> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bono.network/llms.txt
> Use this file to discover all available pages before exploring further.
# AI Research
> Research legal questions with AI-powered analysis and verified sources
## What is AI Research?
AI Research is Arbiter's legal research assistant. Ask questions in plain language and get comprehensive analysis backed by real sources. Whether you need a quick definition, case law research, or multi-jurisdictional analysis, Arbiter finds relevant authorities and synthesizes them into actionable answers.
**What you can research:**
* Case law and precedents
* Statutory and regulatory questions
* Multi-jurisdictional comparisons
* Strategic analysis and risk assessment
* Contract interpretation questions
* Procedural and practice questions
**Works globally.** Arbiter researches in any jurisdiction and language. Ask about German contract law, French administrative procedure, or compare GDPR with CCPA - Arbiter searches in native legal terminology for better results.
***
## Getting Started
### Starting a Research Session
1. Click **Legal Research** on the dashboard, or switch to AI Research mode from the sidebar
2. Click **New Chat** to start a fresh conversation
3. Type your question and press Enter
### Writing Good Questions
The more context you provide, the better the answer. Include:
* **Jurisdiction** - "Under New York law\..." or "In the UK..."
* **Parties** - Who is involved and their relationship
* **Key facts** - Relevant circumstances
* **What you need** - Analysis, comparison, risk assessment, etc.
**Example - Vague:**
> "Is this non-compete enforceable?"
**Example - Specific:**
> "Is a 2-year non-compete for a senior software engineer enforceable under California law if the employee was terminated without cause?"
***
## Research Modes
Choose how deeply you want Arbiter to analyze your question:
### Standard Mode
**Best for:** Quick questions, definitions, straightforward research
* Single AI analyst responds
* Fast (\~30 seconds)
* Web research available
* Good for routine questions
**Example questions:**
* "What is the statute of limitations for breach of contract in Texas?"
* "Define 'piercing the corporate veil'"
* "What are the elements of fraudulent misrepresentation under English law?"
### Deliberation Mode
**Best for:** Complex questions, strategic analysis, multi-jurisdictional issues
* Three AI analysts work together, each with a different perspective
* They see each other's analysis and refine their views
* You get a synthesis combining all perspectives
* Takes 2-5 minutes
**Example questions:**
* "Compare non-compete enforceability in California, Texas, and New York for tech employees"
* "What are the risks and benefits of incorporating in Delaware vs. Nevada for a SaaS company?"
* "Analyze the litigation strategy options for a breach of fiduciary duty claim against a former director"
### Grand Deliberation Mode
**Best for:** High-stakes decisions, maximum analytical depth, critical strategic planning
* Same three-analyst approach with extended reasoning
* Exhaustive multi-phase analysis
* Maximum depth and sophistication
* Takes 30-60 minutes
**Example questions:**
* "Develop a comprehensive IP protection strategy for international expansion into the EU, Asia, and Latin America"
* "Analyze all potential claims and defenses in a complex M\&A dispute involving earn-out provisions"
Start with **Standard Mode** for most questions. Upgrade to **Deliberation** when you need multiple perspectives. Save **Grand Deliberation** for truly critical decisions.
***
## The Three Analysts
In Deliberation and Grand Deliberation modes, three AI analysts with different perspectives analyze your question:
**Strategic Lead**
Balanced analysis weighing precedent, policy, and practical outcomes. Looks for the most defensible position.
**Risk Guardian**
Conservative perspective focused on compliance, ethics, and risk mitigation. Identifies what could go wrong.
**Aggressive Tactician**
Creative and assertive approach. Proposes novel strategies, identifies leverage points, and pushes boundaries.
Each analyst provides their perspective, then Arbiter synthesizes everything into a comprehensive answer. You see all perspectives so you can choose the approach that fits your situation.
***
## Web Research (GitLaw)
Enable web research to have Arbiter search for current sources:
### What It Searches
* Case law databases
* Statutes and regulations
* Legal commentary and analysis
* News and recent developments
* Academic sources
### How to Use It
1. Click the **globe icon** in the chat input to enable web research
2. Choose your mode:
* **GitLaw Premium** - Deep analysis, better citations (\~200 tokens)
* **Fast Search** - Quick results, lower cost (\~40 tokens)
3. Ask your question
4. See the research cards showing sources found
### When to Use Web Research
**Enable for:**
* Current case law ("recent cases on...")
