report and the supervisor approves it before distribution.
⬜ Record the workflow’s steps and owners in the department’s process documentation.
⬜ Label each report as AI-assisted so readers know a drafting tool was used.
⬜ Present the measured time savings at the next managers’ meeting.

A facilities team redesigned its maintenance-request triage around Claude three months ago. The coordinator logged each request’s wait time and every miscategorization. Technicians say wrong categories, not delays, are the main problem. What should guide the first revision?

⬜ Review the logged miscategorizations and revise the steps where errors cluster.
⬜ Review the wait-time records and revise the steps where requests wait longest.
⬜ Move categorization to a more capable model and keep the steps unchanged.
⬜ Ask Claude to assess the workflow against general triage best practices.

A warehouse supervisor sends an associate a one-line request: set up a weekly Claude summary of the site’s safety-incident logs. Before building anything, what should the associate clarify first?

⬜ Who acts on the summary and what decision it feeds.
⬜ What length, format and tone the summary should have.
⬜ Whether a sample summary from last month’s logs looks right to the supervisor.
⬜ Whether a Project should hold the logs so each week’s summary stays consistent.

A catering company wants its first Claude use case to be a recurring task whose output is easy to check against existing material. The kitchen writes notes after every event, and staff turn them into a weekly client summary by hand. Budget details for proposals are agreed in phone calls nobody records. The employee handbook needs a one-time rewrite. Each month the owner picks venue partnership bids based on relationships and risk appetite. Which task is the strongest fit?

⬜ Drafting the weekly client summary from the kitchen notes.
⬜ Rewriting the employee handbook before hiring season.
⬜ Recommending which venue bids to accept each month.
⬜ Drafting client proposals with the agreed budget figures.

A team has finished its planned Claude pilot for summarizing a hotel group’s guest feedback, tracking time saved and error counts against the old manual process. Leadership will decide whether to fund a wider rollout. How should the results be presented?

⬜ Report the measured time savings and error counts, and name the remaining limitations.
⬜ Run a live demo on fresh feedback instead of showing the measurements.
⬜ Compare the results with published benchmarks from other hotel groups.
⬜ Recommend extending the pilot to gather more data first.

A team has mapped its renewal-reminder process and agrees on the problem. Members propose three different redesigns but can’t choose. How should Claude help with the next design step?

⬜ Lay out each redesign with its trade-offs, then pick one to prototype.
⬜ Build a working prototype of the most popular redesign right away.
⬜ Map the current process in more detail before any redesign work.
⬜ Escalate the choice to a developer or architect.

A facilities team’s maintenance approvals are slow, and staff often skip steps in the written procedure. Which two practices help Claude find which internal steps cause the delays? (select two)

⬜ Describe the process to Claude step by step, as the team actually runs it.
⬜ Use Research to compare it with approval workflows at other organizations.
⬜ Paste the written procedure from the policy manual as the description.
⬜ Treat Claude’s suggested bottlenecks as leads to check against the team’s records.

An HR team collected 150 free-text survey responses about problems with its expense-claim process. The redesign must fix the problems employees reported. How should an associate use Claude to produce the requirements list?

⬜ Paste the responses and ask Claude to group them into themes and draft requirements.
⬜ Use Research on how other companies design expense processes.
⬜ Run a second survey with multiple-choice questions first.
⬜ Use only the 20 most detailed responses to keep the input manageable.

A facilities team drafted next year’s maintenance schedule with Claude, then reviewed and approved it. At the budget meeting, a director asks who’s responsible if the schedule is wrong. How should the team lead respond?

⬜ Say the team reviewed the schedule and is accountable for it.
⬜ Suggest treating the schedule as provisional because AI helped produce it.
⬜ Explain Claude’s general strengths and limitations as a drafting tool.
⬜ Present the hours of planning time Claude saved.

A library team runs a written Claude workflow with named inputs and one owner. Members quietly fix wrong outputs during their runs, so at the quarterly review the owner can’t tell which problems keep coming back. What should the team add?

⬜ Record every failure in a shared place when it’s caught.
⬜ Have Claude review each draft for errors before the team sees it.
⬜ Rewrite the Project instructions to warn against common mistakes.
⬜ Ask members to describe recent problems from memory at each review.

An analyst is using Claude to synthesize research from many external sources for a planning report. Which two practices keep the synthesis grounded and checkable? (select two)

⬜ Ask for a supporting source alongside each claim as it’s made, not gathered at the end.
⬜ Ask Claude to present the findings as one confident consensus so the report reads decisively.
⬜ Leave the research question open-ended so no possibly relevant source is excluded.
⬜ Require that disagreements between sources be reported, not smoothed into one account.

A director notes that a team’s Claude-assisted monthly board report has produced accurate drafts for six months and asks whether the human review step can now be dropped. How should the associate respond?

⬜ Explain that confident errors are still possible despite the track record, and recommend keeping review.
⬜ Say removing review is impossible because Anthropic’s policies require human approval of every output.
⬜ Keep reviewing quietly while telling the director the step was removed.
⬜ Agree, since six accurate months show the tool no longer needs checking.

An associate will demo a Claude-assisted vendor-quote comparison workflow to department heads deciding whether to adopt it. The pilot produced typical outputs and one case where the human review step caught an error. Which examples should the demo include?

⬜ Typical outputs, including the case where review caught an error.
⬜ Only the strongest outputs, to make the most persuasive case.
⬜ No concrete outputs, just the benefits in general terms.
⬜ A full list of every error the workflow has ever produced, first.

An error in a Claude-drafted client renewal summary got past review at an insurance brokerage, and the client spotted it. Leadership asks the associate what happened. Which response is appropriate?

⬜ Acknowledge the error, note that confident mistakes are a known limitation, and explain how the review step will be strengthened.
⬜ Call it a one-time software glitch that’s already fixed.
⬜ Promise that a small change to the prompt will prevent similar errors.
⬜ Retire the workflow and go back to fully manual drafting.

A grants team’s Claude workflow has reliably drafted internal meeting recaps through a two-month pilot. An enthusiastic executive asks the associate to tell the board it can also handle grant compliance reports. What should the associate say?

⬜ Present the recap results as evidence for recaps only, and note that compliance reports would need their own pilot.
⬜ Tell the board it handles compliance reports too, since the same model produced the good recaps.
⬜ Run one compliance report through it before the meeting and present that as proof.
⬜ Don’t mention the recap results, to avoid encouraging risky expansion.

An associate used Claude to turn a manager’s rambling notes into requirement statements for a new visitor-feedback process. Before handing them to another team, which test best shows whether each statement is finished?

⬜ A colleague could pick up each statement and act on it without the associate there to explain.
⬜ Claude reviews the set and confirms it’s complete.
⬜ Each statement is at least a full paragraph long.
⬜ Each statement uses formal “must” and “shall” wording.

An apprenticeship team uses Claude in an approved workflow where people keep authority over external publication and personnel decisions. Which two activities fit Claude’s assisting role? (select two)

⬜ Updating placement records directly when Claude finds a better staffing match.
⬜ Drafting a structured project update from approved meeting notes for a person to review.
⬜ Posting the final announcement to the public website once the draft meets the style guide.
⬜ Synthesizing a literature review from approved uploaded research, keeping attributions.

A project manager must find out within a month whether Claude can turn meeting notes into action registers across the organization. The notes range from typed minutes to rough bullets with no owners. Which approach gives the most reliable evidence?

⬜ Test on the best-formatted minutes to see what the workflow can do at its best.
⬜ Pilot with a small group on representative notes, including the rough ones, and score each register against defined criteria.
⬜ Roll out to all teams with a feedback form, since broad use surfaces problems faster.
⬜ Recommend rollout if Claude takes less than half the time staff do now.

Teachers piloting Claude-drafted lesson briefs say the objectives are clear, but activities often run past class time and assume materials schools don’t have. The improvement steps are: (1) generate new briefs for a representative set of lessons; (2) roll out to the wider teaching team; (3) add class-time and available-materials limits to the prompt; (4) collect specific examples of activities that overran or needed missing materials; (5) compare the new briefs with the pilot results on the same criteria. What’s the right order?

