Using Generative AI tools in the practice of law is becoming an essential skill.  It not only helps lawyers streamline their workflows and improve efficiency, but it is becoming crucial to meet the evolving demands of clients.

In order to get the most out of a GenAI tool when used in your practice, you need to prompt the tool in a way that takes full advantage of its capabilities.  Therefore, it is important to become proficient in “engineering” your prompts.  

What exactly is “prompt engineering?”  According to Perplexity (one of our current go-to GenAI tools), prompt engineering is “the practice of designing and refining prompts – questions, instructions, or contextual information – to effectively communicate with and elicit desired responses from artificial intelligence (AI) models ...”  More simply, it refers to how users interact with GenAI tools.

All lawyers seek to ensure that every word in their legal documents and communications has meaning and carries weight.  So, it should be no surprise that this holds true for effective AI prompt engineering. Here are some tips to help lawyers master the art of AI prompt engineering:

1. Be Precise: Accuracy and precision are critical when it comes to the law. Avoid ambiguity in your prompts by using clear, concise and plain language. Aim for a conversational tone, avoid legalese and explain to the tool the meaning of any acronyms and other abbreviations used in the prompt.
2. Context is Key: GenAI tools are trained on massive amounts of data, so in order to generate a useful output, you need to give context to your issue or problem when crafting your prompts. This could include details like the field of law, the regulatory regime involved, governing law, target audience and any specific documents you want analyzed.  The context windows for some of the most commonly used GenAI tools are more than large enough to allow you to give as much context to your prompt as needed (for example, OpenAI’s GPT-4 Turbo has a context window of roughly 100,000 words or 300 pages of text).  However, beware that providing too much context may not be helpful – finding the right balance is key. 
3. Specify the Why and the What: Clearly state your goal. What information are you seeking? What action do you want the AI tool to take? Once the intent is established, provide specific instructions on the desired output. For example, specify the “voice” you are looking for (e.g., formal, casual, personal, adversarial), and indicate the format you want for the output (e.g., a summary, an analysis, a draft, a bullet-point list, a table, a deck slide, an email correspondence, a blog-post, etc.).
4. Manage Biases and Fairness:  Consider how to adjust your prompt in order to minimize biases and to allow the AI responses to be as fair and objective as possible, especially in sensitive areas.  For example, if you are seeking GenAI assistance in drafting a public-facing recruitment policy, consider asking the tool to “generate recruitment content that emphasizes the importance of diverse skills and experiences in contributing to our team's success” rather than “generate recruitment content that highlights the need for young, ambitious and assertive new talent to join our team.”  You can even explicitly instruct the GenAI model to avoid certain pitfalls like bias, factual errors and hallucinations.
5. Iterate and Refine:  Leverage the chat-based nature of GenAI tools. You can have a back-and-forth conversation with the tool, providing additional context and tweaking prompts on an iterative basis. Try starting with a basic question, then keep refining the prompt (such as by adding more context and specificity, identifying problems with the output from an earlier prompt, and/or changing the tone or terminology) as needed to steer the AI tool towards a more relevant and focused output. 
6. Practice Makes … Better: Just as with the practice of law, crafting better GenAI prompts takes practice. Even the most experienced “prompters” adjust their approach not only to align with advances in AI technology, but to improve their prompt engineering as they learn what works and what doesn’t. 

Remember that while AI can be a powerful tool, it should only be used to supplement, rather than replace, your legal expertise. Use that legal expertise to engineer effective prompts and always critically review and analyze the AI tool's output.