# Identifying Appropriate vs Inappropriate AI Use Cases > Before building a workflow around Claude, judge whether a task is one Claude is allowed and well-suited to do, not just whether it is technically possible. Anthropic's Usage Policy designates legal, financial, and employment-related tasks as "high-risk," requiring human-in-the-loop oversight and AI disclosure specifically when the output is consumer-facing; a separate, small set of categories (surveillance and biometric profiling, weapons design, malicious cyber operations, deceptive political content) is prohibited outright regardless of who consumes the output. **Domain:** CCAOF-D6 · Governance, Risk & Responsible Use (15% of the exam) **Canonical:** https://claudearchitectcertification.com/concepts/appropriate-ai-use-cases **Last reviewed:** 2026-05-04 ## Quick stats - **High-risk categories named:** 3 (legal, financial, employment) - **Exam domain:** CCA-A D6 (15%) - **High-risk safeguards:** 2 (human review + AI disclosure) - **Safeguard trigger:** Consumer-facing output - **Prohibited category groups:** 4 - **Task-suitability checks:** 3 (specified, verifiable, error-tolerant) ## What it is Before building a workflow around Claude, a business user has to judge whether the task is one Claude is *allowed* and *well-suited* to do, not just whether it is technically possible. These are two separate questions. Anthropic's Usage Policy is the authoritative framework for permissibility - what should and should not run on Claude at all, applying to everyone using Anthropic's products, consumer and enterprise alike. Passing the policy check is necessary but not sufficient: a task can be fully permitted by policy and still be a poor fit for an LLM in practice, which is a separate, task-suitability judgment about the nature of the task itself. The policy sorts tasks along two independent axes an associate needs to check separately: whether the task category itself is designated high-risk or prohibited, and whether the specific output reaches an end user directly or gets reviewed and re-packaged first. A task can be high-risk and low-risk in effect depending on that second axis, while a prohibited category stays prohibited no matter how it is delivered. On top of that, a third, independent check applies regardless of policy category: is the task itself well-specified, verifiable, and tolerant of the kind of error an LLM can make? A permitted task that is vague, unverifiable, and error-intolerant (e.g., an ungoverned final decision with no review step) is still a bad candidate for Claude even though nothing in the Usage Policy blocks it. ## How it works High-risk categories are named explicitly. Anthropic designates certain categories as "high-risk use cases," ones with public welfare and social equity implications, explicitly including legal, financial, and employment-related uses of Claude. Landing in one of these categories does not mean the task is off-limits, it means extra conditions attach before it ships. Those extra safeguards are human-in-the-loop oversight and AI disclosure. High-risk use cases carry additional safeguards: a human-in-the-loop oversight path and a requirement to disclose to end users that they are interacting with an AI system. The August 2025 update scoped those safeguards to consumer-facing output. Anthropic clarified that the extra requirements apply specifically when the model's output is consumer-facing, not for business-to-business interactions. An internal analyst using Claude to draft an employment-policy summary for HR sits in a different risk tier than a consumer-facing chatbot giving individualized legal or financial advice, even though both are "employment" or "legal" in category. A separate set of categories is prohibited outright, regardless of consumer-facing or B2B framing. The policy prohibits surveillance, tracking, profiling, and biometric-identity inference; weapons design; malicious cyber operations; and deceptive political or election content, no safeguard level makes these appropriate. Law enforcement use is permitted for a carefully tailored set of use cases (back-office support, document summarization) but restricted for surveillance or profiling, showing the two axes, category and delivery context, apply even within one type of user. The Usage Policy is a living document. Anthropic frames it as revised periodically as capabilities and regulation evolve, so current guidance can date quickly, an architect or business user should verify against the live policy page before quoting specifics rather than relying on a memorized snapshot. Task-suitability is a separate check, and policy-permitted does not mean well-suited. Even inside a fully allowed category, ask three things about the task itself: is it well-specified (a clear, bounded definition of what a correct output looks like, versus an open-ended, ambiguous ask), is it verifiable (can a human or a downstream system actually check the output before it has consequences, versus a one-shot answer nobody reviews), and is it tolerant of the kind of error an LLM can make (a wrong draft that gets edited is low-cost; a wrong number that silently drives a payment is not). A task can clear every Usage Policy check and still be the wrong fit for Claude if it fails these three - the policy answers "is this allowed," not "is this a good idea." ## Where you'll see it in production ### Employee benefits chat widget Consumer-facing eligibility Q&A is employment-related and reaches the employee directly, so it needs a human-in-the-loop review path for edge cases and a clear AI-disclosure notice. ### Internal HR FAQ drafting The same underlying task, used internally with an HR specialist reviewing before publishing, is a materially lower-risk, B2B-style use of Claude even though the subject matter is identical. ## Decision tree 1. **Does the task fall into an explicitly prohibited category (surveillance, tracking, profiling, biometric-identity inference, weapons design, malicious cyber operations, deceptive political or election content)?** - **Yes:** Stop, this is prohibited regardless of consumer-facing or B2B framing; no safeguard level makes it permitted. - **No:** Not prohibited outright, continue to the high-risk check. 2. **Does the task fall into a designated high-risk category (legal, financial, or employment-related)?** - **Yes:** It may need the high-risk safeguards, check whether the output is consumer-facing to know if they apply. - **No:** Standard usage-policy terms apply, no high-risk safeguard requirement from this axis. 3. **Is the model's output consumer-facing (an end user is the direct recipient), rather than reviewed and re-packaged by an internal analyst first?