Summary
Eve's Reasoning Mode is an advanced feature designed for complex legal analysis and multi-step thinking tasks. In this guide, you'll learn how to effectively utilize Reasoning Mode for deep case analysis, strategy building and more.
Unlike Chat Mode's quick conversational responses, Reasoning Mode provides thorough, structured outputs ideal for sophisticated legal work. While it requires more processing time and context, it provides more comprehensive answers for complex document analysis, multi-claim evaluations, and detailed strategic planning. This guide will walk you through the technical requirements, limitations, and proven strategies to maximize your use of this advanced feature.
Some Background on Reasoning Mode
Before using Reasoning Mode, it's important to understand its technical requirements and limitations. Unlike Chat Mode, Reasoning Mode requires extensive context upfront and operates with longer processing times - typically several minutes rather than instant responses. This increased processing time is due to the sophisticated reasoning models working to provide thorough, structured analysis of complex legal tasks.
Reasoning Mode consumes significantly more tokens than Chat Mode, which means it requires more computational resources to process your requests. This higher token usage is necessary to support the deep analysis and multi-step thinking capabilities that make Reasoning Mode particularly valuable for complex legal work. The feature is specifically integrated with playbook functionality and is accessible through the blue lightbulb icon in your playbooks.
When using Reasoning Mode, you should be prepared to provide comprehensive context at the beginning of your interaction rather than through back-and-forth dialogue. The system works best when given complete information upfront, as it's designed to process and analyze all relevant information in a single, thorough pass rather than through iterative conversations. This approach, while more demanding in initial setup, results in more comprehensive and structured outputs suitable for sophisticated legal analysis and documentation.
When to use Reasoning Mode vs Chat Mode
If you want quick, iterative dialogue, go with Chat Mode. If you need thorough, step-by-step legal reasoning or multi-step tasks over large sets of data, Reasoning Mode is your best bet—just be ready for longer wait times and more tokens.
Examples of when to use Chat Mode
Simple Q&A: “Eve, what does this document say about retaliation?”
Fast iteration: “Draft a short paragraph,” “Revise for tone,” “Provide a bullet list.”
Ongoing conversation: stepping through small tasks or clarifications.
Examples of when to use Reasoning Mode
Deep case analysis: “Review this case and identify FEHA, PAGA, and FLSA claims.”
Strategy building: “Draft a detailed litigation roadmap, including recommended motions.”
Cross-checking multiple documents: “Identify contradictions between witness testimonies and client affidavits.”
Step-by-Step Instructions
Step 1: Access Reasoning Mode
To turn on “Reasoning Mode” simply toggle the “Reasoning” Button in Eve’s chat interface.
Step 2: Structure Your Prompt
When using Reasoning Mode, we recommend that you structure your prompt using these five essential sections:
1. Context
Establish the overall situation or background for the prompt.
Provide relevant information that the AI model needs to understand the legal scenario.
Example: "You are a legal assistant supporting a law firm in preparing for client consultations and case analysis."
2. Action
Define the specific task you want the AI model to perform.
Be clear about the desired output and format.
Example: "Identify the key legal issues and suggest possible courses of action based on the facts provided."
3. Subject
Identify the specific topic, entity, or data that the AI model will be working with.
Provide specific details or examples to narrow down the scope of the task.
Example: "The client is a small business owner facing a potential breach of contract claim from a vendor."
4. Example
Provide examples of the desired input-output format or specific cases to guide the AI model's response.
This helps the model understand the expected structure and style of the output.
Example: "For instance, if a client reports that a vendor failed to deliver goods on time as per the contract, the response could include:
Identifying whether the delay constitutes a material breach
Reviewing the contract for any force majeure or penalty clauses
Suggesting negotiation, mediation, or potential litigation as next steps"
Combined Example for Law Firms
Prompt Structure:
Context:
You are a legal assistant supporting attorneys at a law firm.
Action: Summarize the main legal issues and recommend initial steps for the attorney to consider.
Subject:
The client is an employee who believes they were wrongfully terminated.
Example:
Input: The client was dismissed after reporting safety violations to management.
Output:
Identify potential claims for wrongful termination and whistleblower protection
Advise gathering documentation, such as emails and termination letters
Suggest reviewing state and federal employment laws relevant to retaliation
Step 3: Submit and Wait
After submitting your prompt, be prepared for longer processing times (typically several minutes). Reasoning Mode is performing deep analysis and multi-step thinking to provide comprehensive results. Do not refresh or close the window during processing.
Step 4: Review and Refine
Once Reasoning Mode completes its analysis, you'll receive a structured, detailed response. As with all AI-generated content, review the output to ensure it meets your needs. If necessary, you can submit follow-up prompts, but remember to maintain the same structured approach and provide comprehensive context for any additional requests.
Reasoning Mode in Playbooks
If you would like to use Reasoning Mode while running a playbook, navigate to your playbook and locate the blue lightbulb icon. Click on it to activate Reasoning Mode for your playbook instance.
Notes and Tips
Pro Tip: For optimal results with Reasoning Mode, focus on front-loading your context rather than engaging in back-and-forth dialogue. Unlike Chat Mode, Reasoning Mode works best when given comprehensive information upfront. This allows the system to process all relevant details in a single, thorough analysis.
Best Practices for Complex Tasks:
When analyzing multiple documents, provide clear references and specify which documents relate to which aspects of your request
For strategy building, clearly outline your objectives and any constraints or considerations that should influence the analysis
When drafting complex documents, provide examples of preferred formatting or structure to guide the output
Consider breaking very large tasks into logical segments, but maintain comprehensive context for each segment
Common Issues to Avoid:
Avoid vague or open-ended requests; Reasoning Mode works best with specific, well-defined objectives
Don't interrupt the processing time by refreshing or starting new requests; allow the system to complete its analysis
Remember that Reasoning Mode isn't designed for quick, iterative conversations - save those for Chat Mode
Be mindful of token usage; extremely large context dumps may need to be streamlined to focus on the most relevant information
If you have any questions about Reasoning Mode in the Eve platform, don't hesitate to reach out to our support team at support@eve.legal.