⬜ 3, 4, 1, 2, 5
⬜ 1, 3, 4, 5, 2
⬜ 4, 3, 1, 5, 2
⬜ 4, 1, 3, 2, 5

Claude drafts replies to school-group booking requests, and a staff member must confirm dates, charges and cancellation terms before each reply goes out. How should the handoff to that reviewer be designed?

⬜ Document what each draft must include and the booking facts, policy checks and approval conditions the reviewer must verify.
⬜ Have Claude write an approval checklist from its own draft and use it as the full review standard.
⬜ Have a second Claude chat approve each draft and the reviewer confirm the two agree.
⬜ Have reviewers check closely only when Claude flags uncertainty, and spot-check the rest.

A team is redesigning its slow but important monthly executive brief with Claude. The steps are: (1) have a reviewer approve the reconciled facts, decisions and open issues; (2) gather updates from each approved source, recording where each came from; (3) send the presentation to the executives; (4) flag conflicting updates for review; (5) build the slides with Claude. What’s the right order?

⬜ 2, 5, 4, 1, 3
⬜ 2, 4, 1, 5, 3
⬜ 4, 2, 5, 1, 3
⬜ 2, 4, 5, 3, 1

A consultancy is preparing a market-entry plan from the client’s internal capability documents and current regulatory, competitor and customer evidence. Partners review it in two weeks, and every external claim must be traceable. Which workflow fits best?

⬜ Put the capability documents in a Project and rely on Claude’s general knowledge for the market.
⬜ Draft the plan first, then use Research to find sources that back each recommendation.
⬜ Keep the capability documents in a Project, use Research for current external evidence and verify the key cited claims.
⬜ Run web searches in a regular chat as questions come up and paste findings into the draft.

An operations team spends six hours a week copying the same data into three documents, reconciling inconsistent labels and writing similar summaries by hand, and the three versions often disagree. Which improvement should come first?

⬜ Keep the three documents and have Claude rewrite each one after staff finish copying.
⬜ Drop the weekly review and generate the reports straight from raw extracts.
⬜ Standardize the source fields into one reviewed dataset and have Claude generate each summary from it.
⬜ Add a fourth document that records how the other three differ each week.

An operations lead is explaining which tasks fit Claude’s language abilities in an ordinary chat, with no live data, calculation tools or authority to make decisions. Which two fit most clearly? (select two)

⬜ Drafting a customer email in the right tone for its audience.
⬜ Calculating exact multi-year budget totals from a long list of typed-in figures.
⬜ Writing an executive summary from long documents supplied in the chat.
⬜ Reporting the live status of production equipment each shift.

A team is asked to use Claude to speed up proposal review across several departments. Each department reviews differently, and leaders disagree about where the delays are. What should the team establish first?

⬜ The most advanced Claude feature available, to design the process around it.
⬜ A target of removing every human review step.
⬜ The review objectives, current bottlenecks, required evidence, decision owners, constraints and success measures.
⬜ The capabilities most likely to impress in a demonstration.

A museum is adding Claude to its weekly acquisitions summary, with curators keeping responsibility for final recommendations. The steps are: (1) pilot with a small group and gather feedback; (2) identify where Claude can help and where people must stay responsible; (3) document the current steps, owners and decision points; (4) update the documentation and roll out to the wider team; (5) define Claude’s inputs, expected outputs and human handoffs. What’s the right order?

⬜ 2, 3, 5, 1, 4
⬜ 3, 5, 2, 1, 4
⬜ 3, 2, 5, 1, 4
⬜ 3, 2, 1, 5, 4

A food-distribution network uses Claude’s code execution and file creation to build a weekly allocation table from approved data. Which two practices make it dependable as a recurring workflow? (select two)

⬜ Rely on Claude’s code to catch data errors, since code applies the same logic every time.
⬜ Standardize input fields, units and date formats, define the table’s structure and document how missing or unusual data is handled.
⬜ Format the workbook with totals and highlighting so reviewers can scan it.
⬜ Have the responsible reviewer check key totals against the inputs and resolve unexpected results before use.

Claude extracts three conflicting requirements for a mobile-library schedule: one group wants fixed visiting times, another wants on-demand rescheduling and a third assumes staff can guarantee both. What should the associate do next?

⬜ Flag the conflict, keep each requirement linked to its source and ask the stakeholder owners to agree a resolution.
⬜ Ask Claude for a compromise schedule and adopt it as the requirement.
⬜ Record all three as alternatives and resolve them during implementation.
⬜ Adopt the requirement backed by the largest group.

A reviewer accepts most of a Claude-drafted grant brief but asks for three specific corrections. The accepted sections already meet the requirements. What should the associate do next?

⬜ Make the three corrections and also tighten the wording in nearby sections.
⬜ Make the three corrections, leave the accepted content as it is and record each change against the feedback.
⬜ Make the corrections and regenerate the whole brief for consistent style.
⬜ Make the corrections and add a new section the associate thinks reviewers will want.

A prototype uses Claude to turn field reports into maintenance priorities. Accuracy is acceptable, but reviewers spend almost as long fixing inconsistent categories as the old manual process took. What should the team do next, before any wider rollout?

⬜ Remove human review next cycle to improve the time savings.
⬜ Refine the category definitions and examples, retest on representative reports and compare total time with the baseline.
⬜ Scale now, since accuracy is acceptable.
⬜ Rename the priority labels so they look consistent.

A Claude workflow for supplier summaries was piloted in three departments. Two saw large time savings, but reviewers in the third rewrote most drafts, because its source files use different field names and its review rules are stricter. What should the project lead do next?

⬜ Roll out everywhere, since two pilots succeeded.
⬜ Stop measuring quality in the third department.
⬜ Adapt the workflow to the third department’s fields and rules, retest there and compare results with the other pilots.
⬜ Cancel the workflow everywhere.

---

Domain 5: Configuration and Knowledge Management (12%)

A training coordinator answers enrollment questions in a private Project. The course catalog is a Google Doc that colleagues edit several times a week, and answers must reflect the current version. What should the coordinator do?

⬜ Add the catalog to the Project’s files through the Google Drive connector.
⬜ Paste the current catalog into the first message of each chat.
⬜ Upload each new catalog version alongside the earlier ones.
⬜ Download and re-upload the latest copy every Monday.

A long-running Project keeps mentioning a coordinator who has left and a cancelled office move date. The Project’s instructions and knowledge files were reviewed and updated last month. What should the manager do?

⬜ Review the memory Claude has saved and edit or delete the stale entries.
⬜ Rewrite the instructions to match the team’s current process.
⬜ Start each new chat by telling Claude to ignore the old details.
⬜ Replace the knowledge files with current versions.

A coordinator typed the team’s tone and formatting rules into a Project’s description. Drafts still ignore the rules. She also uses the same account for unrelated work. What should she do?

⬜ Move the rules into the Project’s instructions and save them.
⬜ Rewrite the description with stronger, more detailed wording.
⬜ Add the rules to her account-wide Instructions for Claude.
⬜ Upload the rules as a document in the Project’s knowledge base.

A membership team drafts renewal reminders in a Project. Reviewers keep fixing two problems: formal legal wording and a missing renewal deadline. Which text should be saved as the Project’s instructions?

⬜ “Use friendly, plain language and always state the member’s renewal deadline.”
⬜ The full membership handbook, pasted in as the instructions
⬜ “Write excellent, error-free emails that make every member feel valued.”
⬜ “Keep every email under 100 words and close with the coordinator’s name and title.”

An HR coordinator maintains a shared Team-plan Project that colleagues use daily. In its Files section, adding from Google Drive is disabled, though it works in her private Projects. How should she add a Google Doc to the Project’s knowledge?

⬜ Download the Google Doc and upload the file directly to the Project.
⬜ Ask the workspace Owner to enable the Google Drive connector.
⬜ Ask a member with Can edit access to add the file from Google Drive.
⬜ Create a private Project for the document and add it there.

A team uploaded a custom Skill that formats its monthly report. The team changed the report template last month, but Claude still produces the old layout. What should the team do?

