Better Prompts, Better Results: How Legal Professionals Can Get More from AI

Artificial intelligence (AI) is quickly becoming a valuable and increasingly common tool in legal practice. From drafting correspondence and summarizing case law to reviewing contracts and organizing research, AI can help attorneys work more efficiently and spend more time on strategic legal work.

However, the quality of AI-generated results depends heavily on one thing: the prompt.

A prompt is the instruction given to an AI tool describing what you want it to generate. The clearer, more detailed, and more structured the prompt, the more useful and accurate the response is likely to be.

For legal professionals, learning how to write effective prompts for different scenarios is becoming an important skill. This includes understanding how to structure prompts effectively, refine results, avoid common mistakes, and use AI responsibly within the ethical obligations of legal practice.

What is prompt engineering?

Prompt engineering is the process of thoughtfully writing and refining instructions that guide AI tools toward producing accurate, relevant, and useful outputs. Unlike attorneys, AI does not truly reason or apply legal judgment—it identifies patterns and generates responses based on the information provided to it. As a result, the quality of the output depends largely on the quality of the input.

Because legal work requires precision, context awareness, and structured analysis, poorly written prompts can contribute to missed legal issues, incorrect jurisdictional assumptions, or broad and unreliable responses. In contrast, well-structured prompts help improve efficiency, consistency, and overall reliability of AI-generated output.

The difference between a vague prompt and a well-structured one can significantly impact the usefulness of the result. Here are some examples:

Vague prompt: “Explain contract law.”

This prompt is too broad and gives the AI very little direction. The result will likely be generic, overly simplified, or not useful for practical legal work.

More effective prompt: “Provide a concise summary of enforceability requirements for contracts under New York law, including offer, acceptance, consideration, and common defenses. Format the response as bullet points suitable for client education.”

This prompt is more useful because it specifies key concepts including the jurisdiction, desired format, and intended audience.

Highly detailed prompt: “Act as a senior commercial litigation attorney practicing in New York. Prepare a client-facing memorandum analyzing whether an unsigned services agreement between two businesses may still be enforceable under New York law where: (1) the parties exchanged multiple draft agreements by email, (2) services were partially performed for three months, and (3) invoices were paid during that period. Discuss offer, acceptance, consideration, partial performance, and any applicable statute of frauds concerns. Include relevant New York legal principles, organize the response using IRAC headings, and conclude with a practical risk assessment for the client. Limit the response to approximately 750 words and maintain a professional but plain-English tone suitable for a business owner.”

This version provides the AI with substantially more direction, including the role it should assume, the relevant factual and legal context, and limitations such as length and scope. While not every prompt requires this level of detail, understanding how to build more sophisticated prompts can be extremely valuable when handling complex legal analysis, drafting, or communications.

Core principles of effective prompt engineering

An effective prompt avoids ambiguity and provides the AI with clear and detailed facts. AI performs best when given relevant facts. When writing your prompt, consider including some, or all, of these core areas of specificity:

Purpose

Define the task you want the AI to complete (e.g., an email, demand letter, contract, analysis, or memorandum of law).

Context

Define the situation (e.g., contentious litigation, firm negotiation, gentle guidance, or transmittal).

Persona

Define what role the AI should assume (e.g., a teacher, kind advocate, dangerous opponent, professional leader, or senior employment attorney).

Tone and Style

Define the tone and style the response should use (e.g., terse demand, persuasive logical motion, summary for the client at a sixth grade reading level).

Format and Output

Direct how the AI should organize the results (e.g., as a bulleted list, chart, memorandum, IRAC, or diagram).

Constraints

Clarify limits to reduce irrelevant output (e.g., define the specific jurisdiction, documents to analyze, word count, time frame).

Use Examples

Provide examples or reference documents to tell the AI (e.g., write me a new client email to match the same tone, format, and word count as this past correspondence).

Let’s break down parts of an example prompt and then put it all together.

• Purpose: Analyze whether the facts support a claim for negligence.

