AI Tips: How to write good prompts in the LLM era?

In the LLM era, you can use natural language to let AI help you work, and prompts are a new skill for communicating with machines. Although talking to AI is very similar to communicating with people, we still need to transform into AI's way of thinking to better implement prompting.

Clarify your goals: the cornerstone of writing good prompts

The first and most critical step in writing prompts is to clarify your desired results. This may seem simple, but it actually requires us to think deeply.

From a human perspective, a good prompt is a statement that can clearly convey its own goals. For example, when we want to write a product promotion copy, we cannot simply tell AI to "write a promotion copy". With such vague instructions, it is difficult for AI to grasp the focus and style of the copy. We need to think about the unique selling points of the promoted product, who the target audience is, and what kind of promotion effect we hope to achieve, and then incorporate this information into the prompt.

From the perspective of AI, although they can generate content based on massive amounts of data, they lack human subjective cognition and situational understanding. Therefore, only when we provide sufficient background information, clarify the expected results, and even give specific guidance, can AI generate content that is more in line with our wishes.

For example, when asking AI to write promotional copy for a smart watch, tell it that this watch has a health monitoring function and the target audience is young people who pursue a healthy lifestyle. You want the copy to highlight the sense of technology and health management advantages, so that AI can create in a targeted manner.

Rich information: Adding building blocks to AI understanding

Importance of Project Background and Role Information

When communicating with AI, providing project background and professional role information can help AI better understand the task. For example, a marketing department operator reports project progress to the company's senior management. "I am responsible for the operation of the marketing department. This is the first time I have to report project progress to the company's senior management." This kind of information supplementation makes AI understand that this is a reporting scenario between superiors and subordinates within the company, involving specific project requirements. Based on this information, it can choose a more professional and more appropriate way of expression for the reporting scenario, avoiding inappropriate words or structures.

If a teacher asks AI to help design teaching courseware, the teacher can tell AI background information such as the subject he teaches, the age level and learning level of the students. Based on this, AI can design courseware content that is more suitable for students of that age group to understand, and select appropriate cases and teaching methods.

Multi-dimensional information assists precise creation

In addition to project background and professional role, other relevant information can also improve the quality of AI creation. For example, when creating a travel guide, tell AI the travel season, budget, travel days, and personal interests and hobbies.

If you like history and culture, AI will recommend historical sites and cultural attractions in the guide; if your budget is limited, AI will recommend cost-effective accommodation and restaurants. These multi-dimensional information helps AI accurately identify needs and create more practical content.

Standardized output: Let AI create with purpose

Clarify content type and structure

Specifying the content type and structure in Prompt can make AI output more in line with expectations. For example, when preparing a project report for the marketing department, it is clearly required that "the report structure is: 1. Market data overview (simple chart or summary); 2. Project results overview (2-3 key points); 3. Next quarter plan", so that AI can organize the content according to the specified structure, highlight key data and highlights, and avoid content confusion.

When writing academic papers, requiring AI to follow the structure of "Abstract - Introduction - Research Methods - Results and Discussion - Conclusion" can make the paper more logical and standardized. At the same time, specifying the content type, such as whether to write a research report or a review article, allows AI to choose the appropriate writing style and tone.

Control output length

It is also important to properly control the length of the output. Different scenarios have different requirements for the length of the content. When reporting work, each part of the content may need to be concise and clear, and controlled within a certain number of words; when writing a novel, the requirements for the number of words are relatively flexible.

For example, "Please help me design a concise report structure, highlighting key data and highlights, and each part should not exceed 100 words." Through such requirements, AI can ensure the conciseness of the content and avoid being lengthy and complicated while satisfying the information transmission.

Example guidance: Helping AI understand quickly

Inspiration from case examples

Providing cases is an effective way to help AI quickly understand needs. Take the "Market Data Overview" section in the report content as an example. Give an example like "In the past three months, our market share has increased by 12%, and monthly user growth has stabilized at around 8%. Sales have increased significantly year-on-year, mainly driven by new product launches and marketing promotions." AI can learn from this how to summarize market performance and highlight key indicators and trends. When creating other types of reports, similar cases can also be provided to help AI master the corresponding writing style and focus.

When asking AI to create poetry, provide some examples of excellent poetry. AI can draw inspiration from them, learn techniques such as poetry rhythm and imagery, and create more poetic works.

The guiding value of frameworks and mind maps

In addition to cases, task frameworks and mind maps can also help AI better understand content structure and key points. For example, when planning an event, the framework of the event planning is given: "1. Event theme; 2. Event purpose; 3. Event time and location; 4. Event process; 5. Promotion plan; 6. Budget arrangement", and AI can fill in specific content based on this framework to ensure the integrity and logic of the event planning.

Mind maps can present the hierarchy and relationship of content in a more intuitive way. When creating a complex article, using a mind map to show the outline of the article and the relationship between the various parts, AI can more clearly grasp the overall structure of the article and create coherent content.

Chain of Thought (CoT): Improve the logic of AI answers

Chain of Thought (CoT) is an effective way to improve the logic of AI answers. When asking questions, by specifically asking the AI ​​to explain the reasoning process in detail, it is prompted to generate answers using a step-by-step reasoning approach.

For example, when analyzing sales strategies for the North American and Asian markets, a normal prompt might simply be “Analyze sales strategies for the North American and Asian markets,” while the prompt with CoT added would be “Please analyze the differences between the North American and Asian markets. Please explain step by step the market performance, driving factors, and differences between the two regions, and explain how these differences will affect future sales strategies.”

In this way, AI no longer simply gives conclusions, but makes the answers clearer and more logical through step-by-step reasoning. In scenarios such as solving mathematical problems and analyzing complex events, CoT can play an important role in helping us obtain more in-depth and reliable answers.

Writing prompts well is the key to communicating with AI efficiently. By clarifying goals, enriching information, standardizing output, guiding with examples, and using thinking chains and other techniques, we can make AI better understand our needs and generate better quality content that meets expectations.

As AI technology continues to develop, mastering these skills will make us more adept in the era of artificial intelligence and make full use of the advantages of AI to improve work and life efficiency.