⬜ Update the Skill’s content to match the new template.
⬜ Replace the old template in the Project knowledge base.
⬜ Wait for the Skill to pick up the new template automatically.
⬜ Describe the new layout in the Project’s instructions.

Using the Gmail connector, an associate asks Claude to summarize a report that arrived as an email attachment. Claude reads the email body but can’t open the attached PDF. Other Gmail searches work fine. What should the associate do?

⬜ Download the PDF from Gmail and upload it to the conversation.
⬜ Rephrase the request so Claude searches Gmail for the attachment by file name.
⬜ Disconnect and reconnect the Gmail connector, then try again.
⬜ Ask the organization Owner to loosen the Gmail connector’s restrictions.

A newsletter Project has several draft chats each week. Every draft must follow a 40-page brand style guide, and a competitor’s sample should inform only one exploratory chat. Which two actions set this up correctly? (select two)

⬜ Add the competitor’s sample to Project knowledge and remove it after the experiment.
⬜ Paste the full style guide into the Project instructions.
⬜ Attach the competitor’s sample as a file in the exploratory chat.
⬜ Upload the brand style guide to the Project knowledge base.

A help-desk Project’s instructions contain the full, current hardware catalog, and Claude invents model numbers for devices the catalog doesn’t list. Which two configuration changes should the team make? (select two)

⬜ Move the catalog out of the instructions and upload it as Project knowledge.
⬜ Set up a monthly review to remove superseded files from Project knowledge.
⬜ Shorten the catalog before pasting it back into the instructions.
⬜ Add an instruction telling Claude to say so when the catalog doesn’t list a device.

An insurance team’s Project drafts customer letters. Every letter still tells customers to fax documents, but the company moved to an online upload portal last month. The Project’s instructions still include the fax step. What should the team do?

⬜ Update the Project instructions to match the new process.
⬜ Clear the Project’s memory to remove the stale context.
⬜ Add a correction at the start of each new chat.
⬜ Upload a document describing the new process to Project knowledge.

A coordinator’s Google Drive connector finds their own files without problems, but Claude can’t find a teammate’s budget document in a Drive folder the coordinator has no access to. What should the coordinator do?

⬜ Ask the teammate to share the folder with the coordinator’s Google account.
⬜ Disconnect and reconnect the Google Drive connector, then search again.
⬜ Ask an organization Owner to enable the Google Drive connector for the account.
⬜ Paste the document’s Drive link into the chat so Claude can open it.

An editor wants Claude to use British spelling in every conversation, inside and outside her Projects. The rule is saved in one Project’s instructions, so chats outside that Project ignore it. What should she do?

⬜ Add the rule to her account-wide Instructions for Claude.
⬜ Upload a spelling style guide to each Project’s knowledge.
⬜ Paste the rule at the start of each new conversation.
⬜ Copy the rule into the instructions of each Project.

A consultant’s Project for one client needs formal replies, while her other client Projects use a casual tone. A five-step briefing procedure must run the same way across all three Projects. Where should each item be set up?

⬜ The tone rule in that Project’s instructions; the procedure as a Skill.
⬜ Both the tone rule and the procedure in her account-wide Instructions for Claude.
⬜ Both written into each of the three Projects’ instructions.
⬜ The tone rule as a knowledge file; the procedure in this Project’s instructions.

A community theater coordinator has used Claude’s memory for a year. Most entries are about past seasons and former staff, so answers include outdated context. She decides a full memory reset is needed. What should she do first?

⬜ Export Claude’s memory as a backup, then reset it.
⬜ Add a standing instruction to ignore outdated memory entries.
⬜ Edit the stale entries one at a time and skip the reset.
⬜ Turn memory off so no stored context is used.

An associate adds a Google Doc from Drive to a Claude chat. Claude summarizes the text accurately but says nothing about three charts embedded in the document as images. What should the associate do so Claude can analyze the charts?

⬜ Save the charts as image files and upload them to the chat.
⬜ Add the document to a private Project’s knowledge instead.
⬜ Ask the owner for edit access to the document in Drive.
⬜ Disconnect and reconnect the Google Drive connector, then add the document again.

A grants coordinator saved one funder’s strict proposal formatting rules in her account-wide Instructions for Claude. Now unrelated conversations follow those rules too. What’s the right fix?

⬜ Move the rules into the instructions of a Project for that funder’s proposals.
⬜ Keep the rules account-wide but add a line saying to apply them only to grant conversations.
⬜ Upload the rules to a Project’s knowledge and leave the account-wide instructions as they are.
⬜ Delete the account-wide entry and paste the rules at the start of each proposal chat.

A team’s Project has the standing instruction “keep replies concise and professional,” yet drafts still run long and swing between formal and casual. What change makes the instruction work?

⬜ Rewrite it as specific, checkable requirements, such as a word limit and named tone traits.
⬜ Switch the Project to a more capable model.
⬜ Move the instruction into Project knowledge as a style-guide document.
⬜ Paste the same instruction into the first message of every chat as well.

A housing cooperative has approved connectors to its shared project workspace and its public reference library. Personnel and payroll systems need a separate review that hasn’t happened, and personal storage is out of scope. Which two connector setups fit these rules? (select two)

⬜ Connect the shared workspace within its approved data and permission scope.
⬜ Connect payroll read-only so staff can check reimbursements while its review is pending.
⬜ Use an approved read-only connector for the public reference library.
⬜ Connect staff members’ personal storage too, so Claude finds files wherever they’re saved.

A shared Project’s instructions tell Claude to write concise updates, keep every detail, use plain language and include extensive technical explanations. Results vary, and team members now add their own conflicting instructions in each chat. What should the Project owner do?

⬜ Repeat the four requirements in stronger wording.
⬜ Resolve the conflicts by setting priorities, the audience, the level of detail and any exceptions in one coherent set of instructions.
⬜ Add a line asking Claude to balance all four as best it can.
⬜ Remove the instructions and let each person describe the style they want.

A shared Project holds 70 files with vague names, duplicate procedures, expired policies and mixed regional versions. Claude often retrieves the wrong guidance, and staff in two regions have received answers based on the other region’s rules. What’s the best fix?

⬜ Keep every version and tell Claude to work out which is current and which region applies.
⬜ Upload extra copies of the key procedures so Claude retrieves them more often.
⬜ Split into one Project per region and copy all 70 files into each.
⬜ Remove obsolete and duplicate files, name the rest clearly and separate each region’s authoritative versions.

A small translation cooperative plans a Claude Project for recurring client briefings. What should the coordinator settle first?

⬜ Who needs access, and share the Project with them.
⬜ The Project’s purpose and the recurring tasks it should support.
⬜ Upload the client files and style guides as reference material.
⬜ Duplicate the most-used existing Project and adjust its instructions.

Several departments share a Project for procurement reviews. Nobody owns its instructions, teams upload conflicting templates and changes go live untested. Last month a template edit broke the approval checklist for two departments. Which improvement is best?

⬜ Assign one owner, document the approved sources and instructions, version changes and test them before release.
⬜ Let each department edit the instructions and templates directly so updates arrive quickly.
⬜ Freeze the configuration permanently.
⬜ Back up the Project before each change and keep releasing directly.

A science center’s Project produces recurring school-visit materials, and approved sample documents are already in its knowledge. Which two items belong in the Project instructions as lasting writing rules? (select two)

⬜ This term’s visit dates and booking contacts.
⬜ The required structure and length for school-visit documents.
⬜ Let each reviewer apply their own style preferences.
⬜ The center’s approved voice and tone.

A consultant supports two unrelated community organizations, each with confidential reference material. Which setup best keeps their work separate?

⬜ Two Projects, each holding the combined research library for both clients.
⬜ One Project with all files, stating the client at the start of each chat.
⬜ A separate Project for each client, with its own knowledge, instructions and memory, sharing controlled, and responses checked for cross-client references.
⬜ One Project with files prefixed by client name and an instruction to use only the right ones.

A ferry team’s Project drafts apologies for cancelled bookings. Its instructions say to be casual, very brief and always offer a voucher. The approved guide requires formal empathy without contractions, enough explanation and vouchers only for eligible cases. Which change aligns the setup?

