What You Need to Know
A task type is the kind of work you hand to Claude. This objective covers four: analysis (finding what material shows), research (gathering from sources), drafting (writing a piece) and brainstorming (generating ideas). Anthropic's documentation shows Claude doing all four: analysing chart data, drafting anything from a text message to a screenplay, and generating viewpoints and recommendations as an ideas partner. One prompting style does not fit all four. Each type rewards a different briefing shape, meaning what the prompt contains.
A lettings agency manager has four jobs this month. She must find out why tenancy renewals fell, compare other agencies' tenant screening, write to landlords about a new fee, and find ways to win more landlords.
Analysis: Material Plus Questions
Analysis starts with the material. Attach the spreadsheet or paste the survey results, because Claude works from what it is given. Uploads carry the real artefacts, even documents hundreds of pages long.
Then name the questions the analysis must answer and the decision it will inform. "Analyse this data" gives Claude no target, so it describes rather than analyses. The manager attaches her renewals spreadsheet and asks why renewals fell, which properties lost most tenants, and what should change before next year. Ask for main findings, supporting evidence and caveats, and for how Claude reached each conclusion, so the reasoning can be checked. Checking that reasoning is Domain 2 work.
Research: A Scoped Question, Not a Topic
For multi-source questions, Research works agentically: Claude runs multiple searches that build on each other and returns a report with easy-to-check citations. For a quick current fact, an ordinary web search suffices. The documented dividing line: web search suits straightforward factual queries answered in one or two lookups, while Research suits thorough gathering across many searches.
Either way, brief a research question, not a topic. State the scope, the angle, and what the findings must inform. The manager asks Research how agencies of her size locally screen tenants, to decide whether to change hers. A bare subject gives the searches no direction and returns an unfocused report. Then verify the citations that matter.
Drafting: The Full Brief
Drafting needs the full five-component brief from objective 1.1: audience, purpose, tone, length and format. Add an exemplar when house style matters, because an example is one of the most reliable ways to steer format, tone and structure. For the fee letter, the manager names landlords as the audience and attaches a past letter. Expect to iterate (objective 1.3). On anything long, stage the work (objective 1.2): outline first, then sections.
Brainstorming: Wide, Then Narrow
Brainstorming reverses the drafting approach: open wide first, asking for quantity and range, such as twenty ideas including some unconventional ones. To get beyond the obvious, ask for it explicitly rather than leaving Claude to infer it. Withhold judgement, because heavy constraints cut the variety that is the whole point. Then converge in a second step: apply your criteria, shortlist, and develop the survivors. The manager asks for twenty ways to win landlords and holds her budget back for the shortlist. This makes Claude a genuine ideas partner instead of a generator of three safe suggestions.
The Diagnostic Habit
Name the type, then pick the shape.
- Analysis: material plus questions.
- Research: a scoped, cited question.
- Drafting: a full brief plus iteration.
- Brainstorming: wide, then narrow.
In the exam
The exam guide names these four task types, and the tested skill is diagnostic: read the workplace situation, name the type, brief in its shape. Wrong answers cross the wires: a tightly constrained brainstorm, a one-line research topic, an analysis with no material attached and a draft with no audience. The last is the recurring wrong answer on drafting. Choosing which product feature or model to use is Domain 3 territory; this objective covers the briefing that follows.
Watch for crossed wires
Most items hide the error in a plausible plan that applies the right briefing to the wrong task type. Check each part of a multi-task option separately: one crossed wire, such as constraints at a brainstorm's divergence step, sinks the option.
Brainstorming's diverge-then-converge sequence is standard facilitation practice applied to Claude, not a rule from a product manual.
Exam Traps
Asking for an analysis without naming the questions it must answer
Material with no questions attached produces a description of the data, not an analysis. The professional brief supplies the material, the questions, and the decision the findings will inform, and asks for findings, evidence and caveats.
Opening a brainstorm with tight constraints and a fixed number of acceptable directions
Constraints applied at the divergence step remove the variety the exercise exists to produce. Criteria belong in the convergence step, where they shortlist a wide field rather than strangling it before it exists.
Briefing research as a topic rather than a scoped question
An unscoped topic returns an unfocused report. A research question states the scope, the angle and what the findings must inform, which tells the searching what to investigate and gives the citations something to support.
Practice Scenario
A university department administrator has four tasks this week: work out why tutorial attendance dropped this term; build a picture of how three other universities run applicant open days; write the new-term welcome letter to students; and generate ideas for lifting open-day attendance. She intends to use Claude on all four. Which working plan matches strategy to task?
Build Exercise
Brief All Four Task Types
What you'll learn
- How to recognise the task type in a real piece of work
- How to brief an analysis with material, questions and the decision it informs
- How to scope a research question with an angle and a purpose
- How to open a brainstorm wide and converge in a second step
- List four real tasks from your own week, one per type: an analysis, a research question, a document to draft, and a problem needing ideas
Why: The diagnostic habit only forms against real work; the four types look obvious on paper and blur in practice.
You should see: Four one-line task descriptions, each labelled with its type.
- Write the analysis brief: attach the actual material, then name the questions it must answer and the decision it informs
Why: Material plus questions is what separates an analysis from a description of the data.
You should see: A brief with the real file attached and two or three named questions.
- Write the research brief as a scoped question with an angle and a purpose, and note whether it warrants Research or a quick web search
Why: A scoped question directs the searching; a bare topic returns an unfocused report.
You should see: A research question stating scope, angle and what the findings must inform, plus a one-line tool call.
- Write the drafting brief with the full five components plus an exemplar, and the brainstorming brief asking for quantity and range with criteria held back
Why: These two shapes are opposites: the draft wants full specification, the brainstorm wants deliberate openness.
You should see: A complete drafting brief with an exemplar attached, and a brainstorming ask for twenty ideas including unconventional ones.
- Run all four, then note where a wrong-shaped brief would have cost you a rework cycle
Why: Seeing the cost of a crossed wire on your own work is what makes the diagnostic habit permanent.
You should see: Four outputs plus a short note per task naming the crossed wire you avoided.
Sources
- Use research on Claude — Anthropic
- When should I use web search, extended thinking, and research? — Anthropic
- What are some things I can use Claude for? — Anthropic
- Upload files to Claude — Anthropic
- Prompting best practices — Anthropic