How to Write Effective Claude Prompts: Practical Examples for 7 Professional Roles

Claude can write an email in seconds, summarize a document, analyze customer feedback, generate ideas, review code, or help prepare a presentation.

But there is a big difference between asking Claude to “write something” and giving it enough information to produce something you can actually use.

The difference is usually not the AI model itself. It is the quality of the instruction.

A prompt that simply says:

“Write a marketing strategy for my company.”

leaves Claude guessing.

What kind of company?
Who are the customers?
What is the business goal?
What market are you targeting?
What resources do you have?
What should the final strategy look like?

A better prompt gives Claude the information it needs to make useful decisions.

This guide explains a practical approach to writing Claude prompts and, more importantly, shows how prompting varies by professional role.

The examples below are designed around real workplace situations rather than abstract prompt-engineering exercises.

What Makes a Claude Prompt Actually Useful?

There is no single “magic formula” that makes every prompt work.

However, strong prompts usually give the model five things:

1. A clear task
2. Relevant context
3. A specific objective
4. Constraints or criteria
5. A clear output format

Anthropic’s own prompting guidance emphasizes clear instructions, useful context, examples, and explicit output requirements.

Think about it this way:

Weak prompt

Write a LinkedIn post about AI.

Better prompt

You are a B2B content strategist. Write a LinkedIn post for HR leaders at mid-sized technology companies. The topic is how AI is changing employee onboarding. The goal is to generate discussion rather than promote a product. Use a professional but conversational tone and finish with one specific question for the audience.

The second prompt gives Claude a direction.

It doesn’t simply ask for text. It explains who the audience is, what the content is about, what the objective is, and what the response should look like.

That distinction becomes even more important when you use Claude for professional work.

The 5-Part Prompt Structure I Recommend

For most professional tasks, start with this structure:

1. Role

Tell Claude what perspective or expertise it should use.

For example:

Act as a senior B2B marketing strategist.

or:

Act as an experienced technical recruiter specializing in software engineering roles.

The role is useful when the task requires a particular professional perspective.

2. Context

Give Claude the information it would otherwise have to guess.

This might include:

• company information
• target audience
• project background
• customer information
• existing content
• goals
• constraints
• examples

For example:

Our company sells HR software to technology companies with 100–500 employees. We are launching a new employee onboarding feature next month.

This is much more useful than simply saying:

Write an announcement for our new feature.

3. Task

State exactly what you want Claude to do.

Instead of:

Help me with recruitment.

Try:

Review the job description below and identify the three most important skills candidates should demonstrate during the first interview.

The more concrete the task, the easier it is to evaluate the response.

4. Criteria

Explain what a good result should achieve.

For example:

Prioritize practical experience over theoretical knowledge. Avoid generic recommendations. If information is missing, identify the gap instead of inventing an answer.

This is particularly useful for professional work where accuracy matters.

5. Output format

Tell Claude how you want the answer presented.

For example:

Return the result as a table with four columns: Skill, Why It Matters, Interview Question, and Strong Answer Signals.

This small addition can dramatically improve usability.

Anthropic also recommends explicitly describing the desired output and using examples or structured formatting when you need consistent results.

A Simple Prompt Formula

You can combine the five elements into a reusable structure:

Act as [ROLE].
Context: [BACKGROUND INFORMATION]
Task: [WHAT YOU WANT CLAUDE TO DO]
Criteria: [WHAT A GOOD RESULT SHOULD ACHIEVE]
Output: [FORMAT YOU WANT]

You don’t need to use this exact wording every time.

The important idea is to give Claude enough context to understand the task, rather than expecting it to infer your intentions.

Example 1: HR and Recruitment

Recruiters often use AI for job descriptions, candidate screening, interview preparation, and employer branding.

But a prompt such as:

Write interview questions for a marketing manager.

is too broad.

Claude doesn’t know what kind of marketing manager you need.

A better prompt

Act as a senior recruitment specialist with experience hiring marketing leaders for B2B SaaS companies.
We are interviewing candidates for a Marketing Manager position in a 70-person SaaS company. 
The person will own content marketing, LinkedIn, email campaigns and basic demand generation. They will work with a small team and report directly to the Head of Marketing.
Create 12 interview questions divided into four categories:

• strategic thinking
• practical marketing skills
• analytical thinking
• collaboration and communication

For each question, provide:

1. What the question is designed to evaluate
2. What a strong answer should demonstrate
3. One potential red flag

Keep the questions practical rather than theoretical.

