Prompts are key in 2023: Twenty-five tips to help you unlock the potential of generative AI

In the last few years, generative AI has seen exponential growth. Language models like GPT-3.5-Turbo and GPT-4 on Azure OpenAI Service can automate content generation and conversational experiences, making it easier and more efficient to communicate with customers and end users.

AI input prompts defined

A prompt, in the context of AI, particularly in large language models, refers to the input or instruction given by users to elicit a specific type of response. To get the most out of large language models like GPT-4, it’s imperative to craft prompts that yield effective results. The challenge lies in choosing the best combination of words, expressions, symbols, and structures to steer the model toward producing accurate and pertinent content.

Why prompts matter

Just like when communicating in real life, how you ask for what you want can limit—or expand—the type of information you receive.

Prompts help specify the user’s intent and expectation from AI, hence more precise prompts lead to more accurate and relevant results. They allow users to obtain a wide variety of responses, from answering questions, creating stories in the tone of your favorite author, generating poetry, and even performing code-related tasks.

Similar prompts can lead to varying responses based on the underlying model, its training data, or even subtle variations in how you phrase your request.

Following you’ll find prompt tips to help you create the kind of content you need, whether at work or play.

Get started! Twenty-five prompt tips for your best content creation to date

  1. Know what you want
    Clearly outline the problem or need you’re trying to address. For instance, “I need fresh ideas for a marketing campaign geared toward our latest app.”
  2. Start with a simple question
    Begin with a simple question to check the model’s understanding, then build into more complex queries.
  3. Ask open-ended questions
    Generative models work best when they’re free to roam. Instead of asking, “Should I use social media for marketing?” ask, “What are some innovative ways to grab attention across my social media platforms.” You can even specify the specific platform you plan to use (X, LinkedIn, etc.).
  4. Iterate and refine
    Use the feedback you get from initial prompts then build on it. Like an idea about a general content marketing strategy? Follow up with, “How can I implement a content marketing strategy for a tech product designed to monitor how happy my pet is while I’m away from home?”
  5. Provide context
    The more context you provide, the better AI can tailor its response to your unique situation. For example, “I just released an app to track sleep patterns and am looking for low-cost marketing strategies that appeal to businesses concerned with their employee’s well-being.”
  6. State your boundaries
    Mention any constraints (e.g., budget, timeline, resources) upfront. “What are marketing strategies for a new product that can be executed within a 5,000 USD budget and within a two-week period?”
  7. Go big (then small)
    Break down big questions into smaller ones for more actionable insights. Instead of asking “How can I improve my business?” ask “How can I increase online sales?” Then, followed up with “What social media platforms are best for advertising beanbags?”
  8. Opposites attract: Marry creative and analytical requests
    AIs can handle both creative brainstorming and analytical tasks. But divide and conquer for the best results. Brainstorm marketing strategies, then follow up with “What are the pros and cons of influencer marketing?”
  9. Strengths, Weaknesses, Opportunities, Threats (SWOT)
    A SWOT analysis can help get your business on the right path. Ask AI to perform a SWOT analysis using specifics about your business for immediately actionable items you can get started on ASAP.
  10. Ask for examples
    Grasping a concept is easier with an example on hand. Ask for an example then use it as a potential model. You’ll discover what does—and doesn’t—work for your specific case. For example, “Can you provide a case study of a successful influencer marketing campaign?”
  11. Specify the format
    Want answers in bullet points, a paragraph, or a list? Make mention of that and see your wishes take literal form.
  12. Define tech terms
    If there’s a term or acronym specific to your industry, use a prompt to define it, or ensure the model understands the context in which it’s being used.
  13. Rephrase for precision
    If the initial answer isn’t satisfactory, rephrase your question. Not only will you get your creative communicative juices, but you’ll also find that answers to your questions can prompt new directions in your thinking.
  14. In-depth versus short-form
    Want a summary in a single paragraph or a pages-long academic deep dive? Make mention of your preferences in your prompt.
  15. Use negative instructions for a positive effect
    If you know what you don’t want, specify that. E.g., “Provide marketing strategies excluding online advertisements.”
  16. Multiple answers
    Ask for multiple answers or perspectives to a single question for a more comprehensive and nuanced understanding of your topic.
  17. Request sources
    While most models don’t browse the web in real-time, asking it to base its answer on known sources up to its last training data can provide greater credibility.
  18. Limit bias
    Explicitly ask the model to give an unbiased answer or to consider multiple perspectives.
  19. Context is key
    Ask for compliance requirements for specific industries. “What are the current trends in generative AI?” “What compliance requirements should I keep in mind for healthcare related topics?
  20. Power in numbers
    Quantify whenever possible: Need numbers or percentages? Metrics, distances, or speed? Include this request in your prompt.
  21. Avoid leading questions
    Ensure your question doesn’t steer the model towards a specific answer—unless that’s your intention.
  22. Tone it down (or up)
    If you need fun, out-of-the-box content versus a more academic tone, mention that. E.g., “Provide a fun creative tagline for a green energy campaign.”
  23. Provide real-world implications
    Explain why you need the answer or how it’ll be used for more context. We’re launching a new cookie testing app next month. How should we position it against competitor x?”
  24. Safety and accuracy
    Cross-reference critical information provided by the model with trusted external sources, especially if decisions based on the model’s answer have significant business implications.
  25. Refine over time
    If you’re using the model regularly, note down what types of prompts give you the best results and refine them accordingly.

Think of generative AI prompts as a multifunctional tool built for the digital age, adept at both enhancing business strategies and enriching our home lives. Visit the AI learning and community hub for more tips.

The goal is to use generative AI as a tool in your broader decision-making and brainstorming process. Combining AI’s suggestions with your expertise and knowledge of your business and market will yield the best results. By integrating these AI insights, we’re not merely keeping up with the times; we’re pioneering a future where efficiency meets innovation.

Our commitment to responsible AI

With Responsible AI tools in Azure, Microsoft is empowering organizations to build the next generation of AI apps safely and responsibly. Microsoft has announced the general availability of Azure AI Content Safety, a state-of-the art AI system that helps organizations keep AI-generated content safe and create better online experiences for everyone. Customers—from startup to enterprise – are applying the capabilities of Azure AI Content Safety to social media, education and employee engagement scenarios to help construct AI systems that operationalize fairness, privacy, security, and other responsible AI principles.

Get started with Azure OpenAI Service