* Statutory questions ("what does the statute say about...")
* Regulatory updates
* Jurisdiction-specific questions
**Skip for:**
* General legal principles
* Definitions and concepts
* Hypothetical analysis
* Questions about documents you've uploaded
Web research adds 30-60 seconds to response time. If you just need general analysis, you can skip it.
***
## Adding Documents to Your Research
You can add documents to your research chat for context. Click the **folder icon** in the chat input to:
* Select documents from your library
* Add multiple documents at once
* Give Arbiter full context when answering questions
**Example uses:**
* "Compare the indemnification clauses across these three contracts"
* "Which of these agreements has the most favorable termination provisions for our client?"
* "Summarize the key differences between the draft and final versions"
When working with a Matter, documents from that matter are automatically available as context. You don't need to add them manually.
***
## Visualizations
Ask Arbiter to create visual representations of complex information:
### Timeline
For chronological information - deadlines, events, procedural steps.
> "Create a timeline of the key events in the Enron scandal"
> "Show me the procedural steps for a federal civil lawsuit"
### Comparison Table
For multi-option or multi-jurisdictional analysis.
> "Compare data privacy requirements under GDPR, CCPA, and LGPD in a table"
> "Show the differences between LLC, S-Corp, and C-Corp in a comparison"
### Entity Graph
For relationships between parties, entities, or concepts.
> "Map the corporate structure and relationships in this case"
> "Show the parties and their contractual relationships"
### Decision Tree
For decision frameworks and analysis paths.
> "Create a decision tree for whether this contract is enforceable"
> "Show the analysis framework for determining personal jurisdiction"
Just ask for the visualization you want: "Can you show that as a timeline?" or "Put that in a comparison table."
***
## Reasoning Transparency
Every AI response includes a collapsible **reasoning card** (purple) that shows how the AI reached its conclusion.
### What You'll See
* How it approached the problem
* What factors it considered
* Why it weighted certain issues more heavily
* Its analytical framework
### Why It Matters
* **Client communication** - Explain the basis for advice
* **Compliance** - Document your research process
* **Quality check** - Verify the AI's reasoning makes sense
* **Learning** - Understand the analytical approach
Click the purple card to expand and see the full reasoning.
***
## Example Research Sessions
**Question:** "Our client is a software company in Austin. They want to enforce a non-compete against a departing VP of Engineering who is joining a competitor. The non-compete is 18 months, covers Texas and California, and was signed 3 years ago. What are our chances?"
**Mode:** Deliberation with Web Research
**What you'll get:**
* Analysis of Texas non-compete law (generally enforceable)
* Analysis of California implications (likely unenforceable there)
* Recent case law on tech employee non-competes
* Strategic recommendations from each analyst
* Synthesized advice on how to proceed
**Question:** "A founder is starting a SaaS company. They expect to raise VC funding within 2 years and eventually go public. Should they incorporate in Delaware, their home state of Texas, or somewhere else? What entity type?"
**Mode:** Standard with Web Research
**What you'll get:**
* Comparison of Delaware vs. other jurisdictions for VC-backed companies
* Entity type analysis (C-Corp for VC path)
* Practical considerations (costs, franchise taxes, registered agents)
* Recommendation with reasoning
**Question:** "We're drafting a distribution agreement between a US manufacturer and a German distributor. What are the key differences between US and German law we need to consider? The contract will be governed by New York law."
**Mode:** Deliberation with Web Research
**What you'll get:**
* Comparison table of US vs. German contract law differences
* Specific issues with New York choice of law for German party
* Mandatory German law provisions that may apply despite choice of law
* Risk analysis and drafting recommendations
**Question:** "What is the business judgment rule?"
**Mode:** Standard (no web research needed)
**What you'll get:**
* Clear definition
* Key elements
* When it applies
* Common exceptions
* \~30 second response
**Attached Documents:**
* Original Share Purchase Agreement (SPA)
* Amendment No. 1 to SPA
* Earn-out calculation spreadsheets (3 years)
* Board minutes from 4 meetings
* Email chain between CFOs
* Expert valuation report
* Opposing counsel's demand letter
**Question:**
"URGENT - Board meeting in 48 hours. We represent TechAcquire Inc. in a 47 million dollar earn-out dispute with the former shareholders of DataFlow Systems. The sellers claim we deliberately suppressed revenue recognition and diverted opportunities to avoid triggering the Year 3 earn-out milestone of 12 million dollars.