Why this works better

The prompt gives Claude:

• the hiring context
• company size
• role responsibilities
• reporting structure
• evaluation criteria
• desired format

That means the resulting questions are much more likely to resemble an actual interview rather than a generic list copied from a recruitment blog.

Reusable HR prompt

Act as an experienced [RECRUITMENT/HR] professional.
Context: [COMPANY + ROLE + SENIORITY + TEAM]
Task: [SPECIFIC HR TASK]
Evaluate the result based on: [CRITERIA]
Format the response as: [FORMAT]

Example 2: Marketing Manager

Marketing professionals often ask AI to create content.

But “write me a LinkedIn post” is rarely enough.

Imagine you are promoting a new AI course.

Weak prompt

Write a LinkedIn post promoting our AI course.

Better prompt

Act as a B2B content strategist who specializes in professional education and AI.
We run an online platform that helps professionals discover AI courses, tools and learning resources. 
Our audience includes marketers, HR professionals, managers, analysts and technology professionals.
We are publishing a new guide about practical AI skills for non-technical professionals.

Create a LinkedIn post that:

• starts with a strong observation rather than a promotional headline
• explains why AI literacy is becoming relevant outside technical roles
• gives readers one useful insight they can apply immediately
• avoids exaggerated claims such as “AI will replace everyone”
• sounds like a knowledgeable professional sharing an observation
• ends with one natural question

Keep it under 180 words.

The important difference is that you are not asking Claude to “make it engaging.”

You are defining what engaging means for your particular audience.

Example 3: Sales and Business Development

Sales teams can use Claude for account research, outreach preparation, objection handling, and call preparation.

Suppose you want to prepare for a meeting with a potential customer.

Prompt

Act as a senior B2B SaaS sales strategist.
I am preparing for a discovery call with a company that has approximately 300 employees and is currently expanding its sales team across Europe.
Our product helps sales teams automate lead research and prospect enrichment.
Based only on the information I provide below, prepare:

1. Five hypotheses about their potential business challenges
2. Seven discovery questions
3. Three likely objections
4. A recommended response to each objection
5. Three signals that would indicate strong buying intent

Important: distinguish between facts provided in the context and assumptions. Do not present assumptions as facts.

Here is the company information:
[PASTE INFORMATION]

That last instruction is particularly useful.

AI can produce very convincing statements, even when the underlying information is incomplete.

For business research, asking Claude to distinguish known information from assumptions makes the output easier to review.

Example 4: Product Manager

Product managers often have to turn large amounts of feedback into something a team can actually prioritize.

Instead of:

Analyze this customer feedback.

try:

Act as a senior product manager for a B2B SaaS product.
Below is feedback collected from 45 customers.
Analyze the feedback and identify:

• recurring problems
• feature requests
• usability problems
• pricing concerns
• positive feedback

Group similar comments together rather than treating every comment as a separate issue.

For each major theme, provide:

1. Number of mentions
2. Representative examples
3. Potential business impact
4. Suggested priority: High, Medium or Low
5. Reason for the priority

Do not recommend building a feature simply because several customers requested it. Consider frequency, customer impact, and potential business value.

This is a much more useful instruction because it gives Claude a decision framework, not just a request for summarization.

Example 5: Project Manager

Project managers can use Claude as a planning and communication assistant.

For example, imagine a website project that is already behind schedule.

Prompt

Act as an experienced digital project manager.
We are rebuilding a company website. The original launch date is four weeks away, but development is currently approximately 10 days behind schedule.

The remaining work includes:

• homepage development
• five product pages
• CMS configuration
• analytics setup
• QA
• content migration

The design is approximately 90% complete.

Create a recovery plan for the project.

Identify:

1. Critical-path activities
2. Tasks that could run in parallel
3. Tasks that could potentially be reduced or postponed
4. Risks to the new launch date
5. Decisions that need to be made this week

Present the result as a practical action plan for the project team.

Do not assume that additional staff or budget are available unless explicitly stated.

