Writing effective prompts in Adobe AI Assistant for Acrobat

Prompt engineering involves crafting prompts that guide AI to produce the desired output. Effective prompts enhance the ability of the AI tool to generate succinct and relevant responses of high quality.

Understand prompt engineering

Prompts guide AI tools by providing questions, commands, or statements to generate output with the desired structure, tone, and relevance. Proficient prompt writing is essential to maximize the use of AI Assistant for Acrobat.

Common types of prompts

Here's a list of common types of prompts with examples of how each can be effectively used to achieve specific outcomes with generative AI tools:

Structure of a prompt

A well-crafted prompt for AI Assistant for Acrobat integrates specific tasks, relevant context, clear expectations, and formatting guidelines to guide the AI to generate the desired outcome. 

Here are components of a prompt used to guide AI Assistant for Acrobat to generate precise output:

When crafting a prompt, think of it as directing a helpful assistant—the more specific, the more consistent the outputs. Embrace experimentation in prompt engineering by varying questions and adding context.

Writing effective prompts

To generate the most informative and useful responses, clearly define your goals, provide relevant context, use clear and concise language, employ examples, outline steps, and build on prior interactions. A prompt includes the following information:

• The topic.
• Keywords or phrases associated with the topic.
• The tone of the response.
• The target audience.

Below are key considerations and approaches for crafting effective prompts to quickly generate relevant insights.

Be intentional with your prompts

To craft effective prompts, first understand your task and then clearly specify the desired outcome:

Keep it simple

To enhance clarity and focus in your interactions, keep your prompts simple and break down complex questions into manageable parts:

Communicate constraints

Specify constraints clearly when communicating. If your request has limitations, like a word count or a focus on specific sections or themes, stating these upfront helps tailor the response accordingly.

Customize prompts

Tailor prompts to fit your needs. Specify the output's tone, style, and audience, as this context adds depth and improves response quality. Experiment with various lengths, tones, and voices as desired.

Define desired output format

Specify the desired output format. Clearly indicate if a particular format like paragraphs, bullet points, lists, tables, JSON, or XML adds value. Also, define the style and tone, ranging from formal to casual or persuasive, to align the output with your goals.

Format strategies:

• Clearly define the needed format, such as bullet points, paragraphs, JSON, or XML.
• Choose a format suited to the output's use, like data integration or readability.
• For structured formats, specify essential fields or elements.
• Experiment with different prompts and outputs to refine the response.

Improve results with few-shot learning

Few-shot learning enables AI models to predict accurately with few examples, focusing on generalization over memorization. Include examples to direct the AI Assistant for Acrobat's approach. This method is particularly useful for creating creative, analytical outputs or structuring detailed responses.

Guide with chain-of-thought prompting

Chain-of-thought prompting directs the AI Assistant for Acrobat through sequential reasoning, mimicking human logic to solve problems or answer questions. This method breaks tasks into logical steps, improving AI responses. Effective prompting tips:

• Outline steps - Present each step as a question or statement leading to the next part of the process. Use cues like "Let’s think step by step," "Consider each aspect one by one," or "Firstly, secondly, finally" to signal logical connections. 
• Logical flow - Ensure prompts lead the AI Assistant for Acrobat logically toward the solution.
• Transparency - Prompt the AI Assistant for Acrobat to show its reasoning, creating a clear path to the final decision.

Expand responses using prompt chaining

Prompt chaining involves an iterative process of prompts and responses, where each new prompt builds upon the AI Assistant for Acrobat's previous response. This method is ideal for creative tasks like narrative generation or brainstorming, allowing ideas to evolve gradually. Unlike chain-of-thought prompting, which offers detailed single-response reasoning, prompt chaining is dynamic and involves multiple interactions over time.

Check for accuracy

Occasionally, the AI Assistant for Acrobat may make mistakes. Always double-check its responses for accuracy, grammar, and style.

Iterate and improve

Use previous responses as a foundation for further interactions with AI Assistant for Acrobat, enhancing its effectiveness over time. Refine prompts to better guide the AI, adding context, clarifying objectives, or asking more specific questions if needed. Crafting effective prompts is a skill that improves with practice, so embrace failures as learning experiences and keep experimenting.

Avoid common mistakes while writing prompts

• Vagueness. When prompting, use clear language to receive better-quality responses.
• Requesting inappropriate or unethical content. You are responsible for the content and consequences of your prompts. Always respect local laws, rules, others' rights, and terms of service when using the AI Assistant for Acrobat.
• Slang, jargon, or informal language. Using informal language may lead to low-quality, inappropriate, or unprofessional responses from the AI Assistant for Acrobat.
• Conflicting instructions. Prompts containing conflicting information can confuse the AI Assistant for Acrobat, resulting in lower-quality responses.
• Overcomplicating prompts. Simplify your prompts to ensure clarity for the AI Assistant for Acrobat, focusing on tasks the AI can handle effectively.
• Assuming prior knowledge. Don't assume that the AI Assistant for Acrobat knows what is provided in the prompt, as this can lead to inaccuracies. Always provide context and set clear expectations.

Responsible AI practices

Here are key practices for responsible AI use:

For more information, see Adobe’s commitment to AI ethics.