What You Need to Know
Two questions decide whether a task goes to Claude. Fit asks whether Claude would do the work well: the right shape, the right volume, output someone can check. Domain 4 covers fit. Appropriateness asks whether the task should be delegated to Claude at all. It looks at different evidence: the stakes, the authority the work requires, and who is accountable for the outcome. A task can be a perfect practical fit and still be work that only a qualified human should do.
Fit Is Not Permission
This lesson uses one made-up example throughout: a law firm. Its office manager has two jobs to do. One is a long series of client feedback forms to summarise into themes. The other is a stack of applications for an office assistant's job, to score so that the office manager can pick the person to hire. Scoring the applications is an excellent practical fit: structured, voluminous, checkable. It can still be inappropriate, because of what the scoring decides and who cannot challenge it. The wrong answer that appears often bases its reasoning on quality: Claude would do the task quickly and well, so it must be fine to delegate. The quality of the output does not decide whether the task is appropriate.
The Four Inappropriate Categories
Four recognisable categories make a use inappropriate.
The first is work that needs professional authority: legal, medical, financial and safeguarding judgements. Safeguarding means protecting children and adults at risk from harm. A qualified person must make these and be accountable for them. Claude can help someone prepare, summarise or understand. Claude's recommendation becomes advice only after a qualified person reviews it. At the law firm, legal advice to a client must come from a qualified lawyer. The Usage Policy names the settings where this category applies most. Its High-risk Areas include legal and medical advice, finance, credit, insurance, housing, employment, education and healthcare access. In these areas a qualified person must review Claude's recommendation before it is given as advice or used to make a decision. The policy also states: "The qualified person remains responsible for the accuracy and appropriateness of what is delivered or decided." The policy excludes wellness content, such as education on sleep, stress, nutrition or exercise.
The second is a consequential decision about identifiable people: who gets the job, the award or the penalty. Poor output is not the reason. The person affected needs a decision that an accountable person made and can explain. Many of these decisions are in the Usage Policy's High-risk Areas. For example, Employment includes "screening, ranking, advancing, or rejecting candidates; hiring". In these areas, Claude's recommendation can be part of the decision only with two controls. First, a qualified person meaningfully reviews it, with the authority to change it, and stays responsible for the result. Second, the person affected is told that AI was used. Lesson 2.4 explains the review rule, and lesson 6.4 explains the disclosure rule.
Meaningful review means that the reviewer makes their own judgement. A reviewer who only approves each result at the end leaves the decision with Claude. The scoring of the office assistant applications belongs in this category. If Claude scores the applications and the office manager only approves the scores, she cannot explain the result as her own decision. The usual answer is that Claude prepares the material and the person decides. The policy's requirements do not apply to "Internal drafting, research, summarization, or analysis that is not the ultimate High-risk AI Recommendation delivered". A High-risk AI Recommendation is Claude's recommendation in a High-risk Area.
The third is fabrication presented as real: invented quotes, testimonials, case studies or evidence, such as client testimonials the firm writes itself. Plausibility makes fabrication worse, because a convincing invention is easier to believe. The Usage Policy's prohibitions on misinformation and deceptive practices are the basis of this category.
The fourth is anything the Usage Policy or the organisation's own policy prohibits. Lesson 6.3 covers the organisation's own policy. The Usage Policy's Universal Usage Standards are prohibitions that apply to every user. Office work meets these most often:
- Identifying or tracking people without their consent. The policy names facial recognition. The ban is on identifying or tracking a person. So the office manager can ask Claude when and where people gather in images from the reception camera, but not who they are. Claude also refuses to name people in images.
- Using other people's work without the right to use it. The policy prohibits using Claude to "Infringe, misappropriate, or violate the intellectual property rights of a third party". To misappropriate is to take something for your own use without the right to do so. For example, the firm cannot have Claude write out song lyrics for the screen at a staff party when its licence does not cover those songs.
- Fake reviews, invented personas and bots that claim to be human. These overlap with fabrication. The policy prohibits them because they mislead people about who wrote the content, or whether they are talking to a person.
- Misusing private information. Lesson 6.2 covers this.
Review does not make a prohibited use acceptable. The review rule applies only to uses that the policy allows.
The Substitution Test
One test decides whether a task is appropriate. Picture a capable, well-meaning colleague who has:
- no professional qualification in the area
- no access to the person's history
- no accountability for the outcome
Would you accept this piece of work from them? The test changes an abstract question about appropriateness into an ordinary judgement about delegation. Every professional already makes that judgement. The feedback summary passes: the same colleague could write it, and the office manager would check it. Scoring that the office manager only approves fails on all three grounds. The framework in Anthropic's AI Fluency course names this skill Delegation.