** - **Yes:** If also high-risk, the additional safeguards, a human-in-the-loop review path and a clear AI-disclosure notice, are required as of the August 2025 policy clarification. - **No:** Same underlying task, lower risk tier, an internal reviewer re-packaging output before anyone else sees it is a materially different, B2B-style use even when the task category is identical. 4. **Is this a law-enforcement use case?** - **Yes:** Permitted only for a carefully tailored set of use cases such as back-office support or document summarization, restricted for surveillance or profiling regardless of the other axes. - **No:** This axis does not apply. 5. **Assuming the task clears the Usage Policy, is it well-specified, verifiable, and tolerant of the errors an LLM can make?** - **Yes:** Policy-permitted AND well-suited - proceed. - **No:** Policy permission alone doesn't make this a good use case; redesign the task (narrow the spec, add a verification/review step, lower the blast radius of an error) before treating it as appropriate. ## Exam-pattern questions ### Q1. A benefits team wants Claude to answer employee health-plan eligibility questions directly inside a public-facing chat widget. Is this appropriate, and under what condition? Yes, but only as a high-risk use case requiring a human-in-the-loop review path for edge cases and a clear AI-disclosure notice, because eligibility determination is employment-related AND the chat widget delivers the output directly to the employee. Named distractor: "this is fine without extra safeguards since employees aren't the general public" - the safeguard trigger is consumer-facing delivery, not whether the audience is internal or external to the company. ### Q2. The same team instead uses Claude internally to draft a first-pass FAQ document that an HR specialist reviews before publishing. How does the risk tier change? It drops to a materially lower, B2B-style risk tier, the August 2025 clarification scopes the extra human-in-the-loop and AI-disclosure requirements to consumer-facing output, and here an HR specialist reviews and re-packages the draft before an employee ever sees it. Named distractor: "the task is employment-related so the same high-risk safeguards still apply regardless of the review step" - Anthropic's own clarification ties the extra safeguard requirement to who consumes the output directly, not to the task category alone. ### Q3. A consultant wants to build a public-facing tool that infers a person's identity from photos for a retail loss-prevention client. What's the correct call? This is prohibited outright, the Usage Policy bans surveillance, tracking, profiling, and biometric-identity inference regardless of whether the use is framed as consumer-facing or B2B, so no safeguard level makes it appropriate. Named distractor: "add human-in-the-loop review and an AI-disclosure notice to bring it into policy" - those are the high-risk safeguards for a different category; this category is prohibited, not merely high-risk. ### Q4. A law-enforcement agency wants Claude for both back-office report summarization and real-time crowd-surveillance profiling. How does the Usage Policy treat these two requests? Differently, back-office support and document summarization sit inside the carefully tailored set of use cases law enforcement is permitted, while surveillance and profiling use is restricted under the same prohibited-category rules that apply to anyone. Named distractor: "law-enforcement status exempts both requests from the Usage Policy's restrictions" - the agency's status does not blanket-exempt a prohibited category. ### Q5. A team assumes that because Claude can technically generate a legal-sounding contract clause, using it to give individualized legal advice directly to consumers is automatically an appropriate use case. What's the flaw? Technical capability is not the test, individualized legal advice delivered directly to consumers is a high-risk, consumer-facing use requiring human-in-the-loop oversight and AI disclosure. Named distractor: "if Claude generates a correct-sounding answer, the use case is automatically appropriate" - capability and appropriateness are separate questions, and this is the exam's core trap. ### Q6. A task clears the Usage Policy cleanly - it's not prohibited, not high-risk, and B2B-internal. The task itself is "summarize what the customer probably wants" with no defined success criteria, no review step, and its output auto-triggers an irreversible account action. Is this an appropriate use case? No - passing the Usage Policy check only establishes permissibility, not suitability. This task fails all three task-suitability criteria: it isn't well-specified (no clear definition of a correct output), isn't verifiable (nothing checks the output before it acts), and isn't error-tolerant (a wrong guess triggers an irreversible action). Named distractor: "since the task isn't prohibited or high-risk, it's an appropriate use case" - conflates policy permissibility with task-suitability, which is a separate judgment about the task's own design. ## FAQ ### Q1. Does the Usage Policy apply differently to enterprise customers versus individual consumers? No. The Usage Policy is the authoritative should-and-shouldn't framework and applies to everyone using Anthropic's products, consumer and enterprise alike; what changes the safeguard requirement is whether a given output is consumer-facing, not the account tier. ### Q2. Are legal, financial, and employment tasks always prohibited? No, they're designated high-risk, not prohibited. They're permitted with additional safeguards, human-in-the-loop oversight and AI disclosure, when the output is consumer-facing; a separate, smaller set of categories (surveillance and biometric profiling, weapons design, malicious cyber operations, deceptive political content) is prohibited outright. ### Q3. If the Usage Policy allows a task, is it automatically an appropriate use case for Claude? No. Policy permissibility and task-suitability are separate checks. A permitted task can still be a poor fit if it isn't well-specified, isn't verifiable, or isn't tolerant of the kind of error an LLM can make - those are design questions about the task itself, not questions the Usage Policy answers. --- **Source:** https://claudearchitectcertification.com/concepts/appropriate-ai-use-cases **Last reviewed:** 2026-05-04 **Evidence tiers**, 🟢 official Anthropic doc / API contract · 🟡 partial doc / inferred · 🟠 community-derived · 🔴 disputed.