Building an Autonomous Financial Analyst with o1 and o3-mini

Financial AI platform Endex is developing an AI Analyst that retrieves, synthesizes, and reasons over complex financial data using OpenAI's o1 and o3-mini models. The platform aims to go beyond simple search results, providing structured thinking and deep analysis to assist the workflows of investment professionals.

Unlike traditional RAG (Retrieval-Augmented Generation) approaches, Endex's agent uses OpenAI's reasoning models to review data like a financial analyst. This allows it to identify facts, spot inconsistencies between data, and add context to metrics.

Key features and results include the following:

• Multi-step financial reasoning: The o1 model enables accurate multi-step reasoning without complex prompt chains.
• Improved efficiency and speed: Using o3-mini, latency has been reduced to a third of what it was before, supporting confidential information package (CIP) analysis and automated financial model adjustments.
• Cross-validation: Identifies inconsistencies in financial data and finds restatements noted in footnotes, alerting analysts to them.
• Multimodal analysis: Through o1's vision capabilities, it analyzes tables and charts within prospectuses, internal materials, Excel models, and 8-K reports.
• Automated output generation: Automatically generates reports in various forms, including emails, documents, Excel models, and slide decks.

In actual blind tests, financial professionals preferred responses generated by the o1 model over non-reasoning models by 70%.