What Works: Crafting Effective Prompts
Let’s start with examples of how to frame prompts that yield high-quality results.
Drafting a Contract Clause
Example 1: Vague Prompt
"Draft a non-compete clause."
Why It Doesn’t Work: This prompt is too broad. It fails to specify the type of contract, jurisdiction, industry, duration, and other crucial factors necessary for drafting a legally sound clause. As a result, the AI is likely to generate something overly generic and impractical.
Example 2: Specific Prompt
"Draft a non-compete clause for an employment contract under UK law, restricting the employee for 12 months within the software industry."
Why It Works: This prompt includes key details—contract type, jurisdiction, duration, and industry—allowing the AI to generate a clause that is far more relevant and useful.
Conducting Legal Research
Example 1: Vague Prompt
"Find cases on breach of contract."
Why It Doesn’t Work: Again, this prompt is too general. It does not specify the type of breach, the relevant jurisdiction, or the specific legal issue, which can lead to an overwhelming amount of irrelevant results.
Example 2: Specific Prompt
"Find UK Supreme Court judgments from the past 10 years concerning material breaches of contract in commercial property transactions."
Why It Works: This prompt narrows the scope by specifying the jurisdiction, time frame, type of breach, and legal context. This dramatically increases the likelihood of obtaining relevant and actionable case law.
What Doesn’t Work: Common Pitfalls
Now, let’s examine some common mistakes that can lead to poor AI results—and how to avoid them.
Overloading the Prompt with Too Much Information
While detail is important, an excessively complex prompt can confuse the AI.
Example of an Overloaded Prompt
"Draft a non-compete clause under UK law preventing an employee in the software industry from working for a competitor, using trade secrets or confidential information, within Europe, for 18 months, ensuring compliance with all relevant legal guidelines and regulations."
Why It Doesn’t Work: This prompt tries to accomplish too much at once, which may lead to confusion or an incoherent result. Instead, breaking the task into manageable steps improves clarity and accuracy.
Better Approach:
- Step 1: "Draft a non-compete clause for an employment contract under UK law, limited to 18 months within the software industry."
- Step 2: "Incorporate language to protect trade secrets and confidential information."
- Step 3: "Ensure the clause aligns with relevant UK employment law and case law."
This structured approach helps AI generate legally sound and coherent outputs.
Relying on AI Without Verification
AI is a tool, not a substitute for legal judgment. Always review and verify AI-generated content to ensure accuracy and compliance.
Conclusion
Prompt engineering is becoming an essential skill for modern legal practice. By crafting precise and well-structured prompts, you can significantly enhance your efficiency and the quality of your legal work. However, AI should be seen as an aid—not a replacement for legal expertise. For example, when researching relevant UK Supreme Court cases, treat the AI-generated results as a starting point and verify them using official databases such as BAILII or Westlaw. Once you have identified relevant judgments, AI can then assist in further analysing and extracting insights from them.
Final Thought
Think of prompt engineering as instructing a junior lawyer or trainee: the clearer your instructions, the better the outcome. Just as you would review a trainee’s work, you must also scrutinise AI-generated content. Refine your prompts with the same care as a legal document. The better your prompts, the better the AI output—adding real value to your legal practice.
