Skip to content
Domain 1Task 1.1

Creating Effective Prompts for Business and Technical Tasks

Learn this interactively

What You Need to Know

A prompt is the brief you give Claude: the task and everything needed to do it. Briefing well isn't a technical skill; it's the one you use to delegate work to a colleague. Every other objective in this certification builds on it. A good brief gives the other six domains something to work with, and no feature, model choice or troubleshooting step rescues a bad one.

A roofing firm's office manager wants Claude to draft a one-page update on this month's jobs for its two directors. The one-line version is "write something about this month's jobs".

The Briefing Mindset

Anthropic's guidance is to treat Claude as a brilliant but new employee: highly capable, with no context on your company, project or standards until you supply it. The Help Center gives chat users the same picture: a newly hired contractor who needs very specific instructions. It advises explaining a failing prompt as if to someone with no background knowledge of what you are asking.

An effective prompt contains as much context as possible, because Claude works from what it is given. Like a good delegation, it covers the task, background, audience, format and what good looks like. Neither a new hire at the roofing firm nor Claude knows which jobs ran late.

The Five Components

A complete business prompt has five parts.

  1. Task. Say precisely what you want done, with a verb and a deliverable. "Draft a one-page project update for the steering committee" is a task. "Help me with my project" is not. The office manager's task is a one-page update on this month's jobs for the directors.
  2. Context. The background Claude can't know: who the audience is, what the document is for, what has already been decided, and any constraints that apply. For a writing task, the prompt outlines the requirements, the target audience and the main points.
  3. Source material. Paste or attach the actual inputs, such as the report, the meeting notes or the data. Do not describe them from memory. Claude works from what it is given, and a from-memory summary adds errors and omissions the real document doesn't have. claude.ai accepts documents, spreadsheets and images so the real inputs can go in. The roofing firm attaches its job sheets instead of recalling them.
  4. Output specification. Format, length, tone and structure: three bullet points per workstream, neutral tone, under 400 words.
  5. An example, when standards matter. One sample of a previous document in the house style beats three paragraphs describing the style. Anthropic calls examples one of the most reliable ways to steer format, tone and structure. Last month's update is the roofing firm's sample.

Role and Audience Framing

Opening a prompt with a role, such as "As an HR communications specialist", shapes vocabulary, tone and priorities. Anthropic notes that even a single sentence makes a difference.

Naming the audience changes the output more than most instruction tweaks. The same analysis should read differently for the board and the project team. Claude shapes it if told who is reading, so the roofing firm's brief names the directors. Explaining the motivation behind a requirement also gets a more targeted response.

What Specificity Buys

A vague prompt produces generic output, which needs heavy rework. Being specific about the desired output format and constraints improves Claude's results. A minute on the brief saves a rework cycle. The official usage guidance agrees: be specific and concise, and review and edit your prompt before sending. A well-briefed first attempt costs less time and less usage than repeated corrections. The roofing firm's one-liner gives no job sheets, readers or format.

Technical Tasks at Associate Level

Technical doesn't mean code. At this level it means asking Claude for a spreadsheet formula, a first pass at simple data analysis, or the first draft of a technical document. The skill is the same: brief the task well. The office manager might write: I need a formula that flags any invoice more than 30 days overdue, here is a sample of the data.

In the exam

The recurring wrong answer in this objective is the one-line prompt, "write something about our Q3 results", presented as acceptable because Claude will figure it out. It cannot: none of the five components is there to work from. The formula brief above has done the Domain 1 work; how elegant the formula is was never the question.

Key Concept

When an exam item compares prompts, count the components. The keyed answer almost always supplies the missing task, context, source material, output specification or example. The distractors polish wording and leave the gap open.

Exam Traps

Exam Trap

Accepting the one-line prompt because Claude will figure it out

Claude has none of the context a colleague would need, so the output is generic and the user pays for the skipped brief in rework. The keyed answer briefs the task properly; specificity up front is the documented fix.

Exam Trap

Choosing the option that polishes wording while the real gap goes unfilled

Stems in this objective usually point at a missing component: no audience, no source material, no format. A cosmetic instruction tweak leaves that gap open, which is exactly why it is the tempting second-best rather than the answer.

Exam Trap

Writing the deliverable yourself and using Claude only to proofread

It looks like diligence, but it under-uses the tool. Drafting from a good brief is exactly the work Claude does well; a proofreading-only role leaves the time saving on the table, and the exam treats it as the lazy under-use posture.

Practice Scenario

A facilities coordinator needs Claude to draft a staff memo introducing a new desk-booking system to all office-based staff. She has the supplier's how-to guide and wants a draft she can send without reworking it by hand or going back to Claude for more rounds. Which prompt sets the work up best?

Build Exercise

Build Exercise

Write the Five-Component Brief

Difficulty
20 minutes

What you'll learn

  • How to state a task as a verb plus a deliverable
  • How to supply the context Claude cannot know
  • How to attach real source material instead of describing it
  • How to specify format, length, tone and structure
  • How to use one example document to set the standard
  1. Pick a real recurring task from your own job: a report, a briefing or an email you produce every week

    Why: A task you know well makes gaps in the brief visible; you can tell instantly when the output misses your standard.

    You should see: One line naming the task, the deliverable and how often it recurs.

  2. Write the task line as a verb plus a deliverable, then add the context Claude cannot know: the audience, the purpose, what has already been decided, any constraints

    Why: The task line and the context are the two components most often missing from weak prompts, and the two cheapest to add.

    You should see: A task sentence plus three or four lines of context a new colleague would need on day one.

  3. Attach or paste the actual source material: the report, the notes, the data, not a from-memory summary of them

    Why: Claude works from what it is given; a description from memory adds errors and omissions the real document does not have.

    You should see: The real files or pasted text sitting in the conversation alongside your brief.

  4. Specify the output: format, length, tone and structure, and attach one example of a previous document in the house style

    Why: Examples are among the most reliable ways to steer format, tone and structure; one sample beats paragraphs of description.

    You should see: An output specification of one or two lines plus an attached exemplar.

  5. Run the prompt and compare the draft against what your usual first attempt looks like, noting which component earned its place

    Why: The comparison shows what each part of the brief bought, which is how the briefing habit sticks.

    You should see: A first draft noticeably closer to publishable than a one-line prompt would produce.

Sources