The final sentence is important.

Constraints tell the model what not to assume.

Example 6: Data and Business Analyst

AI can be particularly useful for explaining analytical findings to non-technical stakeholders.

Imagine you have a dataset showing declining customer retention.

Instead of:

Analyze these numbers.

try:

Act as a business analyst preparing insights for a non-technical leadership team.
I will provide customer retention data for the last 12 months.
Analyze the dataset and identify:

• the most significant changes
• unusual patterns
• potential explanations
• questions that require further investigation

Separate observations from hypotheses.

Do not claim that one variable caused another unless the data supports a causal conclusion.

Finish with:

• 5 key findings
• 3 recommended next analyses
• 3 questions leadership should discuss

Use plain business language and avoid unnecessary statistical terminology.

This is a good example of where a prompt can improve not only the answer, but the quality of reasoning around the answer.

Example 7: Developer

Claude can also be used for programming, but developers often make the same mistake: providing too little context.

Weak prompt

Fix this Python code.

Better prompt

Act as a senior Python developer reviewing production code.
The function below processes customer transaction data and returns a monthly summary.
The current problem is that transactions with missing customer IDs cause the function to fail.

Requirements:

• do not change the function’s public interface
• preserve the existing output structure
• handle missing customer IDs safely
• explain the root cause
• provide the corrected version
• identify any other obvious edge cases

Here is the code:

[PASTE CODE]

This gives Claude a much narrower engineering problem.

And if you are working with a larger codebase, don’t expect a single giant prompt to solve everything.

For complex tasks, breaking the work into smaller stages can be more reliable. Anthropic describes this approach as prompt chaining: using several connected prompts rather than trying to solve a complex task in one instruction.

The Same Task Can Produce Completely Different Results

One of the easiest ways to understand prompting is to compare two versions of the same request.

Imagine a marketing manager wants to understand why a campaign performed poorly.

Prompt A

The model has almost no information.

Prompt B

Act as a senior B2B demand-generation strategist.
Analyze the campaign data below.
Campaign objective: generate qualified demo requests.
Target audience: HR leaders at companies with 200–1,000 employees.
Channels: LinkedIn Ads and email.
Campaign duration: 21 days.

Analyze:

• CTR
• conversion rate
• cost per lead
• lead quality
• differences between channels

Identify the three most likely performance problems and explain what additional data would be needed to confirm each hypothesis.

End with five practical experiments for the next campaign.

The second prompt creates a much better starting point because Claude understands the business problem, not just the task.

One of the Most Useful Prompting Techniques: Give Claude an Example

Sometimes explaining what you want isn’t enough.

An example can be much more precise.

For example:

Convert the following product descriptions into short LinkedIn posts.

Use this style:

Input:
“Our analytics platform helps sales teams understand which accounts are most likely to convert.”

Desired output:
“Your sales team probably doesn't need more leads. It needs to know which leads deserve attention first.

That’s where intent data can make a difference.”

Now create three posts using the same writing approach for the following products:
[PRODUCTS]

Anthropic specifically highlights examples as a useful way to improve how consistently Claude follows a desired format or behavior.

This technique is often called few-shot prompting.

You don’t necessarily need ten examples.

One or two strong examples can sometimes communicate the desired style much better than several paragraphs of explanation.

Don’t Ask Claude Only What to Do. Explain What “Good” Looks Like.

This is probably the biggest practical lesson from all of these examples.

Compare:

Create a job description.

with:

Create a job description that would appeal to experienced software engineers without using exaggerated language or generic phrases such as “rockstar developer.” Make the responsibilities specific and distinguish between required and preferred qualifications.

The second prompt gives Claude a quality standard.

This is especially useful when you already know what you don’t want from the output.

However, there is an important nuance: whenever possible, tell Claude what you want it to do rather than building a prompt entirely around prohibitions. Anthropic’s current guidance also recommends positive, explicit instructions for controlling output.

Instead of:

Don't make it generic.

try:

Use specific examples, concrete language, and details relevant to B2B SaaS marketing.

Use Claude as a Collaborator, Not Just a Text Generator

Another useful shift is to stop treating Claude as a machine that should produce the final answer immediately.

For complicated tasks, use several steps.

