Mastering the art of prompt engineering (aka prompt crafting) is touted as the new “must learn” business skill. It is. But just because it has a fancy, technical sounding name does not mean it is complicated or hard to understand. Here’s a very short primer, guaranteed to improve the output you get from ChatGPT, Copilot, Gemini, or most any generative AI platform. You are only minutes away from unlocking your prompt engineering superpowers. Let’s go!
Pre-Prompts Make Prompts Better
Incorporating a pre-prompt that sets a specific context or role can enhance the efficacy and relevance of the responses. It helps align the AI’s responses with the intended audience’s needs and expectations, and the specific scenarios being addressed.
General Pre-prompt Structure
Context Setting: “You are acting as a seasoned marketing strategist with deep expertise in AI applications in marketing.”
Audience Consideration: “Your audience consists of marketing professionals ranging from novices to experts, seeking to integrate AI into their strategies effectively.”
Best Practices in Prompt Engineering
It’s helpful to categorize prompts based on their intended outcomes or the style of interaction they facilitate. Here’s a list of popular types of prompts with short descriptions for each:
Instructive Prompts: These prompts directly instruct the model to perform a specific task, follow a particular format, or adhere to constraints. They are clear and precise about what is expected. Example: “Draft a compelling email for a digital marketing conference, inviting a leading expert in AI for marketing to deliver the keynote address, emphasizing the unique audience they’ll reach and the topics they could cover.”
Creative Prompts: Encourage AI to generate content that is imaginative or unconventional. They often seek stories, poems, or creative ideas. Example: “Generate a narrative about a marketing professional using an innovative AI tool to solve a challenging campaign issue, leading to unexpected success and insights into consumer behavior.”
Informational Prompts: Request factual information, explanations, or summaries. They are used to gather knowledge or clarify concepts. Example: “Provide a detailed explanation of how AI and machine learning technologies are transforming predictive analytics in marketing, including real-world applications and benefits.”
Conversational Prompts: Initiate a dialogue or a series of questions and answers. They mimic a natural conversation and can be open-ended. Example: “Discuss the evolving role of AI in content marketing strategies and how it’s shaping the future of personalized customer engagement.”
Persuasive Prompts: Ask AI to argue for or against a particular viewpoint, persuade the reader of a position, or construct a compelling case. Example: “Make a compelling argument for the integration of chatbots in customer service strategies, focusing on the improvement of customer experience and operational efficiency.”
Reflective Prompts: Encourage introspection or analysis of ideas, experiences, or texts. Though based on programmed data, AI can generate analytical or reflective responses on a given topic. Example: “Analyze the ethical considerations marketers must take into account when implementing AI-driven personalization tactics in their campaigns.”
Scenario-Based Prompts: Present a hypothetical situation and ask for a response or solution to the scenario. They are useful for problem-solving or exploratory thinking. Example: “Outline a strategy for a startup to leverage AI tools for market analysis and customer segmentation in launching a new product in a competitive market.”
Instructional Prompts: Designed to explain how to do something in a step-by-step manner. They are often used for tutorials, recipes, or DIY projects. Example: “Explain the step-by-step process for setting up and optimizing an AI-powered programmatic advertising campaign to maximize ROI.”
Analytical Prompts: Require analysis or critical thinking to break down a concept, argument, or data. They often involve comparison, contrast, or evaluation. Example: “Evaluate the impact of AI on email marketing performance, comparing traditional segmentation methods with AI-driven personalization techniques.”
Role-Play Prompts: Have AI assume a specific character or role in a given scenario, allowing for creative and diverse perspectives in responses. Example: “Assume you’re a marketing consultant specializing in AI. Advise a client on integrating AI into their digital marketing strategy to enhance customer engagement and increase conversions.”
Post-Prompts Help Too
There is a field of study known as “Theory of Mind.” It involves the AI anticipating the knowledge, beliefs, and intentions of the human interlocutor, thereby tailoring responses more closely to the user’s perspective and expectations. “Chain of Thought,” a related idea, instructs the AI to break down its reasoning process into a sequence of logical steps, making complex problem-solving or decision-making processes more transparent and understandable. Depending on how the AI is “feeling,” (I’m anthropomorphizing for fun. AI doesn’t feel — at least not the way humans do.) it may lay out its strategy for dealing with your request, then give you an option to approve it or suggest alternative approaches. Or, it may just do a better job of “reasoning” out the answer. (Reason is an apparent emerging attribute of generative AI that no one has yet offered a definitive explanation for.)
Including post-prompt language like “Let’s work this out step by step” provokes Chain of Thought-style responses and can be particularly effective for certain types of prompts, especially instructional, scenario-based, and analytical categories. This language encourages a structured, methodical response, which is ideal for explaining processes, solving problems, or analyzing complex issues in a clear and accessible manner. It signals the need for a detailed, sequential breakdown, which can be very helpful for readers seeking clarity and actionable guidance.
For other categories, especially those that benefit from creativity, persuasion, or reflection, a different type of post-prompt language might enhance the response quality. Here’s how you could tailor post-prompt language across different categories for improved performance:
For Instructional, Scenario-Based, and Analytical Prompts: “Let’s work this out step by step.” Signals the need for a detailed, sequential explanation or analysis.
For Creative and Role-Play Prompts: “Feel free to explore creative avenues.” Encourages imaginative and unconventional responses, allowing for a broader exploration of ideas.
For Informational Prompts: “Provide a comprehensive overview.” Indicates the need for a thorough and detailed explanation, covering all relevant aspects of the topic.
For Conversational and Reflective Prompts: “Let’s dive deep into this topic.” Suggests a more exploratory or thoughtful approach, encouraging a detailed discussion or reflection.
For Persuasive Prompts: “Build a strong case with compelling arguments.” Directs the response to be argumentative and convincing, emphasizing the construction of a persuasive narrative.
The Perfect Prompt
Each AI model has qualities and characteristics that make it unique. Importantly, these models and interfaces are subject to change without notice. Because models and interfaces are dynamic, there is no “perfect” prompt. That said, the prompt structures outlined above will help you unlock your AI super powers.
Author’s note: Originally posted on Feb 10, 2024, revised on April 7, 2024. This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.