Prompting: The Art of Crafting Precise Requests for AI

Prompting—the art of formulating precise requests to interact with an AI model—is essential for producing high-quality documents. A well-written prompt can help you obtain clear legal analyses, concise summaries, or structured documents tailored to your needs.

Here are the best practices for effective prompting in a legal context, illustrated with concrete examples.

Note: These examples should be entered in the Prompt section of your Templates in Whisperit. To create a template, see this page.

1. Provide Clear and Precise Context

Start by defining the AI’s role and the expected goal. This guides the AI in drafting the content.

Example:

## Context
You are a legal assistant specialized in contract analysis and memorandum drafting. Your goal is to provide clear, structured analyses adapted to Swiss law and to suggest relevant clauses or recommendations. Use Markdown format to structure the results.

Why?

• It steers the AI toward responses that are relevant to the legal context and specific jurisdiction.

2. Use Explicit Instructions

Give clear instructions on sections, legal terms, and formatting requirements.

Example:

### General Guidelines
1. If a clause is missing, clearly state "Clause missing."
2. If a clause is standard, mention that it complies with usual legal norms.
3. Use clear and professional legal language.
4. Avoid interpreting the parties’ intentions if they are not explicitly stated in the contract.

Why?

• Ensures the generated results remain within the legal framework and avoid ambiguity.

3. Specify the Output Format

Clearly state the expected output format—for example, sections, comparative tables, or clause hierarchies.

Example:

Structure the report as follows:
- **Contract Summary**: Include the parties, subject matter, and duration.
- **Clause Analysis**: Mention strengths, weaknesses, and potential risks.
- **Recommendations**: Provide suggestions to improve or correct identified clauses.

Example Output:

## Contract Summary
- **Parties**: Company A (seller) and Company B (buyer).
- **Subject**: Sale of IT equipment.
- **Duration**: Contract valid for 1 year, renewable by tacit agreement.

## Clause Analysis
- **Non-compete clause**: Valid, but excessive duration (5 years).
- **Liability clause**: No limitation of liability for indirect damages.
- **Force majeure**: Standard, in line with common practice.

## Recommendations
- Reduce non-compete clause duration to 2 years.
- Add a financial cap to the liability clause.

Why?

• A clear structure makes it easier for lawyers to directly use the output.

4. Provide Lists of Examples or Specific Terms

Include proper nouns, legal terms, and examples the model must respect.

Example:

Use the correct spelling for the following terms:
- Clauses: non-compete, force majeure, early termination, limitation of liability.
- Legal texts: Swiss Code of Obligations (CO), Federal Data Protection Act (FADP).
- Jurisdictions: Swiss Federal Supreme Court, Cantonal Courts.

Why?

• Improves legal precision and ensures compliance with local standards.

5. Handle Missing or Partial Information

Specify how to handle missing details in a legal document.

Example:

- If a clause is missing, state "Clause missing."
- If a clause is vague, state "Unclear wording."

Example Output:

## Clause Analysis
- **Confidentiality clause**: Missing. Recommendation: add a clause defining covered information.
- **Termination clause**: Unclear wording regarding valid grounds.

Why?

• Highlights potential gaps without unnecessary interpretation.

6. Prioritize Clarity and Precision

Request simple, precise language, avoiding unnecessary jargon.

Example:

Use clear and direct language suitable for legal professionals and non-specialist clients.

Why?

• Improves readability and facilitates communication with external parties.

7. Use Standardized Formats and Units

Define consistent formats for dates and amounts.

Example:

- Write dates in DD.MM.YYYY format and times in HH:MM.
- For amounts, use the currency (CHF, EUR, USD) and a clear number format (e.g., 1'000'000 CHF).

Example Output:

- **Signing date**: 28.12.2024
- **Amount**: 1'500'000 CHF

Why?

• Standardization simplifies interpretation and prevents errors.

8. Test and Refine Your Prompts

Try your prompts with several sample contracts or legal documents to refine results.

Example of an Optimized Prompt:

You are a legal assistant specialized in Swiss law. Analyze the following contract, identifying strengths, weaknesses, and potential risks. Provide recommendations for each problematic clause. Use clear Markdown formatting.

9. Provide Examples to Guide the AI

Add concrete examples to steer the response.

Example:

Here is an example of the expected output:

## Clause Analysis
- **Non-compete clause**: Valid but excessive in duration.
- **Confidentiality clause**: Missing.
- **Termination clause**: Unclear wording.

Generate a similar report based on the following contract.

Why?

• Examples help the AI understand your precise expectations.

Conclusion

By applying these best practices, lawyers can optimize their interactions with AI tools to generate accurate, structured legal analyses.

Take the time to define clear, detailed prompts and include examples to guide the AI. With these strategies, you’ll improve the quality of your legal documents and increase efficiency.