For example, if you need to create a marketing strategy, you could use this sequence:

Prompt 1 — Understand the problem

Analyze this business situation and identify the information that is missing before a marketing strategy can be created.

Prompt 2 — Build the strategy

Based on the information above, develop three possible strategic approaches.

Prompt 3 — Challenge the strategy

Act as a skeptical marketing director. Identify weaknesses, assumptions, and risks in each approach.

Prompt 4 — Improve it

Based on the critique, create the final recommended strategy.

This approach is often more useful than asking:

Create the perfect marketing strategy.

The reason is simple: complex professional work usually involves analysis, decisions and iteration, not one perfect instruction.

A Prompt Template You Can Reuse

Here is a practical template you can adapt to almost any professional role:

Role:
Act as a [PROFESSIONAL ROLE] with experience in [SPECIALIZATION].

Context:
[Explain the situation, company, audience, project, or problem.]

Objective:
I need to [DESCRIBE THE BUSINESS OR PROFESSIONAL GOAL].

Task:
[Tell Claude exactly what you want it to do.]

Criteria:
The result should [DEFINE WHAT GOOD LOOKS LIKE].

Constraints:
[Budget, length, audience, tools, resources, deadlines or other limitations.]

Output:
Present the result as [TABLE / BULLETS / PLAN / EMAIL / REPORT / STEP-BY-STEP GUIDE].

Additional information:
[Paste your data, document, examples or source material.]

Save this template somewhere you can easily access it.

You will probably find that you rarely need to start a professional Claude conversation from scratch again.

What to Do When Claude Gives You a Bad Answer

A poor response doesn’t necessarily mean that Claude cannot perform the task.

Before changing the model, try improving the instruction.

Ask yourself:

Did I provide enough context?

If Claude doesn’t know your audience, business, customer or project, it has to fill in the gaps.

Did I define the actual objective?

“Write a report” isn’t an objective.

“Help the leadership team decide whether to continue investing in this channel” is.

Did I explain what should be included?

If something is important, say it explicitly.

Did I define the output?

If you need a table, checklist, executive summary or action plan, tell Claude.

Did I give an example?

If you want a specific style or structure, an example can be more effective than a long explanation.

Am I asking Claude to solve too much at once?

Break complex work into smaller stages.

A Better Way to Think About Prompt Engineering

Prompt engineering is sometimes presented as a collection of secret phrases that unlock better AI responses.

In everyday professional work, it is much less mysterious.

A good prompt is essentially a well-written brief.

A manager gives an employee a vague assignment:

“Figure out our marketing.”

The result will probably be vague.

A manager provides:

“Here is our target market, current performance, budget, deadline, and business objective. Analyze three possible approaches and recommend one based on these criteria.”

Now the employee has something actionable.

Claude works similarly.

The goal isn’t to find a magical combination of words.

The goal is to provide enough context, direction, and criteria for the model to understand what a useful result actually means.

The Prompting Checklist

Before sending an important prompt to Claude, ask:

• Did I explain the context?
• Is the task specific?
• Did I define the desired outcome?
• Did I provide the information Claude needs?
• Did I explain important constraints?
• Did I define what a good answer should contain?
• Did I specify the output format?
• Would an example make my expectation clearer?
• Am I asking Claude to make assumptions that I could clarify?
• Is the task complex enough that I should break it into several steps?

If you can answer “yes” to most of these questions, your prompt is probably much stronger than a one-line request.

Final Takeaway

The biggest improvement you can make to your Claude results is not learning dozens of complicated prompting tricks.

It is learning to brief AI the way you would brief a capable professional colleague.

Explain the situation.

Give it the relevant information.

Define the goal.

Set the boundaries.

Describe what a useful result looks like.

Then review the output and continue the conversation.

For simple tasks, one well-structured prompt may be enough. For more complex professional work, Claude can become much more useful when you treat the interaction as an iterative process rather than a single question-and-answer exchange.

And that is ultimately what makes AI skills valuable.

Related Posts

Mastering the Coding Interview: A Deep Dive into Educative.io’s Flagship Course

Read More

Prompt Engineering & AI Literacy: The New Core Skill of the AI Era

Read More

AI Governance and Policy Compliance

Read More