⬜ Upload the guide and voucher policy as knowledge and leave the instructions as they are.
⬜ Upload past apologies as examples and keep the current instructions.
⬜ Rewrite the instructions to require formal, empathetic wording without contractions, enough detail to explain the cancellation and vouchers only for eligible cases.
⬜ Replace the tone rule with a general reference to the brand guide and remove the voucher guidance.

An energy adviser maintains a Project with approved terminology, eligibility rules and reference documents. Answers still look useful, but the rules change from time to time. How should the adviser maintain it?

⬜ Update it only when users report a wrong answer.
⬜ Add each new rule document as it’s published and keep the old versions too.
⬜ Ask Claude monthly to flag anything that looks outdated.
⬜ Review the instructions and references against current policy on a regular schedule, and update them whenever a verified change happens.

An education team is writing instructions for a new curriculum Project. Teachers will upload final standards, draft notes and older course files, and earlier Projects mixed draft ideas into lessons as if they were approved. Which instruction approach is best?

⬜ Tell Claude to use any uploaded file it finds relevant.
⬜ State the Project’s purpose, name the final standards as authoritative and say how drafts and older files may be used.
⬜ Write one line asking Claude to be accurate and follow the curriculum.
⬜ Paste every standard’s full text into the instructions.

A support coordinator uses the Gmail connector under an approved setup that enables only search and read tools. Policy requires approval for each connector action and requires staff to send messages themselves in Gmail. Which workflow follows the setup and still uses Claude to prepare a reply?

⬜ Approve the first search, let follow-up reads ride on that approval and have Claude save a Gmail draft.
⬜ Search and read with approval, but write every reply by hand without Claude.
⬜ Approve each permitted search or read, have Claude draft the reply text in the chat and have the coordinator review and send it in Gmail.
⬜ Search and read with approval, and let Claude send routine template replies.

An instructional designer’s Project for induction modules needs a trusted Skill for its reusable design procedure, approved course references and standing task guidance. Which setup combines these correctly?

⬜ Upload the references, leave the Skill off and rebuild the procedure in each chat.
⬜ Turn on the Skill and leave the Project’s knowledge and instructions empty.
⬜ Enable the trusted Skill through the normal controls, add the references to Project knowledge and write instructions that say how to use both.
⬜ Upload the Skill’s files to Project knowledge and describe it in the instructions.

---

Domain 6: Governance, Risk, and Responsible Use (15%)

An organization’s AI policy predates a new Claude feature and doesn’t mention it. The policy names an owner for questions, and teammates already use the feature. What should an associate do before using it for client work?

⬜ Ask the policy owner whether the feature may be used.
⬜ Follow the teammates’ practice as the team’s working standard.
⬜ Document the team’s practice and raise it at the next policy review.
⬜ Use the feature only for internal drafts until the policy is revised.

A recruiter wants Claude to score 200 job applications and automatically email rejections to the lowest-scoring half. Names are already removed and each rejection discloses AI assistance. What change does the plan still need?

⬜ Claude summarizes and organizes applications, and a person decides each rejection.
⬜ Remove applicant names and contact details before uploading.
⬜ Add a line to each rejection disclosing AI assistance.
⬜ Have a hiring manager review a random sample of rejections.

An analyst removed names from a small customer dataset before uploading it to Claude. The trend analysis needs only purchase totals by month and region, but account numbers and rare job titles remain. Which two actions are appropriate? (select two)

⬜ Check whether the organization’s data policy allows uploading the de-identified file.
⬜ Remove or replace the remaining account numbers and job titles before uploading.
⬜ Ask compliance for an approved way to upload with the identifiers kept.
⬜ Tell Claude at the start of the chat to treat the remaining fields as confidential.

An office manager wants Claude to draft replies to routine appointment-scheduling emails from customers. The replies go outside the organization, and a wrong reply is easy to correct. What should the office manager do?

⬜ Have Claude draft each reply, then check it before sending.
⬜ Ask the AI governance team to approve the use case first.
⬜ Write the replies without Claude because they go to customers.
⬜ Have Claude draft and send replies with no review.

A committee plans to use Claude to help score thousands of scholarship applications. There’s an AI governance team, but no policy covers this use. After reasoning through who is affected and what fairness requires, the question is still unresolved. What should the committee do?

⬜ Escalate to the AI governance team, with the documented reasoning.
⬜ Decide within the committee, document the reasoning, and begin scoring.
⬜ Disclose AI assistance to applicants, then proceed.
⬜ Proceed, with a committee member reviewing each Claude-assisted score.

A manager used Claude to draft recognition write-ups about each team member for the company newsletter, working from her own notes, and verified each draft against the notes. Policy doesn’t require disclosing AI assistance. What should she still do before sharing?

⬜ Compare the write-ups for consistent standards and fair tone across team members.
⬜ Verify that each write-up reflects the original notes.
⬜ Add a line to each write-up disclosing that Claude helped.
⬜ Redo the drafts with employee names replaced by labels.

An analyst used Claude to help draft an article for an industry magazine. The submission form asks authors to list any material assistance. Company policy says nothing about AI disclosure. How should the analyst handle disclosure?

⬜ List the Claude assistance on the magazine’s submission form.
⬜ Add a note in the article telling readers that AI helped draft it.
⬜ Leave Claude off the form as ordinary writing tooling.
⬜ Hold the submission until the company issues AI guidance.

A manager has approval to draft a confidential merger announcement with Claude, and the draft needs the deal terms. She doesn’t want those details saved to memory or shown in chat history. What should she do?

⬜ Do the drafting in an incognito chat.
⬜ Move the work into a Project with instructions marking the notes confidential.
⬜ Replace the deal terms with placeholders before pasting the notes.
⬜ Turn off web search before starting.

An associate holds a partner company’s unreleased sales figures under a confidentiality agreement. The file contains no personal or regulated data. What should the associate do before asking Claude to analyze it?

⬜ Confirm what the confidentiality agreement permits before uploading.
⬜ Check for regulated data categories, then upload if none appear.
⬜ Remove personal identifiers, then upload.
⬜ Use an incognito chat so the conversation isn’t saved.

An associate’s analysis needs the account numbers in a customer file, but the associate isn’t sure whether the file counts as internal data or regulated financial data. Which two actions are appropriate? (select two)

⬜ Upload a few sample rows to test which fields the analysis uses.
⬜ Ask the policy owner whether an approved path exists for this data.
⬜ Remove the account numbers and run the analysis on the remaining fields.
⬜ Treat the file as regulated data until its classification is confirmed.

An associate must send a supplier a contract translated into Japanese. The signed translation will be legally binding, and nobody at the associate’s company reads Japanese. How should the associate handle it?

⬜ Have a professional translator prepare the signed version, and use Claude’s draft to understand it.
⬜ Have Claude translate it, translate it back into English, and compare the two.
⬜ Have the legal team review the English original, then send Claude’s translation for signature.
⬜ Ask a colleague on the legal team who reads some Japanese to look over Claude’s translation.

A written AI policy says client contracts must not be uploaded to Claude in any form. An associate’s teammates upload redacted copies anyway, and the associate agrees the rule seems outdated. What should the associate do?

⬜ Keep the contracts out of Claude and raise the rule with the policy owner.
⬜ Remove client names and deal terms, then upload the redacted versions.
⬜ Follow the teammates’ practice until the policy owner updates the rule.
⬜ Send the policy owner a case for change and keep uploading while they review it.

A hiring panel member uses Claude to summarize interview notes for each candidate. The prompt asks, “Explain why this candidate might struggle in the role.” Every summary comes back negative. What should the member do first?

⬜ Reword the prompt to ask for both strengths and weaknesses.
⬜ Report the pattern of negative summaries to the AI governance team.
⬜ Add a note to each summary saying AI helped draft it.
⬜ Write the summaries by hand instead of using Claude.

An associate at a property firm wants to use Claude. The firm’s attorney personally answers all tenant questions about leases, and the emergency repair list goes straight to repair crews, who act on it without anyone checking it. Which two tasks are appropriate for the associate to do with Claude? (select two)

⬜ Summarize last quarter’s repair requests to find buildings with repeat problems.
⬜ Sort today’s incoming repair requests into the emergency and routine lists.
⬜ Draft replies to tenants asking whether their leases allow subletting.
⬜ Explain confusing clauses in the standard lease so the associate understands them.

