Skip to main content

​
Overview

Finance Agent uses advanced AI to analyze your questions and retrieve relevant information from earnings transcripts, 10-K filings, and news sources. Understanding how to structure your questions will help you get more accurate and comprehensive answers.

​
Answer Modes

The agent automatically determines the appropriate answer mode based on your question’s complexity:
1

Direct Mode

For simple factual lookups (e.g., “What was AAPL’s revenue in Q1 2025?”).
  • Max iterations: 2
  • Max tokens: 2,000
  • Best for: Quick facts, single data points
2

Standard Mode

For moderate analysis questions (e.g., “How did MSFT’s cloud revenue grow in 2024?”).
  • Max iterations: 3
  • Max tokens: 6,000
  • Best for: Quarterly comparisons, trend analysis
3

Detailed Mode

For comprehensive research (e.g., “Analyze TSLA’s profitability trends over the last 8 quarters”).
  • Max iterations: 4
  • Max tokens: 16,000
  • Best for: Multi-quarter analysis, deep dives
4

Deep Search Mode

For exhaustive research requiring maximum iteration.
  • Max iterations: 10
  • Max tokens: 20,000
  • Best for: Complex multi-company comparisons, comprehensive reports

​
Ticker References

Always use ticker symbols with the $ prefix for best results: 
$AAPL
$MSFT
$GOOGL
$TSLA
The agent supports ticker aliases (e.g., TSMC → TSM), but using official ticker symbols ensures accurate results.

​
Temporal References

Specify time periods clearly to get data from the right quarters:

​
Quarter Format

2025_q1
2024_q4
2023_q2

​
Available Data

  • Earnings Transcripts: Quarterly data starting around 2023 (varies by company)
  • 10-K Filings: Annual data from fiscal year 2019 onward (FY2019, FY2020, etc.)
  • News: Recent developments and announcements
The system has processing limits:
  • Max quarters: 12 (3 years of quarterly data)
  • Max tickers: 8 companies per query
Queries exceeding these limits will be automatically capped with a notification.

​
Question Types

The agent recognizes different question types and routes them to appropriate data sources:

​
Single Company Questions

Examples
What did $AAPL say about iPhone sales in Q1 2025?
Show me $MSFT's revenue breakdown from their latest 10-K
What are $TSLA's biggest risks mentioned in their 2024 10-K?

​
Multi-Company Comparisons

Examples
Compare gross margins for $AAPL and $MSFT in Q4 2024
Which had better revenue growth: $GOOGL or $META?
How do $NVDA, $AMD, and $INTC compare on R&D spending?

​
General/Broad Questions

Examples
Which tech companies mentioned AI in their latest earnings?
What are the top risks in the semiconductor industry?
Show me companies with strong margin expansion

​
Data Source Hints

You can explicitly request specific data sources: 
What did $AAPL say about services in their latest earnings call?
From the Q1 2025 earnings transcript, what did management say about margins?

​
Conversation Context

The agent maintains conversation history (last 5 exchanges, max 4,000 chars per message) for follow-up questions:
Example Conversation
User: What was AAPL's revenue in Q1 2025?
Agent: [Provides answer]

User: What about their services revenue?
// Agent understands "their" refers to AAPL from previous context

User: Compare them to MSFT
// Agent compares AAPL (from context) with MSFT

User: What about $WMT?
// Agent switches to WMT (explicit new ticker overrides context)
When asking follow-up questions:
  • Use pronouns like “they”, “their”, “those companies” to reference prior context
  • Name a specific ticker (e.g., “What about $WMT?”) to switch to a new company
  • The agent automatically inherits time periods from previous questions unless you specify otherwise

​
Best Practices

1

Be Specific with Time Periods

Instead of “recent”, use “Q1 2025” or “last 8 quarters”
2

Use Official Tickers

$AAPL is better than “Apple Inc.”
3

Request Specific Sections

“Show me the risk factors from $MSFT’s 2024 10-K” is more precise than “What are MSFT’s risks?”
4

Leverage Multi-Source Queries

“What did $NVDA say about data center revenue in Q4 2024 earnings, and what does their 10-K say about segment performance?”
5

Use Follow-ups Effectively

Build on previous questions: “What about operating expenses?” (after asking about revenue)

​
Advanced Query Examples

Comprehensive Analysis
Analyze $AAPL's gross margin trends over the last 8 quarters and compare with guidance from their latest 10-K
Cross-Source Research
What did $MSFT say about Azure growth in Q2 2024 earnings, and how does this compare to the cloud segment revenue in their FY2024 10-K?
Competitive Intelligence
Compare R&D spending as a percentage of revenue for $GOOGL, $META, and $AMZN over the last 3 years
Risk Analysis
What are the top 5 risk factors mentioned in $TSLA's latest 10-K and did they address any of these in Q4 2024 earnings?

​
Iterative Improvement

The agent uses an iterative self-improvement process:
  1. Initial Answer: Generated from retrieved context
  2. Evaluation: Assesses confidence and completeness (uses GPT-4.1-mini)
  3. Follow-up Search: Identifies gaps and retrieves additional context if needed
  4. Re-generation: Produces improved answer with expanded context
  5. Repeat: Continues until max iterations or high confidence (0.90+ threshold)
The agent uses all available iterations by default to ensure comprehensive answers. Early stopping only occurs if confidence exceeds 0.95 (near-perfect).

Build docs developers (and LLMs) love

Get started for freeTalk to us