• Context: We represent the pedestrian who was struck in a crosswalk by a driver who was texting. The incident occurred in Illinois.

• Persona: Act as a senior litigation attorney.

• Tone and Style: Use a professional and objective tone suitable for internal case evaluation, while keeping the analysis concise.

• Format and Output: Provide analysis using IRAC and include key elements of negligence.

• Constraints: Limit to 400 words and focus on Illinois law.

• Use Examples: Model the analysis after a traditional pre-litigation case assessment memorandum and include practical observations about strengths, weaknesses, and potential defenses.

Now, here’s the full prompt:

“I need to analyze whether the facts support a claim for negligence. We represent the pedestrian who was struck in a crosswalk by a driver who was texting. The incident occurred in Illinois. Act as a senior litigation attorney. Use a professional and objective tone suitable for internal case evaluation, while keeping the analysis concise. Provide analysis using IRAC and include key elements of negligence. Limit to 400 words and focus on Illinois law. Model the analysis after a traditional pre-litigation case assessment memorandum and include practical observations about strengths, weaknesses, and potential defenses.”

AI prompt use cases examples

To help you begin incorporating AI prompt templates into your day-to-day legal workflows, consider the following practical examples and use cases:

• Summarize recent Supreme Court decisions (2020–present) addressing administrative deference doctrines.

• Draft a professional cease-and-desist letter regarding trademark infringement, including formal legal tone and standard demand language.

• Review the following indemnification clause and identify risks to the indemnified party under [state] law.

• Identify and categorize potential legal issues in the following fact pattern, including tort, contract, and regulatory concerns.

• Summarize key elements of fraud under [state] law in bullet-point format.

• Using the provided example, draft a mutual NDA with standard confidentiality provisions and a 2-year term.

• Evaluate potential liability exposure for breach of fiduciary duty under [state] law.

Effectively review and refine AI outputs

AI may not produce the perfect result on the first attempt. Effective prompting is an iterative and collaborative process between the user and the AI. Refining prompts is often helpful to improve the quality, focus, and usefulness of the output.

After reviewing the initial response, continue guiding the AI by clarifying instructions, requesting additional detail, adjusting tone, narrowing scope, or adding constraints. Small refinements can significantly improve the final result.

Common AI prompting mistakes

• If you receive a generic result, your instructions may be too vague. Provide precise instructions and refine the result.

• If the incorrect law was applied, add the jurisdiction and area of law.

• If you receive an unfocused, meandering response, the task may be confusing or too complex. Break it into steps so the AI can complete the task sequentially.

• If you receive an output that is hard to use or convoluted, provide structure to the AI: paragraphs, chart, bulled points.

Always maintain professional responsibility

Most importantly, attorneys remain fully responsible for all legal work product generated with the assistance of AI. AI output should always be carefully reviewed, verified, and edited before use. Attorneys should independently verify all legal authorities, citations, and factual assertions generated by AI before relying on them in legal work.

A helpful way to approach AI-generated content is to treat it like a draft prepared by a junior associate—it may provide a useful starting point, but it still requires attorney oversight, legal judgment, and due diligence.

Lastly, attorneys should also avoid entering confidential or sensitive client information into open AI platforms. Before using any AI tool, it is important to understand the platform’s privacy policies, data handling practices, and applicable ethical obligations related to client confidentiality.

Start building better legal AI workflows

As AI continues to become more integrated into legal practice, the ability to write effective prompts is quickly becoming an important professional skill. Well-structured prompts help legal professionals generate more accurate, relevant, and useful outputs while improving efficiency across research, drafting, analysis, and client communication.

At the same time, AI should always be approached thoughtfully and responsibly. When used strategically and with proper oversight, AI can serve as a valuable tool that supports—rather than replaces—the expertise and insight of legal professionals. Visit LEAP's legal AI solutions page to see how our suite of integrated AI tools, including a comprehensive library of pre-built prompt templates spanning multiple practice areas and jurisdictions, can help your firm work faster and smarter.