Which two practices help a team keep its everyday Claude use in line with its organization’s AI policy over time? (select two)

⬜ Apply the policy strictly to high-stakes work and rely on judgment for routine tasks.
⬜ Periodically compare the team’s actual Claude use with what the policy requires.
⬜ Have each member read and sign the policy once when joining the team.
⬜ Recheck existing workflows whenever the policy or Claude’s features change.

A warehouse safety coordinator is behind on the daily forklift checklist, which records physical checks of each machine. She’s thinking of asking Claude to fill it in using last month’s completed forms. What should she do?

⬜ Do the physical checks herself and use Claude only to format her notes.
⬜ Have Claude draft the entries from past forms and review each one before filing.
⬜ Ask her supervisor to approve moving today’s inspection to tomorrow.
⬜ Ask Claude to find recurring problems in past forms and check only those forklifts.

An associate finds a custom Skill shared in a public forum that would automate a recurring report. The organization’s AI policy requires a review of any new software before it’s installed. What should the associate do first?

⬜ Submit the Skill to the policy’s software review process before enabling it.
⬜ Turn off memory in the chats where the Skill will run.
⬜ Test the Skill on anonymized internal files before relying on it.
⬜ Check the Skill’s author and permissions personally, then enable it.

A community college office uses Claude to draft several documents in one week. Which draft most needs a check for bias and fairness before it’s used?

⬜ Summaries of applicant essays that will rank students for limited tutoring places.
⬜ A reformatted version of last semester’s public event calendar.
⬜ An internal email quoting figures for an unannounced tuition increase.
⬜ A plain-language summary of a vendor contract’s payment deadlines.

An associate wants Claude to summarize a competitor’s annual report for an internal comparison. The competitor has published it on its public investor website. How should the associate handle the file?

⬜ Upload the report and ask Claude for the summary.
⬜ Remove the financial figures before uploading it.
⬜ Ask the compliance team to approve the upload first.
⬜ Hold it until the policy owner confirms its sensitivity tier.

A credit union associate wants Claude to analyze member account records. The internal AI policy allows uploading internal operational data, but a financial regulation restricts how member account data may be processed. What should the associate do?

⬜ Keep the records out of Claude and raise the conflict with the compliance team.
⬜ Follow what the rest of the team does with similar records.
⬜ Upload the records, since the internal policy allows it.
⬜ Remove member names, then upload the records.

An associate worked through the terms of a confidential acquisition in a regular chat with memory on. Team policy says acquisition details must not be kept in memory across chats. What should the associate do now?

⬜ Review the saved memories and delete the acquisition details.
⬜ Use incognito chats for any further work on the acquisition.
⬜ Turn off memory so future chats save nothing new.
⬜ Start a new chat and tell Claude to treat the acquisition as confidential.

A coordinator prepares safety-incident reports from a 12-person plant team for Claude to analyze. She deletes employee names but leaves each report’s distinctive job title and exact incident date. What should she do before uploading?

⬜ Remove or generalize the distinctive job titles and exact dates.
⬜ Tell Claude in the first message to treat every report as anonymous, then upload.
⬜ Use an incognito chat and upload the reports unchanged.
⬜ Upload them as prepared, since removing names is enough.

Which two screening decisions correctly identify a task as inappropriate to hand to Claude? (select two)

⬜ Ruling out Claude for a large document set because a person would need days to read it.
⬜ Declining to have Claude produce a legal or medical determination someone will act on without expert review.
⬜ Keeping a task with a person when its outcome would be hard to reverse and serious for the people affected.
⬜ Avoiding Claude whenever the subject is highly specialized rather than everyday business writing.

Which two uses of Claude are appropriate as long as a person reviews the output before it’s used? (select two)

⬜ Choosing which employees get year-end bonuses, with the manager reading Claude’s reasoning after amounts are announced.
⬜ Analyzing an unredacted file of patient health records, since the final report will be reviewed before it circulates.
⬜ Drafting performance-review summaries from a manager’s own notes, with the manager checking each before delivery.
⬜ Writing a first draft of the customer newsletter, which a marketer edits and approves before it’s sent.

A new ten-page organizational AI policy has just been issued. An associate needs to work out which parts actually constrain their everyday Claude use. What’s the effective way to read it?

⬜ Pick out the explicit requirements, prohibitions and permissions, note what each restricts, and list the ones that apply to their work.
⬜ Treat anything not explicitly permitted as prohibited until every section is clarified.
⬜ Read only the introduction and summary, since the details repeat them.
⬜ Give every sentence equal binding weight so no rule is missed.

A support team plans to route customer chat questions through Claude and send the replies under a made-up human agent’s name, without mentioning AI. How should an associate assess the plan?

⬜ Advise against it, because customers would think they’re talking to a person, and recommend disclosing Claude’s role.
⬜ Support it, since a made-up name protects real employees’ privacy.
⬜ Support it, since a person reviewing each reply makes the messages human-written.
⬜ Object only if reply accuracy can’t be checked before sending.

A credit union team plans to have Claude generate personalized retirement-savings recommendations and email them to members automatically, with nobody reading them first. What should an associate flag?

⬜ A qualified person must review each recommendation before it’s sent, and members must be told AI helped produce it.
⬜ Claude shouldn’t be used for any financial content at all.
⬜ A disclaimer that the advice is automated makes human review unnecessary.
⬜ Stronger prompts about accuracy and tone make the recommendations safe to send directly.

A team’s usage audit finds two gaps: a required review step was skipped under deadline pressure, and a restricted document was uploaded. Which two actions address these correctly? (select two)

⬜ Make the review step a fixed gate in the workflow so deadline pressure can’t skip it.
⬜ Treat both as one-off mistakes needing no follow-up, since nobody acted maliciously.
⬜ Delete the conversation with the restricted document and consider it resolved.
⬜ Report the restricted upload through the organization’s data-exposure process.

An associate in an external certification program, whose rules require submitted essays to be the participant’s own original work, is thinking of having Claude write the essay. What should the associate recognize?

⬜ Submitting it would be plagiarism; Claude should only provide permitted help, such as feedback on a self-written draft.
⬜ Heavy editing of Claude’s draft makes it the associate’s original work.
⬜ Originality rules only cover copying people, not tools.
⬜ It’s fine as long as the conversation stays private.

A teammate pilots a new Claude workflow, finds the output consistently accurate, and says the use case is therefore appropriate to adopt without further screening. What should the associate point out?

⬜ Accuracy doesn’t decide appropriateness; the task’s stakes, consequences and accountability do.
⬜ Nothing; consistently accurate output is the best evidence a use case is appropriate.
⬜ The pilot was too small, so a longer trial is needed first.
⬜ The conclusion only holds if the pilot used the most capable model.

Which two writing tasks should an associate keep for themselves rather than hand to Claude, even though each is short and easy to revise? (select two)

⬜ An apology to a teammate after a heated disagreement in yesterday’s meeting.
⬜ A personal condolence message to a colleague whose parent has died.
⬜ Turning rough notes into a first draft of an internal how-to guide.
⬜ The agenda and timings for next week’s recurring status meeting.

A team screened a Claude workflow that summarizes internal support tickets and judged it appropriate. A teammate now wants to reuse the same workflow to draft replies that go straight to customers. What should the associate recognize?

⬜ The customer-facing version is a new use case that needs its stakes and consequences screened afresh.
⬜ Customer-facing communication is never suitable for Claude.
⬜ Reuse is fine if the same prompts and Project setup are kept.
⬜ The earlier approval covers it, since the task is essentially the same.

Before pasting part of a working document into a Claude conversation, which two questions belong in a quick data-sensitivity check? (select two)

⬜ Whether Claude can be told at the start to keep the content confidential.
⬜ Whether deleting the conversation later would fully undo any exposure.
⬜ Whether the task actually needs the sensitive parts, or they can be removed first.
⬜ Whether any policy, contract or regulation places obligations on the material.

