Mastering advanced AI prompts for better legal insights

If you’ve mastered the basics of prompting, you’re ready to take the next step. Advanced prompting allows you to obtain nuanced outputs from AI systems that can transform how you approach complex legal tasks. At Robin AI, we’ve designed our tools to respond to advanced prompts, giving you the opportunity to customize solutions for your unique needs.

‍

Principles of Advanced Prompting

Leverage Context:

• Always provide relevant background to guide the AI, which are usually general purpose tools. Even while using specialised AI Assistants like Robin AI, there can be several different contexts within the legal sector in which a user asks questions - from managing a private equity fund’s obligations in a side letter, to a SaaS provider handling the fallout from a data breach. For the most tailored outputs, add any relevant assumptions and parameters.
• Example: "You are helping me with a task for a client in the pharmaceutical industry. Draft an intellectual property clause for a clinical trial agreement."

Granular Instructions:

• Instead of a single, broad instruction, include several, smaller instructions, clearly setting out your expectations for the AI’s response. The more detail in your instructions, the higher the quality of the AI’s output will be.
• Example: "Draft a service agreement. Specifically, ensure you include clauses for 1. termination for convenience by the service provider, 2. penalties for the customer’s breaches, and 3. arbitration as the dispute resolution method."

Format and Structure, with an Example:

• Specify the desired output format and structure. You could even provide the AI an example of a similar task you have previously done, and ask it to follow the same writing style, usage of sub-headings, etc.
• Example: "Summarise this lease agreement in bullet points, categorised by the landlord’s obligations and rights. As an example, here is a summary of a different contract that I have previously written: [insert summary].”

‍

Common Pitfalls and How to Avoid Them

Underestimating Capabilities:

• To begin with, don’t be afraid to ask for complex tasks, AI can often tackle much more than you think. State of the art AI models, like Claude Sonnet 3.7 which you can access via Robin AI’s Legal Assistant, have been shown to perform sophisticated analysis and excel at finding-a-needle-in-the-haystack type situations. However, if you don’t immediately get the response you’re seeking, then…

... Avoid Overloading Prompts:

• Sometimes, requiring the AI to perform a complicated task in a single go leads to poor results. Instead, split your task into granular sub-tasks. In this way, you can build a chain of prompts, using the output of the first sub-task as the input for the second, and so on.
• For example, you can first ask the AI to “Summarise the key risks for [X] Party in this agreement, categorised by severity.” Once it gives you its response, you can follow up with: “Propose mitigation strategies for each serious risk.”

Unclear Goals:

• Clearly define what you want from the AI’s output to avoid unhelpful, vague or incomplete responses. You should specify exactly what you want in the AI’s response, and can even define the expected word count, style, etc. Remember that, similar to any assistant, the AI won’t be able to provide a satisfactory response unless it is given accurate instructions.
  ‍

Match Your Prompt to the Outcome You Need

As your prompting becomes more advanced, it’s not just the structure of the prompt that matters - it’s the outcome you're aiming for.

At Robin AI, both Reports and the Legal AI Assistant are built on the same powerful underlying technology. But they’re designed for different purposes - so your prompts should reflect what you want to achieve.

• Use Reports when your goal is to analyse and compare information at scale - for example, extracting liability language across thousands of contracts or running a structured risk assessment. Your prompts here should focus on collecting and organising information clearly and consistently.
• Use the Legal AI Assistant when your goal is to act on that information - whether that’s drafting a revised clause, writing an internal summary, or generating a client-ready email. This is where more creative, iterative prompting comes into play.

Advanced prompting works best when your instruction matches your objective. If you ask Reports to draft a clause, or use the Assistant to scan hundreds of contracts, you won’t get the results you’re after - not because the AI isn’t capable, but because the task is misaligned with the workflow.

To get the most out of advanced prompting, focus not just on how you ask - but also on where and why. Let the prompt fit the tool, and the tool fit the task.

‍

Conclusion

Advanced prompting is about unlocking the full potential of AI tools, enabling you to use AI for the most complex legal challenges with precision and confidence. Ready to explore more? Dive deeper into Robin AI University for additional resources and tips.