Peer to Peer: ILTA's Quarterly Magazine
Issue link: https://epubs.iltanet.org/i/1538025
P E E R T O P E E R M A G A Z I N E ยท S U M M E R 2 0 2 5 77 However, lawyers, paralegals, and other legal professionals who want to take their AI capabilities to the next level can unlock even more productivity by mastering the art of creating their own prompts to automate and further enhance tasks. Prompt engineering is a valuable skill when pre-built AI prompts are not quite delivering the results you are looking for. Creating prompts enables you to fine-tune, iterate, and optimize, resulting in outputs tailored to your specific needs. Effective prompt engineering requires providing instructions that guide GenAI models to deliver the best results. Generally, it is not a one-and-done task, such as entering a few words into a search engine. Prompt engineering involves trial and error, adjusting methods to unlock the desired outputs. While this process takes time to learn, users who master prompt engineering can gain more control of their AI systems and expand the value it delivers beyond built-in options to nearly endless use cases. Whether your organization is using AI in a legal case management solution or an ediscovery tool, or a standalone AI solution such as Harvey, Google Gemini, or Microsoft Copilot, this article will offer a look under the hood of AI, how it works, how it knows things (and what it does not know), and practical tips on improving your prompt writing. WHAT HAPPENS WHEN I PROMPT AI? Before building prompts, it is essential to understand the fundamentals of how generative AI operates. GenAI leverages pretrained models to analyze and generate content. For text-based solutions, these are large language models (LLMs) that are trained on extensive datasets, including books, websites, and other sources, to capture the intricacies of language. The model tokenizes the prompt by breaking it into smaller, processable pieces. The AI then predicts the most likely sequence of tokens based on what it has learned during training. Essentially, it is pattern prediction driven by complex algorithms and vast amounts of data. GenAI systems are increasingly adaptive, with fine-tuning options that allow customization for specific industries or tasks. GenAI systems also generally fall into two categories: white- box and black-box. Black-box AI delivers results but keeps its decision-making process hidden. Prompt engineering is a valuable skill when pre-built AI prompts are not quite delivering the results you are looking for.