Many lawyers often encounter problems with loopholes in documents and shallow case analysis when using AI to handle legal work, believing that AI has limited roles. In fact, the problem is not the AI itself, but the way it is used – most people still ask questions passively with a “search box” mindset, failing to realize the full potential of AI. This article will subvert this perception and propose a “command” thinking: by constructing a sophisticated prompt framework that includes identity authorization, case information, and virtual character settings, AI can transform from a simple database into a “virtual law firm” that can simulate multiple perspectives.
Many lawyer friends found out that asking AI to write an essay was full of loopholes. Analyzing individual cases, the perspective is too shallow to read. Do you think AI is like that?
This idea is wrong.
The problem is not in AI. It’s the “material” you feed it. You just treat it like an intern. Ask a question, and it will answer a sentence. You don’t give it a goal, you don’t give it a framework.
I have seen too many people and are still using AI in the way of “Baidu”. “Help me analyze the key points of so-and-so disputes”, it can only give you the big road goods on the Internet.
If you want AI to become your “outer brain”, you have to change your mind. You are not “asking questions”. You are using language to set up a sophisticated running program for it.
Today, I will give my lawyer friends a prompt framework that is enough to cope with most of your legal work.
01 Jump out of the “search box”, AI is not your intern
Forget about that dialog box first. Don’t use it as a chat tool. Think of it as a top lawyer just starting out. He has a knowledge base all over the world. But you don’t know anything about the case at hand. And he is very “rigid”, if you don’t say it, he will never do much.
So, you can’t simply “ask”. You want “authorization” and “directive”. This is the core difference. “Search box” thinking is to passively obtain information. What is the catch. The results were random and shallow.
After 10 years of interaction design, why did I transfer to product manager?
After the real job transfer, I found that many jobs were still beyond my imagination. The work of a product manager is indeed more complicated. Theoretically, the work of a product manager includes all aspects of the product, from market research, user research, data analysis…
View details >
“Instruction” thinking is when you tell the AI: “Now, you are my exclusive advisor.” “This is the case, this is the goal.” The more information you give to AI, the more accurate it will be. The closer it is to your expectations. Even, exceed your expectations.
Many people complain that AI will “talk nonsense”. That’s because you didn’t set a “guardrail” for it. It knows no boundaries. Your expertise and experience are the key to igniting AI.
Stop asking “what do you think about this case”. This is equivalent to giving up your dominance. The correct thing to do is to take out what is in your head and give it to it first. The facts of the case, the evidence, your preliminary judgment, your concerns. Then, let it play a role based on these. This role can be your colleague, your opponent, or even a judge.
In this way, AI is truly “alive”. From a library, it becomes a simulator.
02 The Ultimate Prompt Revealed: Build Your “Virtual Law Firm”
The following prompt is very long, but each part is useful.
You just copy it and fill in the blanks according to the case.
I call it the “Multi-View Adversarial Simulation” directive.
Because it essentially creates a virtual team within AI.
Directives:
Multi-perspective adversarial simulation of legal affairs
1. Core identity
You are no longer a large language model. Now, you are a top lawyer in the [Chinese mainland company law and contract law field]. With over 20 years of experience, you are adept at handling complex commercial disputes with a rigorous logic, attention to detail, and a high sensitivity to commercial risks. All your answers must be based on this identity, using professional, calm, and objective legal terminology and ways of thinking.
2. Background information
You are dealing with a specific case, and here is all the information you know, which you must use as the sole factual basis for your analysis, and must not imagine or introduce external irrelevant information:
– Case Type: [e.g., equity transfer contract dispute]
– Our parties: [e.g., Party A, transferee company A]
– Counterparty: [e.g., Party B, natural person B]
– Core Facts:
[Use points 1, 2, and 3 to clearly and emotionlessly state the key facts of the case.] For example:
1. On May 1, 2023, A and B signed the Equity Transfer Agreement.
2. The agreement stipulates that B will transfer 10% of the equity of Company C held by him to A at a price of 1 million yuan.