A heritage-program manager used Claude to draft a briefing for an outside partner. Which action best reflects the AI Fluency competency of Diligence before it’s sent?

⬜ Add a statement that Claude helped, describing which sections it drafted.
⬜ Identify the claims that need checking, verify them against authoritative records and take responsibility for the version that’s sent.
⬜ Have a separate Claude chat review the briefing and apply its suggested fixes.
⬜ Have a colleague proofread it for tone, clarity and house style.

A communications team is using Claude to draft a public impact report from internal analysis. Leaders want readers to see the evidence behind each conclusion and know who stands behind the report. Which two practices fit? (select two)

⬜ Name the people who review and approve the report and are accountable for its conclusions.
⬜ State that Claude generated the analysis, so the conclusions read as the model’s view.
⬜ Keep source attribution for key claims and disclose important uncertainties and limits in the evidence.
⬜ Present every conclusion with the same confidence so the report reads consistently.

A warehouse supervisor wants Claude to turn approved procedure notes into a draft safety checklist. No personal data is involved, and the supervisor will review it and test it on the floor before anyone uses it. Is this an appropriate use?

⬜ No, because anything affecting workplace safety must be written entirely by people.
⬜ Yes, and since the notes are approved the checklist can be used as soon as Claude produces it.
⬜ Yes, because Claude drafts from approved notes and the supervisor checks and tests the result before use.
⬜ No, because building a checklist counts as workflow design that needs a developer first.

A customer-relations associate is preparing data for a routine Claude task. Which data most clearly needs extra handling, such as removal, redaction or separate approval?

⬜ Complaint summaries that quote messages but replace names with ticket numbers.
⬜ Government ID numbers and full payment-card numbers linked to named customers.
⬜ An internal list of staff names and office extensions for routing escalations.
⬜ Satisfaction scores aggregated by region.

An HR team wants Claude to find themes in 300 exit-interview notes. Policy allows this only with data minimization and accountable HR review. The notes include names, health disclosures and unresolved allegations, and leaders want only aggregate themes. Which two actions should the team take? (select two)

⬜ Run the analysis in an incognito chat so the notes stay out of history and memory.
⬜ Remove names and any details the themes don’t need, including individual health information.
⬜ Upload the notes unchanged and tell Claude to leave names and health details out of its output.
⬜ Have an accountable HR reviewer validate the themes and approve any action that follows.

An associate suspects a Claude-drafted internal note repeats a restricted funding figure that shouldn’t appear in it. The note hasn’t been shared, and policy requires reporting suspected restricted-data reproduction even before any external disclosure. What should the associate do now?

⬜ Remove the figure, use the corrected note and log the change.
⬜ Delete the note, regenerate it and report only if the figure appears again.
⬜ Stop using the note, keep the prompt and evidence securely and report it through the governance channel.
⬜ Start a new chat with a tighter prompt and carry on, since the note wasn’t shared.

An operations team lets Claude work in shared files and draft outbound messages. Some tasks can delete records or send messages to customers. The team wants to keep the efficiency gains without losing control. Which control fits best?

⬜ Give Claude broad access and review a weekly log of its actions.
⬜ Tell Claude to be careful with deletions and external messages instead of setting permissions.
⬜ Limit access to the folders and tools each task needs, and require approval before deletions, external messages and other consequential actions.
⬜ Require approval for every file read, and let deletions and messages proceed freely.

An employer uses Claude to draft interview questions for one role, and every candidate will get the same questions in the same order. Which review best supports fairness?

⬜ Keep every question general so no one’s background gives an advantage.
⬜ Check that each question applies equally to all candidates, avoids assumptions about backgrounds and tests qualifications the role really needs.
⬜ Ask Claude to review the set for bias and approve it if Claude says it’s neutral.
⬜ Add tailored follow-up questions from each candidate’s resume.

An analyst wants Claude to find recurring service themes in support records containing names, account numbers, complaint details and payment information. The data policy sets rules by sensitivity level. The steps are: (1) remove fields the analysis doesn’t need, such as account and payment numbers; (2) confirm which Claude environment the policy approves for that classification; (3) upload the minimized records; (4) classify the records under the data policy; (5) identify which fields the analysis actually needs. What’s the right order?

⬜ 5, 1, 4, 2, 3
⬜ 4, 2, 5, 1, 3
⬜ 4, 5, 1, 3, 2
⬜ 2, 4, 5, 1, 3

Reviewers notice that Claude’s draft profiles of students chosen for an advanced program keep praising assertiveness and visible leadership while saying little about quieter students’ strong analytical work. Which two concerns should they raise? (select two)

⬜ The drafts may reflect a narrow view of merit that disadvantages students with different communication styles.
⬜ The profiles are too short and should give every student equal space.
⬜ Using the drafts uncorrected could reinforce exclusion and skew later judgments about students’ potential.
⬜ Claude should help set the selection criteria, since AI applies criteria more consistently.

A public-service unit has approval to use Claude only for low-risk summaries of permitted operational feedback and internal training material. Employment, clinical and legal decision support need separate approval it doesn’t have. Which two proposals fit its current approval? (select two)

⬜ Draft an internal training summary from the unit’s current approved documents.
⬜ Summarize performance reviews for promotion recommendations, with managers deciding.
⬜ Summarize themes in approved operational feedback using the unit’s aggregation rules.
⬜ Draft summaries of client legal case files for attorneys to finalize.

A shared Project can read procurement files and draft supplier messages. Some members may approve contracts, and others may only prepare analysis. An internal audit next quarter will check who approved each supplier commitment. Which control design fits best?

⬜ Give every member the same broad access to keep collaboration simple.
⬜ Match access and approvals to each person’s authority, limit sources to what’s needed and log key review decisions.
⬜ Tell Claude in the instructions to decline requests that exceed a user’s authority.
⬜ Let the most experienced member approve everything informally.

A communications specialist uses Claude to draft a public apology after a service outage. The draft accepts responsibility for causes the ongoing investigation hasn’t confirmed, and legal has asked to review any public statement. What should the specialist do?

⬜ Publish it as is, since accepting broad responsibility shows the outage is taken seriously.
⬜ Separate confirmed facts, acknowledged impact and commitments from open findings, then send it for legal review.
⬜ Ask Claude to phrase the unconfirmed causes more vaguely.
⬜ Remove all mention of customer impact.

A coordinator plans to upload a customer account workbook for analysis with code execution. Which two preparations best support sound data handling? (select two)

⬜ Mark the file confidential in its filename and properties.
⬜ Inspect the workbook, including hidden sheets and columns, and remove or redact what the analysis doesn’t need.
⬜ Note in the request that results will be shared only within the team.
⬜ Check the analysis against the data classification rules and escalate anything outside the approved scope before upload.

An employer proposes having Claude issue final applicant-selection decisions with no human review. Applying the AI Fluency competency of Delegation, what’s the right assessment?

⬜ Call the output a recommendation and have staff confirm all outcomes in bulk.
⬜ Reject full delegation, keep an accountable person deciding each outcome and limit Claude to an assisting role.
⬜ Approve it once Claude reproduces past hiring outcomes on a validation sample.
⬜ Allow it, with HR reviewing a random monthly sample.

A residents association needs a documented assessment before approving a Claude correspondence workflow. The steps are: (1) record the decision and its rationale; (2) identify the data types and sensitive information involved; (3) decide to approve, modify or decline; (4) identify the audience, decisions, potential impact and reversibility of the outputs; (5) compare the use case and data with policies and regulations. What’s the right order?

⬜ 4, 2, 5, 3, 1
⬜ 2, 4, 5, 1, 3
⬜ 4, 5, 2, 3, 1
⬜ 5, 4, 2, 3, 1

A marketing team needs Claude to analyze confidential campaign results before an internal budget review. Policy allows this data only in approved private Projects open to authorized staff, and bans sharing it outside the company before approval. One analyst suggests using a personal account to move faster. What’s the right action?

⬜ Use the personal account and delete the chat afterwards.
⬜ Publish the analysis as a shared artifact so the outside agency can comment.
⬜ Work in an approved private Project with authorized access only, and keep outputs inside the required review process.
⬜ Leave the policy constraints out of the prompt and settle access later.