Most Work Is Appropriate
The broad range of appropriate uses matters as much as the four categories. Official documentation describes a deliberately broad range of appropriate uses: writing across formats, summarising long documents, brainstorming, analysing images and data, translating. Drafting, reformatting, analysing material you hold, and preparing someone to make a decision are all appropriate. The inappropriate categories are narrow and specific. If you treat everything near them as forbidden, you abandon work that a simple safeguard would enable. The office manager can still ask Claude to summarise each application, then read the applications and choose.
In the exam
This objective asks a different question from Domain 4. The substitution test is one way to check the first two categories. For the third and fourth, you need to know what counts as fabrication and what the policies prohibit. Too much caution is also a mistake: when a safeguard lets the work go ahead, refusing the work gains nothing.
What to notice in the exam
A quick sign-off at the end does not make a delegated decision acceptable. In a High-risk Area, the Usage Policy states: "A qualified person must meaningfully review a High-risk AI Recommendation, with the authority to change or modify it, before it is delivered as advice". A reader who approves each result without weighing the case does not meet this rule. So an option that only adds a reviewer at the end does not fix the delegation. The better option has Claude prepare the material and the accountable person decide.
Checked 29 October 2026: the version of the Usage Policy quoted in this lesson takes effect on 12 November 2026.
Anthropic's AI Fluency course defines Delegation as deciding what work to do with AI and what to do yourself. This lesson's reading, "deciding what to hand to AI and what to keep", follows that definition. The Help Center's use-case article lists what Claude is for and contains no prohibitions. The inappropriate categories are based on the Usage Policy and the exam guide.
Exam Traps
Adding a human reviewer to an inappropriate delegation and treating that as the fix
A reviewer who only approves a decision made elsewhere has not made the judgement and cannot give their own reasons. The Usage Policy asks for meaningful review with the authority to change the result. Let Claude prepare, and keep the decision with an accountable person.
Accepting the argument that a use is fine because the results stay inside the organisation
Keeping an output in-house restores neither accountability for the decision it contains nor the affected person's ability to challenge it. Keep the decision with a named, accountable person.
Declining both halves of a mixed proposal
Declining everything looks cautious but loses help Claude can appropriately give. When a proposal mixes appropriate and inappropriate work, keep the appropriate half and decline only the inappropriate one.
Practice Scenario
A university department manager is asked to use Claude on the annual student survey. Two proposals arrive together: summarise three hundred free-text responses into themes for the course review, and score individual students' responses to flag which should be called to an adviser meeting. The course review needs the themes. University policy allows Claude on the survey text and bars any tool from scoring students for adviser meetings. What should the manager do?
Build Exercise
Sort a Fortnight of Real Work
What you'll learn
- How to run the substitution test on real tasks
- How to name which of the four categories makes a task inappropriate
- How to write the enabling safeguard for a borderline task
- How to check your own judgement against the mistake of being too cautious
- List the last ten pieces of work you gave Claude, or considered giving it, plus two you would never hand over
Why: The test needs real delegations to work on; invented examples make the results look better than they are and hide your actual habits.
You should see: A list of twelve tasks in your own words, each one line long.
- Run the substitution test on each: would you accept it from a capable colleague with no qualification in the area, no access to the people involved, and no accountability for the outcome?
Why: The test converts appropriateness into a delegation judgement you already know how to make.
You should see: A pass or fail against each task, with the failing ground named: qualification, history or accountability.
- Sort the list into appropriate and inappropriate, and for each inappropriate item name which of the four categories it falls under
Why: Naming the category is the exam skill: questions reward candidates who can say why a use is inappropriate, not just that it feels risky.
You should see: Each inappropriate task labelled with one category: professional authority, consequential decision, fabrication, or policy prohibition.
- For any borderline item, write the enabling safeguard that would let it proceed: what changes, the material, the route, or the authorisation
Why: The right answers almost always make the work possible; the professional skill is finding the safeguard, not finding reasons to refuse.
You should see: A one-line safeguard per borderline task, specific enough to act on tomorrow.
- Check the appropriate pile is the larger one; if it is not, check your test results again and remember that most work is appropriate
Why: Over-caution is the mistake that the wrong options in this objective are based on; a mostly-inappropriate list usually means the test was run too harshly.
You should see: An appropriate pile of eight or more of the twelve tasks, or a written reason why your work is genuinely exceptional.
Sources
- Usage Policy — Anthropic
- Vision — Anthropic
- What are some things I can use Claude for? — Anthropic
- AI Fluency: Framework and Foundations: The 4D Framework — Anthropic
- Claude Certified Associate – Foundations Exam Guide — Anthropic. The Help Center article describes appropriate uses but lists no prohibitions. The inappropriate categories come from the Usage Policy and the exam guide's objective.