Key facts: (1) We changed revenue recognition from point-of-sale to subscription model 6 months post-closing, reducing recognized revenue by approximately 8 million in Year 3; (2) Three enterprise deals totaling 4.2 million were closed by our parent company's sales team rather than the DataFlow unit; (3) The CFO email chain shows internal discussion about 'managing the earn-out exposure'; (4) The SPA earn-out clause references 'revenue recognized in accordance with GAAP as consistently applied' but doesn't define the baseline; (5) Delaware law governs.
I need: (1) Comprehensive liability analysis - are we exposed on the accounting change and/or opportunity diversion theories?; (2) Assessment of the implied covenant of good faith claim under Delaware law; (3) Damages exposure range; (4) Litigation risk assessment if this goes to Chancery Court; (5) Settlement strategy and recommended range; (6) Analysis of whether the board can approve a settlement or if shareholder approval is needed given the amount; (7) Comparison of our position versus their position with percentage likelihood assessments; (8) Recommended talking points for the board presentation; (9) Alternative structures for settlement (installments, equity, escrow release); (10) Research any recent Delaware Chancery cases on earn-out manipulation claims from 2022-present."
**Mode:** Grand Deliberation with Web Research
**What you'll get:**
* Deep analysis from all three analysts with competing perspectives
* Arbiter's balanced assessment of liability exposure
* Anais's conservative risk analysis flagging worst-case scenarios
* Abaddon's aggressive defense strategies and counterattack options
* Synthesized liability assessment with percentage estimates
* Recent Chancery Court case law on earn-out disputes
* Settlement range recommendation with supporting analysis
* Board presentation framework
* Detailed comparison table of both parties' positions
* Decision tree for settlement vs. litigation path
***
## Chat Management
### Your Conversations
All your research chats are saved in the sidebar. You can:
* **Continue** any previous conversation
* **Search** past chats by content
* **Rename** chats for better organization
* **Delete** chats you no longer need
### Chat Titles
Arbiter automatically generates descriptive titles from your first message. You can rename them anytime by clicking the chat menu.
### Context Within Chats
Each chat maintains context, so you can ask follow-up questions:
> "What about if the non-compete was signed in California?"
> "How would this analysis change for a junior employee?"
> "Can you expand on the third point?"
***
## Tips for Better Research
Include jurisdiction, parties, key facts, and what you need. Vague questions get vague answers.
Standard for quick questions, Deliberation for complex analysis, Grand Deliberation for critical decisions.
If you need citations, ask: "Please cite relevant case law" or enable web research.
For comparisons, timelines, or relationships, ask for a visual. It often clarifies complex information.
Don't start a new chat for related questions. Follow up in the same conversation to maintain context.
***
## Troubleshooting
Response times vary by mode:
* Standard: \~30 seconds
* Deliberation: 2-5 minutes
* Grand Deliberation: 30-60 minutes
This is normal - the AI is doing deep analysis. Don't refresh the page.
* Try Deliberation mode for more depth
* Be more specific in your question
* Ask follow-up questions
* Enable web research for current sources
* **Switch to GitLaw Premium mode** - Fast Search uses semantic matching which works well for common topics, but Premium mode retrieves content contextually with deeper analysis, finding more relevant and nuanced sources
* Rephrase with more specific legal terms
* Include the jurisdiction explicitly
* Try different search terms in a follow-up
* Some niche topics may have limited online sources
Use Document Analysis instead. Upload your document and chat with Arbiter about it there. The AI will have the full document context.
***
## Next Steps
Learn more about web research capabilities
Deep dive into multi-analyst analysis
Research within your documents
Research across multiple documents