What is a Prompt?
Prompts are instructions or questions that prompt AI to generate specific responses. They act as the “language” through which humans communicate with AI, bridging the gap between human intent and machine execution. The quality of the prompt significantly influences the quality of the AI’s output. A well-constructed prompt can make the difference between a generic, unhelpful response and a valuable insight.
Why Learn Prompting?
AI is here to stay, offering new possibilities across various fields—from text and image creation to data analysis and more. Those who use AI in their professions report significant time savings and reduced workload. As AI technology becomes more integrated into job roles, mastering AI tools will likely become a valuable skill.
You don’t need to become a Prompting Engineer to benefit from AI. Learning the key rules for effective prompting and practicing regularly will improve your proficiency.
6 Key Aspects of a Good Prompt
1. Clarity
The clarity of a prompt is crucial. The clearer the request, the better the AI understands what is required. Avoid vague statements and be as specific as possible.
2. Context
While AI has extensive knowledge, it needs context to provide relevant answers. Including context in your prompt ensures more accurate and specific responses.
3. Goal Orientation
Define the goal of your prompt in advance. Whether it’s answering a question, generating ideas, providing a detailed description, or a brief summary, include this in your prompt.
4. Step-by-Step Refinement
If the initial response is unsatisfactory, refine the prompt to focus on specific aspects. Building on previous prompts helps guide the AI to the desired outcome, aiding in learning how to craft better prompts initially.
5. Keywords
Use keywords to help the AI focus on the relevant topics, especially for broad subjects. Integrate central keywords naturally into the prompt to maintain readability.
6. Creativity
Effective AI use requires creativity. Different phrasings yield different results, so experiment to find what works best. Encourage the AI to take on specific roles or explore its boundaries.
7. Providing Examples
Show the AI what you want by providing examples. For instance, input a complete article or paragraph to help the AI mimic the tone and style or write an article with a similar structure.
Structured Prompts
The more structured your prompt is, the clearer the AI can understand what you want from it. Include aspects like these, depending on what is important to you:
- Perspective: Specify the perspective from which the AI should respond (e.g., “You are a 60-year-old man interested in our marine conservation project who enjoys sailing.”)
- Context: Explain the context in which the AI should understand your request (e.g., I am a fundraiser for an environmental organization and want to understand what is important in a donation appeal to persuade someone like you to donate.)
- Task: Use verbs to describe what the AI should do (e.g., research, write, summarize, analyze, improve, list …)
- Length: Give the AI specific instructions on how long its response should be, which is particularly relevant for website and blog texts (e.g., Write a donation appeal in 500 characters.)
- Target Audience: Specify the target audience for the AI (e.g., Gen X, women living in rural areas, highly educated managers, young fathers …)
- Format: The AI can deliver its responses in different formats (e.g., table, list, comparison, table of contents, article, bullet points …)
- Tone: The AI can also adjust its tone according to your wishes (e.g., factual, informative, humorous, casual, sarcastic, …)
Examples: Good Prompt vs. Bad Prompt
To illustrate the above aspects, here are some examples:
Example 1: “Write a donation appeal”
The problem here is that the prompt is too general. It doesn’t provide the AI with information about the organization, projects, or target audience. For a targeted and engaging result, phrase it better like this:
“Write a donation appeal to support educational programs for children in rural areas of India, aimed at potential new donors in the education sector.”
Example 2: “Analyze the effectiveness of our donation campaigns.”
Without specific data or criteria for “effectiveness,” the AI doesn’t know what exactly to analyze. It’s better to say:
“Compare the total donations generated by our email campaign versus our social media campaign in the last quarter and identify which campaign had the highest ROI.” – of course, the relevant data would need to be included (but anonymized for privacy!).
Example 3: “Create posts for our social media pages.”
Again, this prompt is too generic. The AI needs more information to produce useful responses. A better prompt would be:
“Create inspiring and informative social media posts about environmental protection that appeal to young adults while maintaining our optimistic and action-oriented brand tone.”
Prompts for Image and Video AI
Recently, OpenAI gave a first look at their new video AI, Sora. In the linked video below, you can see not only the results but also the preceding prompts. These are surprisingly detailed and read almost like literary prose. The same goes for generating images with AI: it requires a precise idea (lots of imagination) and the ability to express it accurately in words.