An educator uses Claude to compile a learner resource guide from several external publications, including direct quotations and facts learners can check. The educator is responsible for everything learners receive. Which practice best supports transparency and accountability?

⬜ Present the guide as the educator’s own original work.
⬜ Include every link Claude suggests, even unchecked ones.
⬜ List Claude as the single source.
⬜ Cite the original sources, verify the quotes and claims, disclose AI help appropriately and take ownership of the final guide.

A collections team proposes automatically sending Claude-drafted debt-collection messages built from account notes that may be incomplete, outdated or disputed. Some recipients are vulnerable, and the messages carry legal and reputational risk. The main justification is efficiency. Which two concerns matter most? (select two)

⬜ Messages could pressure or mislead customers when the notes behind them are wrong or disputed.
⬜ Fewer hand-drafted messages may change staffing needs.
⬜ Automatic sending removes the human check on consequential messages that need case-specific judgment.
⬜ Generated messages may be shorter than the ones staff write.

You’re doing an ethics review of a Claude-assisted volunteer recruitment brochure before a festival publishes it. The steps are: (1) make targeted revisions or escalate concerns beyond your authority; (2) identify the audience and the brochure’s potential impact on them; (3) document the outcome, corrections and escalations; (4) check for unsupported generalizations, fairness and accuracy issues; (5) confirm the AI-use disclosure and other transparency requirements. What’s the right order?

⬜ 4, 2, 5, 1, 3
⬜ 2, 4, 5, 1, 3
⬜ 2, 5, 1, 4, 3
⬜ 2, 4, 1, 3, 5

An HR specialist wants Claude to find themes in exit interviews that include names, health disclosures, pay details and misconduct allegations against named managers. Policy allows anonymized HR analysis in the approved workspace, and misconduct allegations must go to the investigations team. What should the specialist do?

⬜ Upload the full records and let Claude strip names during the analysis.
⬜ Route the allegations to the investigations team, then analyze themes from records with names and unneeded sensitive fields removed.
⬜ Analyze the allegations with Claude first to give investigators an organized summary.
⬜ Treat the records as ordinary business data.

A people-operations team is reviewing ethical risks in a Claude-supported workflow. Which two issues deserve the highest priority because of their potential effect on employees? (select two)

⬜ Inconsistent formatting of policy references in HR letters.
⬜ Outputs that systematically disadvantage employees with particular protected characteristics.
⬜ Slow responses during busy periods.
⬜ Employment decisions taking effect without an accountable human owner.

A conservation charity is considering a third-party Skill, from a publisher it hasn’t used before, that formats donor summaries. It keeps a register of approved Skills and prohibits unapproved transfers of donor data. Which two checks best support the trust decision? (select two)

⬜ Enable it after seeing strong ratings and use by other nonprofits.
⬜ Review its instructions, bundled scripts, dependencies and network connections with IT, and check its data access against the donor-data rules.
⬜ Test it on synthetic records, then enable it once the formatting looks right.
⬜ Check the publisher and the Skill against the approved register, and send it for review if it isn’t listed.

An HR team may use Claude for anonymized analysis of employee-relations notes. A manager proposes also scoring individual employees’ “attitude risk” from the notes and sharing the scores with line managers. Which two ethical concerns should HR raise? (select two)

⬜ Scoring individuals from informal notes could unfairly label employees and affect them with no chance to respond.
⬜ Generating scores could delay the monthly people report.
⬜ Employees would be judged on data collected for another purpose, which they didn’t expect or agree to.
⬜ Sharing would be fine if names were removed first.

A supervisor is explaining why the organization keeps an AI governance framework. Which description best states its practical purpose?

⬜ It explains how to enable and connect the approved AI products.
⬜ It defines approved tools, permitted data, required reviews and reporting routes for accountable AI use.
⬜ It lists the laws and regulators that apply to AI use.
⬜ It records which AI products were bought, who’s licensed and renewal dates.

---

Domain 7: Troubleshooting and Optimization (10%)

Members of a shared Project paste the same three items into each monthly report chat: a tone rule, a product list and the report-building steps. Outputs vary because members run the steps differently. Which fix should come first?

⬜ Turn the report-building steps into a Skill for the team.
⬜ Add the tone rule to the Project’s instructions.
⬜ Save all three items as one prompt to paste each month.
⬜ Upload the product list to the Project’s knowledge base.

A logistics team has just moved the repeated background for its weekly client report into a Project knowledge file and added a standing tone instruction. What should the team do before fully relying on the new setup?

⬜ Keep the old approach available for a few cycles and compare the outputs.
⬜ Measure the time saved per cycle and report it to managers.
⬜ Turn the whole report procedure into a Skill.
⬜ Schedule a quarterly review of the Project’s instructions and files.

A museum newsletter editor drafts each issue in a Project. Last week, the in-chat feedback “use shorter sentences and name the featured exhibition first” fixed the draft. This week the same problems returned. What should the editor do?

⬜ Add the wording rules to the Project’s instructions.
⬜ Save the wording rules as a document in the Project’s knowledge base.
⬜ Add the wording rules to the Project’s name and description.
⬜ Give Claude the same feedback again in this week’s chat.

A fundraising manager iterates with Claude on a one-time donor letter. Feedback rounds one to three clearly improved it, rounds four and five only changed minor wording, and one short phrase still reads awkwardly. What should she do?

⬜ Edit the remaining phrase by hand and treat the iteration as finished.
⬜ Save the refined prompt as Project instructions for reuse.
⬜ Start a new chat and resend the prompt with all earlier feedback.
⬜ Send another feedback round asking Claude to adjust that phrase.

A procurement team’s monthly vendor review with Claude takes about an hour per run, and nobody can say which steps waste the time. The team wants to speed up the workflow. What should it do first?

⬜ Watch one full run and note every step done by hand.
⬜ Move the team’s reference documents into Project knowledge.
⬜ Switch the Project’s chats to a faster Claude model.
⬜ Turn the review procedure into a shared Skill.

A Claude Project holds dozens of long reports. A staff member confirms one report is in Project knowledge and has a clear, descriptive file name, yet Claude’s answer leaves out that report’s findings. What should the member try first?

⬜ Ask again and name that report’s exact file name in the prompt.
⬜ Rename the Project’s files with clearer, more descriptive names.
⬜ Check whether the report was actually added to Project knowledge.
⬜ Add more background about the question’s purpose and send it again.

In a new chat, a tourism office employee asks Claude for a one-time press release. The structure, tone and length are right, but the quotes section misses the mayor’s key message. What should the employee do next?

⬜ Give specific feedback on the quotes section only, keeping the rest.
⬜ Restate the whole prompt from scratch in a new conversation.
⬜ Ask Claude to summarize the chat and continue in a new conversation.
⬜ Create a Project with a standing instruction about quotes.

A utility company’s billing team redesigned its Claude workflow for drafting customer letters because factual errors were reaching customers. Which measurement best shows whether the redesign worked?

⬜ The number of factual errors that reach the final letters each cycle.
⬜ How long each letter takes now compared with before.
⬜ How many revision rounds each letter needs compared with before.
⬜ A team survey on whether the new workflow feels easier.

A clinic coordinator has refined a detailed, multi-step procedure for turning monthly shift data into a coverage report. She retypes it in every chat, both inside and outside her Projects. What should she do?

⬜ Save the procedure as a Skill.
⬜ Ask Claude at the start of each chat to recall the procedure from earlier chats.
⬜ Upload a document describing the procedure to a Project’s knowledge.
⬜ Add the procedure to her account-wide custom instructions.

A grants coordinator’s Skill formats the nonprofit’s monthly impact reports. In new chats Claude drafts the reports but ignores the format, even when asked to use the Skill by name. What should the coordinator check first?

⬜ That the Skill is switched on in Customize > Skills.
⬜ Whether the Skill’s instructions could be simpler.
⬜ Whether the Skill’s description clearly says when to use it.
⬜ Whether asking for the Skill by name at the start of each chat helps.

