AI Agents in Financial Reporting

Introduction

• Speaker: Richie
• Topic: AI agents and their application in financial reporting
• Purpose: Automate tedious parts of jobs using AI, specifically report writing

Overview of AI Agents

• AI agents help automate boring, repetitive tasks
• Generative AI enhances automation capabilities
• Focus on building agents for financial reporting

Session Structure

• Interactive session with a focus on understanding possibilities with AI agents
• More of a watch-along than a code-along due to complexity
• Emphasis on understanding architecture and code patterns

Guest Introduction

• Guest Speaker: Jita Tunder
• Credentials: Award-winning lead data scientist at FIT Group
• Achievements: Listed as one of 100 most influential AI leaders in the USA

Key Concepts Covered

AI Agents vs. Agentic AI

• AI agents: Task-oriented, limited autonomy
• Agentic AI: Goal-oriented, higher autonomy
• Importance of human oversight in AI agent processes

Building AI Agents

• Use of open-source technologies to build financial agents
• Components of AI Agents:
  • Task requests
  • Interaction with tools and memory
  • Planning and task execution

Application in Financial Services

• Streamlining operations by automating repetitive tasks
• Enhancing customer engagement through personalized advice
• Compliance monitoring can benefit from AI agents

Implementation

Tools and Setup

• Use of Grok and Agno platforms
• Setting up API Keys
• Grok: Open-source AI inference
• Agno: Framework for chaining agentic workflows

Building AI Agents

Agent 1: Research Agent

• Function: Conducts web search and generates reports
• Toolkits Used: DuckDuckGo, Newspaper3k
• Instructions: Step-by-step process for generating reports
• Output: Executive summaries, key findings, impact analysis

Agent 2: Rag-Based Query

• Function: Retrieval Augmented Generation with knowledge base
• Tools: PG Vector database (Open-source)
• Embedding Model: Sentence Transformers from Hugging Face

Agent 3: Stock Analysis

• Function: Analyzing stocks and generating financial comparisons
• Data Source: Yahoo Finance API
• Output: Stock performance reports, Comparative analysis

Agent 4: Evaluation Framework

• Function: Use LLM as a judge to evaluate agent outputs
• Metrics: Faithfulness, Context relevance, Completeness

Important Considerations

• Importance of structuring outputs and creating evaluation frameworks
• Setting clear instructions and output formats
• Ethical considerations in AI agent deployment
• Limitations of using open-source tools and ensuring accuracy

Conclusion

• AI agents can significantly enhance productivity and efficiency in financial reporting
• Importance of continuous evaluation and updating of AI agent frameworks
• Encouragement to explore agentic AI and its potential applications

Additional Resources

• Recommended readings and courses for further learning on AI agents
• Contact information and support for questions