ChatGPT Prompts for Lawyers: Guide to Mastering Generative AI
How to Create Effective Input for AI Tools
Artificial intelligence (AI) isn’t just for sci-fi movies or Silicon Valley anymore; it’s crashing into the world of law, ushering in a new era of legal strategy and client service. This new era begs the question: what does it take to be your law firm’s AI superhero? Years of coding experience? A multimillion-dollar budget? A robot sidekick named J.A.R.V.I.S.?
Actually, none of the above. All you need are strong prompt engineering skills. Crafting effective prompts for ChatGPT and other AI tools involves designing inputs to guide AI chatbots toward responses that capture the right tone, context, and relevant information. While chatbots will still respond to general law-related questions, well-crafted prompts are essential for eliciting tailored, authentic responses that align with your specific needs. There are five key elements behind creating effective Gen AI prompts for lawyers:
- Personas
- Context
- Desired outcomes
- Relevant information
- Follow-up adjustments
.
Personas
One of the most powerful tools in prompt engineering is the use of personas. By assigning a specific identity, role, or expertise to the prompt, you can shape the tone, style, and depth of the response. An effective prompt will begin with an identification of a relevant persona before explaining the context of the query. Using personas effectively can transform your interactions with AI, making the responses more engaging, contextually appropriate, and highly relevant to your specific legal needs.
For example, if you need a detailed explanation of contract law intricacies, your prompt can begin with:
.
“Imagine you are a veteran contract attorney…”
Or, when preparing for a client meeting on intellectual property law, you may begin your prompt with:
“Assume the role of an experienced intellectual property attorney…”
.
Context
Providing detailed, specific context helps the AI understand the background and objectives, leading to more accurate and relevant responses. This is particularly important for legal scenarios where the specifics of a case or dispute are unique, and the AI requires complete contextual information to reason and respond properly.
To exemplify this, let’s continue the prompts from the previous example, starting with the prompt about contract law. To add context, you might add:
.
“…advising a tech startup on the key elements of contract enforceability. The startup recently signed a software development contract with a major corporation, but the corporation is now trying to amend the terms retroactively.”
.
For the intellectual property example, if you are advising a client about a dispute, you might contextualize your prompt by adding:
.
“…mediating a dispute between two companies. Company A claims that Company B has infringed on their patent for a new type of eco-friendly packaging, while Company B argues that their design is sufficiently different and non-infringing.”
.
Relevant information
While ChatGPT has access to a large dataset, it lacks access to proprietary, paywalled, or recent legal information. A solution is to use a tool like DeeDee®, Dye & Durham’s always-on legal information AI assistant, which integrates a robust, up-to-date legal database. DeeDee AI provides more accurate and context-specific legal information than ChatGPT’s basic training data.
.
Desired outcomes
Specifying the desired outcome or product you expect from the AI-generated response is a crucial element in crafting effective prompts. This goes beyond providing context to outlining the exact information, format, and length you need to meet your objectives effectively.
There is no limit to how specific and customized your queries can be with ChatGPT and most generative AI tools, so feel free to include as many details about your desired outcome as you like ─ the more the better!
Now, let’s add a specific desired outcome to each of the previous examples:
.
“… Provide a comprehensive legal analysis in the form of a detailed memorandum. The memorandum should include an overview of contract law principles applicable to modification, a thorough examination of the original contract terms, a comparative analysis of proposed amendments, and recommendations for negotiation strategies. The document should be approximately 1,000 words in length, structured with clear headings for each section.”
“… Provide a concise legal opinion outlining the likelihood of success for Company A’s infringement claim. The opinion should briefly summarize the key points of contention, cite relevant patent laws and precedents, and conclude with a recommendation for potential next steps. Limit the response to 3-4 paragraphs to ensure clarity and focus.”
.
Follow-up adjustments
Last but not least, a skill that is integral in creating the best ChatGPT prompts for lawyers is writing follow-up adjustments. Using the conversational nature of AI chatbots, you can continue to tune an AI’s response, long after your first query. Chatbots almost never get it right on their first try, so making follow-up adjustments allows you to further customize your perfect response.
To write an effective follow-up adjustment, ensure you:
- Are specific about what needs clarification or enhancement
- Provide updated context as necessary
- Ask targeted questions to deepen understanding
- Give constructive feedback to refine the AI’s responses over time.
Some examples of effective follow-up adjustments are:
.
“Could you revise the response to convey a more formal tone suitable for a client briefing?”
“Could you delve deeper into the legal implications of the recent regulatory changes on our intellectual property strategy?”
“Please incorporate recent court decisions on non-compete agreements into your analysis of our employee contract dispute.”
.
Follow-ups can be used to generate sub-content from previously generated responses. For example:
.
“You suggested several negotiation tactics for our contract dispute. Can you outline specific steps to implement these strategies effectively?”
“From your discussion on contract enforceability issues, what alternative legal theories or defenses could be explored to strengthen our position?”
The final prompts look like this:
“Imagine you are a veteran contract attorney advising a tech startup on the key elements of contract enforceability. The startup recently signed a software development contract with a major corporation, but the corporation is now trying to amend the terms retroactively. Provide a comprehensive legal analysis in the form of a detailed memorandum. The memorandum should include an overview of contract law principles applicable to modification, a thorough examination of the original contract terms, a comparative analysis of proposed amendments, and recommendations for negotiation strategies. The document should be approximately 1,000 words in length, structured with clear headings for each section.”
.
“Assume the role of an experienced intellectual property attorney mediating a dispute between two companies. Company A claims that Company B has infringed on their patent for a new type of eco-friendly packaging, while Company B argues that their design is sufficiently different and non-infringing. Provide a concise legal opinion outlining the likelihood of success for Company A’s infringement claim. The opinion should briefly summarize the key points of contention, cite relevant patent laws and precedents, and conclude with a recommendation for potential next steps. Limit the response to 3-4 paragraphs to ensure clarity and focus.”
Risks, Guardrails and Best Practice
While chatbots like ChatGPT can be a powerful, useful tools, it is crucial that they are used properly and carefully. Some risks involved with the misuse of legal AI tools include:
- Confidentiality breaches: Many generative AI tools lack guarantees of confidentiality, as their algorithms are often trained using input provided by users, which can be stored or processed by the provider. This means any private or sensitive information shared with these tools could be at risk of being exposed, either through data breaches or unintended usage within the AI’s training datasets.
- Best practice: Do not disclose any confidential client or proprietary information in your chatbot prompts.
- Lack of current information: ChatGPT has knowledge cutoffs (GPT-3.5: 2021, GPT-4: 2023) that limit its ability to pull current information from the internet.
- Best practice: Do not rely solely on chatbots for important or current information. Always consult and verify from reliable, verifiable information sources.
- Hallucinations: When a chatbot does not have sufficient information to answer a query, it will sometimes generate completely fabricated information. These instances are called “hallucinations”. It can be hard to identify hallucinations in chatbot responses, and if one flies under your radar, it can be detrimental to the credibility of your firm.
- Best practice: independently verify all factual information received from chatbots and manually provide key information when necessary.
- Bias: Chatbots like ChatGPT are trained on information from the internet, so they are prone to bias. These biases can lead to discriminatory or stereotypical content generation, which can damage public trust in your firm.
- Best practice: Write neutral prompts that avoid bias, and use follow ups to remove bias when identified.
Ready to elevate your practice?
Contact us to learn more, request a demo or get started with Unity® Practice Management, the all-in-one cloud-based software solution designed to help you do more with less effort.