A trainer is refining an onboarding handout artifact with Claude. Version two was strong, but three later rounds of feedback made it worse. She wants every version in the same conversation. How can she try a new direction from version two?

⬜ Edit the message that produced version two and describe the new direction.
⬜ Ask Claude to summarize the chat, then continue in a new chat from the summary.
⬜ Copy the version-two text into a new conversation and continue there.
⬜ Highlight the weak passages in the latest version and use Edit with Claude.

An associate sends Claude a one-off request that combines several tasks, and Claude’s very first answer is vague and unhelpful. Which two actions should the associate take next? (select two)

⬜ Send the request again with more relevant background.
⬜ Break the request into smaller steps and run them in order.
⬜ Create a Project and put the request’s details in its instructions.
⬜ Ask Claude to summarize the chat and continue in a new one.

A grants team optimized its monthly reporting workflow in a Claude Project. It set a 30-minute target per run at the start; runs now take 20 minutes, and the output format is consistent. A teammate proposes another tuning round. How should the team respond?

⬜ Keep the current setup and stop optimizing.
⬜ Keep adjusting the Project instructions until the time stops falling.
⬜ Set a lower target and schedule more tuning rounds.
⬜ Audit one more full run to list any remaining manual steps.

A team’s Project produced reliable weekly briefs for months. Lately every new conversation gives noticeably weaker output, although the prompts haven’t changed. What’s the most likely cause to check first?

⬜ Outdated configuration: review the Project’s instructions and knowledge for content that’s gone stale.
⬜ Context overload: start each session with a summary of the previous conversation.
⬜ Wrong model: move every conversation to a more capable model.
⬜ Underspecified prompts: rewrite the weekly prompt from scratch.

An associate traced a recurring flaw in Claude-drafted client memos to a vague instruction in the team’s Project and rewrote it to be more specific. What should the associate do before treating the flaw as fixed?

⬜ Run the memo task again on a comparable request and confirm the output improved.
⬜ Consider it fixed, since updated instructions apply to every new chat automatically.
⬜ Add more instructions for other possible flaws while the configuration is open.
⬜ Delete the conversations that produced the flawed memos.

An associate keeps Claude on the fastest model for quick daily tasks. Detailed, well-specified prompts asking for deep analysis of several long reports keep returning shallow results, however the wording is revised. What should the associate change?

⬜ Use a more capable Claude model for the analysis.
⬜ Start a new conversation and send the same prompt again.
⬜ Break the analysis into many small prompts for the fast model.
⬜ Add stronger wording demanding more depth.

Reviewers comment on a Claude-assisted sponsorship announcement. One comment is unclear, and several others ask for specific changes. Which two responses support a productive revision? (select two)

⬜ Apply every suggestion in a single pass.
⬜ Ask the reviewer what change they mean when a comment is ambiguous.
⬜ Ask Claude to interpret the unclear comment and apply its best guess.
⬜ Track each accepted comment against the change that addressed it.

A housing cooperative runs a weekly feedback cycle for Claude-drafted maintenance briefings. The steps are: (1) run the next cycle and compare with the previous one; (2) choose one or two targeted prompt changes; (3) decide whether to keep iterating, change approach or escalate; (4) capture specific feedback from the briefing review; (5) document the changes, their effects and remaining problems. What’s the right order?

⬜ 2, 4, 1, 5, 3
⬜ 4, 2, 5, 1, 3
⬜ 4, 2, 1, 5, 3
⬜ 3, 4, 2, 1, 5

A weekly sales report built in a Project ran reliably for six months. This week finance renamed two columns and added a region, and the new file replaced the old one in Project knowledge. Now the report omits the region and mislabels two metrics. What’s the best diagnosis?

⬜ Project knowledge is still serving the old spreadsheet.
⬜ The Project instructions still name the old columns and regions, so they no longer match the new file.
⬜ The extra region made the report too long, so Claude drops the last rows.
⬜ The underlying model changed and now reads the same instructions differently.

A recurring Claude workflow re-uploads the same reference files every run and sometimes sends unchecked drafts into a scheduling system. Which two changes most directly fix those problems? (select two)

⬜ Keep the approved reference files in shared Project knowledge.
⬜ Rely on the scheduling system’s own validation to reject bad drafts.
⬜ Add a defined quality check before drafts go into the scheduling system.
⬜ Give each workflow step a clearer name.

After nine revisions, a Claude-drafted policy summary still has an unresolved gap, and the last five rounds made no measurable progress. What’s the right conclusion?

⬜ Recognize diminishing returns and get human help or a substantially different approach to the gap.
⬜ Accept the draft with a note listing the gap.
⬜ Narrow the requirement so it fits what the prompt can achieve.
⬜ Start a fresh chat and repeat the same request.

A consulting team automated research summaries with Claude. Reviewers now check every minor claim sentence by sentence but still sometimes miss material errors that could change a client recommendation. The team wants less review effort without weaker assurance. Which redesign fits best?

⬜ Keep reviewing every sentence with equal intensity.
⬜ Have Claude mark statements as uncertain and review only those.
⬜ Remove review from routine summaries.
⬜ Set risk-based review tiers: verify material claims fully, sample low-risk content and track recurring errors.

Claude produced six campaign concepts that all meet the brief but share the same cautious audience insight, tone and format. The team needs genuinely different directions within brand and compliance limits. What’s the best next prompt adjustment?

⬜ Ask for concepts that vary audience insight, emotional angle, format and creative risk within the brief’s limits.
⬜ Rerun the same prompt several times and rely on natural variation.
⬜ Remove the brand and compliance limits for this round.
⬜ Pick the first concept, since the sameness suggests it’s safest.

A consulting team wants to shorten a recurring research workflow that checks sources, records uncertainties, gets expert approval and then writes a client brief. Each analyst sets up the research questions from scratch every cycle. Which two changes improve efficiency without weakening quality? (select two)

⬜ Replace expert approval with Claude’s confidence statement when the draft looks consistent.
⬜ Draft the brief only from approved findings, and send unresolved claims straight to the relevant expert.
⬜ Drop source checks, since Research already adds citations.
⬜ Reuse a maintained research template with standard questions, evidence fields and built-in checkpoints.

An associate’s prompt gives the audience, objective, channels and constraints for a launch, plus one past campaign as an example. Every idea Claude returns copies that example’s structure and theme with small changes, across several attempts. What’s the most likely cause?

⬜ Too much detail about audience and constraints leaves no room for originality.
⬜ The model is too small for creative work.
⬜ The single example is anchoring the output, and the prompt doesn’t ask for ideas that differ from it.
⬜ Creative tasks can’t be prompted well, so repetition is unavoidable.

A nonprofit is mapping its grant-review workflow to find its capacity constraint. One stage’s queue has grown for six reporting periods in a row because more work arrives than gets finished. Another stage takes longer per case but runs in parallel with no backlog. Which evidence best identifies the bottleneck?

⬜ The stage with the longest average processing time per case.
⬜ The stage staff call most demanding.
⬜ The stage with the most manual steps.
⬜ The stage whose queue keeps growing because arrivals exceed completions.

Three coordinators each rebuild the same weekly planning pack by hand, a checklist, a calendar summary and a status note, all to the same house rules. Results vary between them, and the procedure rarely changes. The steps to package it as a Skill are: (1) test the Skill on recent weeks’ inputs and compare with approved packs; (2) assign an owner to update the Skill when the procedure changes; (3) write down the procedure, formatting rules and example outputs; (4) share the Skill with the coordinators and retire the manual steps; (5) create the Skill from those instructions and examples. What’s the right order?

⬜ 5, 3, 1, 4, 2
⬜ 3, 5, 1, 4, 2
⬜ 3, 5, 4, 1, 2
⬜ 3, 1, 5, 4, 2

An admissions-support team wants to optimize its recurring Claude drafting workflow. Which description best captures what optimization means here?

⬜ Measure time saved per draft and report it as the main result.
⬜ Merge all review stages into one final check.
⬜ Remove redundant and low-value work, use Claude where it adds measurable value and keep the quality checks that protect each admission decision.
⬜ Use Claude for as many steps as possible.

---