Description:
AI-powered Financial Analyst workflow using CrewAI, GPT-4o, and Serper API to automate company research and generate one-page analyst briefs.
This project demonstrates how to apply agentic AI to financial analysis, automating the research and reporting process for financial firms. In this case, I focused on Morgan Stanley.
The workflow is designed to simulate the work of a junior financial analyst, delivering concise, professional reports that highlight company performance, opportunities, and market outlook; exactly the kind of deliverables recruiters and consulting managers expect to see.
The system uses two agents:
- Researcher Agent → Collects structured insights (financial health, services, risks, opportunities, outlook).
- Analyst Agent → Transforms insights into a polished one-page financial analyst brief with executive summary, analysis, and conclusions.
The system also uses two tasks
- Research Task → Conducts structured research on {company}, gathering insights on financial health, historical performance, key services, risks, opportunities, and future outlook. Produces an organized research brief with clear sections.
- Analysis Task → Reviews the research brief and creates a polished one-page Financial Analyst Report. Includes an executive summary, financial highlights, service strengths, market opportunities, and a forward-looking conclusion.
- CrewAI → multi-agent orchestration
- Python → core implementation
- LiteLLM / OpenAI GPT-4o-mini → natural language generation
- Serper API → live web search for up-to-date financial data
- Dotenv → secure environment variable management
This project can run in two modes:
When the crew runs without internet access, the Researcher Agent relies only on the model’s built-in knowledge.
The generated report still follows the correct structure but is based on general knowledge up to the model’s cutoff.
Example Screenshots:
When SERPER_API_KEY is provided, the Researcher Agent uses live search to pull the most recent company information.
This makes the final financial analyst brief more accurate, up-to-date, and relevant for 2025.
Execution Flow:
- Research agent queries live data sources
- Findings are structured into insights
- Analyst agent turns insights into a professional one-page report
Example Screenshots:
- Offline Mode → General analyst report based on stored knowledge.
- Online Mode → Live data integration, producing reports with current financials, news, and trends (2025).
The details in the report may not be accurate. I am trying to improve the program so that it can be updated with accurate information.