Breaking the Code: A Lawyer’s Take on Prompt Engineering

Last month, I published an article on the importance of legal professionals knowing how and what to write as an input processed by a generative AI solution like ChatGPT, Perplexity or Claude. The article socialized the difference between searching for knowledge versus analyzing resources to convey actionable insights. My opinion is the former type of discovery is static, while the latter is dynamic, rewarding curiosity and the ability to ask successive questions, especially, ones that build on the previous. Doesn't this sound like a deposition strategy or the foundation of inductive or deductive reasoning, the necessary pillars for effective persuasion.

Consequently, lawyers I would argue are the knowledge workers who have the most to gain pivoting from the habits of Google search to the higher effort, higher reward possibilities afforded by prompt engineering for GenAI chatbots.

To illustrate this point, the article's remaining sections will introduce the tactics of prompt engineering and how the informed use of common prompt designs (a fancy phrase synonymous with "asking a good question") can both solve the current problem, and be a tool for future use and self-directed learning.

Why Knowledge Workers Will Prefer Prompting versus Searching

Imagine you're drafting a complex commercial contract and need to include a limitation of liability provision. You might start by searching on Google for something like "sample limitation of liability clauses." The search results will provide a variety of examples culled from different contracts across the web.

This is you acquiring knowledge. You are able to ask a question and get information.

Now consider inputting a more targeted query into an AI legal assistant like Claude: "Draft a limitation of liability clause for a software licensing agreement that caps direct damages at the contract price and excludes indirect and consequential damages." Claude can analyze this specific request, identify the key components (type of agreement, direct damages cap, exclusion of indirect/consequential damages), and generate a tailored draft clause that addresses these precise parameters.

This is you applying knowledge. You can ask many questions that convey an analysis and actionable insights, not just information.

By crafting a detailed, context-specific prompt, you can direct AI to generate work product that aligns closely with your unique needs - in this case, a bespoke limitation of liability provision for a software deal.

So, What is Prompt Engineering?

Prompt engineering, the art and science of designing effective inputs for AI models, offers a particularly promising avenue for lawyers to leverage this technology.

At its core, prompt engineering involves constructing inputs (prompts) that guide AI language models towards generating desired outputs.

It's about understanding how to frame questions and provide context in a way that elicits accurate, relevant, and insightful responses from AI.

Prompt Design Patterns for Legal Applications

Several prompt design patterns align especially well with the kinds of analytical tasks common in legal work:

That's it? It's not. As we delve deeper into prompt engineering, we find techniques that align very well with our legal expertise. Below are some basics:

1. Role Prompting: We can instruct the AI to adopt specific legal roles or perspectives. Imagine if you could anticipate a counterparty's strongest arguments and their leverage prior to a commercial transaction meeting. You can by directing the AI to assume a role with a given perspective, within a given context, and then generate likely outcomes and responses. Would a junior attorney find this form of coaching helpful?

2. Comparative Analysis Prompting: This technique asks the AI to compare and contrast multiple fact patterns with a variety of assumptions. A well informed hypothetical scenario, or two, or three, allows for:

a. Scenario planning and contingency development - inhouse counsel manage a spectrum of risks that are not created equal; so, knowing what mitigations are effective in one circumstance and not others enables the attorney to not only provide options, but a recommended course of action depending on what fact pattern requires which action.

b. Comparative Analysis - a litigator can create personas of different judicial dispositions to a legal philosophy and present their arguments to an AI and learn the strengths and weaknesses of their client's position, not just under the applicable law and precedent, but also under the judicial philosophy to anticipate what facts are material to this judge more often than not, and then developing persuasive positions.

c. Comprehensive Analysis - an inventor who seeks to understand what is patentable subject matter may ask a lawyer what is the best, most likely claim to be awarded, grounded in what is novel with the invention. But if the lawyer is not a subject matter expert, they could use an AI to learn about the industry and technology generally, and know what is potentially new and innovative, and what must be drafted in the application to not fall into a prior art morass.

The Ethical Imperative

As with any powerful tool, the use of AI and prompt engineering in law must be approached thoughtfully and with a strong ethical framework. Maintaining the standards of competence, confidentiality, and professional judgment that define legal practice should always remain at the forefront.

A Vision for the Future

The integration of prompt engineering and AI into legal work is a tremendous opportunity. By combining deep legal expertise with carefully crafted prompts, lawyers can enhance their efficiency and analytical capabilities in service of their clients and the pursuit of justice.

This is a frontier that the legal profession is uniquely well-prepared to navigate. The same commitment to rigor, ethics, and lifelong learning that has always characterized the practice of law will enable attorneys to harness these new tools in transformative ways.

As we explore the potential of prompt engineering together, we can look forward to a future where the power of artificial intelligence is intelligently leveraged to make the law more accessible, efficient, and equitable for all. The challenges are real, but so too are the possibilities - and the legal community has the skill and vision to make the most of this exciting new era.

This was the second of a multiple article series on GenAI and the legal profession. Be on the lookout for podcasts on these published articles. When they become available you can find the links in comments to the article the podcasters will discuss for that podcast, and you will see updated posts on my LinkedIn feed.