3. A has paid 500,000 yuan, and B has not yet registered for industrial and commercial changes.
4. A recently discovered that Company C has a huge amount of undisclosed external guarantees…]
– Key evidence:
[List the key evidence we have.] For example:
1. A scanned copy of the Equity Transfer Agreement.
2. Bank transfer voucher.
3. Company C’s latest industrial and commercial information…]
– Our demand: [Clarify our objectives.] For example, the “Equity Transfer Agreement” was terminated, requiring B to return the paid 500,000 yuan and compensate for the loss of interest. ]
3. The core task is based on the above identity and background information, you need to play as three different virtual personalities at the same time, and conduct a comprehensive and confrontational internal discussion of the case from their perspectives. The result is a structured analysis report.
4. Virtual personality setting
– Personality A: Our host lawyer
-Position: Positive and optimistic, with the primary goal of realizing our demands.
– Task: Design the most persuasive legal argument path; sort out the advantages of our case; Estimating the likelihood of winning the lawsuit and preliminary litigation strategy.
– Personality B: Insider Risk Officer
– Position: Extremely prudent, pessimistic, focused on finding all potential risks and evidentiary flaws in our case.
– Task: Point out the logical loopholes in our argument; simulate the opposing lawyer’s most favorable counterargument; Assess the risk points of our chain of evidence.
– Personality C: Senior Partner/Strategic Advisor
– Position: Neutral, objective, oriented by business interests and final solutions.
– Task: Weigh the points of view A and B; evaluating solutions beyond litigation (e.g., negotiation, mediation); Put forward comprehensive action suggestions.
5. Output format
Please follow the following Markdown format to ensure clear logic:
[Internal Discussion Report of the Case]
Case Number: [Auto-generated]
Date: [Today’s Date]
1. Personality A: Analysis by our host lawyer
1. Core Argumentation Path:
2. Analysis of our advantages:
3. Preliminary litigation strategy and odds assessment:
2. Personality B: Insider risk officer questioning
1. We demonstrate the risk points:
2. The other party may refute the point simulation:
3. Evaluation of our evidence defects:
3. Personality C: Senior partner decision-making suggestions
1. Comprehensive trade-off analysis:
2. Alternative Solution Evaluation:
3. Final Recommendations for Action:
This prompt is not a “problem”, it is a “system”. If you throw this system to AI, it will have a “meeting” with itself inside. I am the attacker, I am the defender, and finally I am the referee. This is the essence. It is also a dry product that others don’t know.
03 From beginner to mastery: the “magic change” gameplay of prompts
The prompt above is a “panacea” template. True masters know how to “magic change”. Fine-tune its settings according to different scenarios.
Give you a few ideas to discover for yourself.
Gameplay 1: Contract review
Just change it to “virtual personality”.
Personality A: Party A (the strong party) legal affairs, explain how the terms are beneficial.
Personality B: Party B’s (weak party) legal affairs, pointing out the “pits” in the contract.
Personality C: Chief legal officer, assessing the overall risk and giving advice on signing.
In the background information, paste the entire contract text in. A perfect contract review bot is there.
Gameplay 2: Negotiation strategy formulation
Also prepare for an important negotiation. Then change it to “virtual personality”.
Personality A: “Our negotiator”, formulating our goals and arguments.
Personality B: “Opponent’s negotiator (simulation)”, predicting the other party’s requirements and simulating the opponent’s attack.
Personality C: “Senior negotiator”, designing response strategies and backup plans.
Print out the results and look at them before the meeting, and you will have a bottom in your heart.
Gameplay 3: Chain Questioning and Stress Test (Advanced)
This is the most ruthless operation. After the first run, you get the report. Don’t stop. Copy a sharp question from “Personality B” (Risk Officer). Start a new conversation and ask him “chain questions”.
The prompt can be written like this:
“Continue the discussion just now. You are still the risk officer. In response to the ‘XX risk’ you raised, please expand further. Detailed Description:
1. What laws or precedents might judges base on to support this risk?
2. How will the opposing lawyer take advantage of this?
3. How do we refute? ”
See? You are using the AI spear to attack the AI shield. Let it “bar” with itself. Through this stress test, you can dig out the risk points of a case.
