Snowflake-Labs/sfguide-agentic-ai-for-asset-management

Snowflake-Labs/sfguide-agentic-ai-for-asset-management

Language: Python

License: Apache-2.0

Stars: 11

Forks: 11

Open issues: 0

Created: 2025-12-04T17:56:40Z

Pushed: 2026-04-08T22:31:36Z

Default branch: main

Fork: no

Archived: no

README:

Agentic AI for Asset Management Using Snowflake Public Data

Overview

Simulated Asset Management (SAM) is a fictional multi-asset investment firm managing multiple strategies. This guide showcases how Snowflake Intelligence transforms investment management through Cortex Agents that seamlessly orchestrate:

• Structured Data: Real securities from SEC filings, portfolio holdings, factor exposures, ESG scores
• Unstructured Documents: Broker research reports, earnings transcripts, press releases, policy documents
• Real-Time Analytics: Cortex Analyst converts natural language to SQL
• Document Intelligence: Cortex Search enables RAG-powered document synthesis

What Makes This Guide Unique

| Capability | Description | |------------|-------------| | Real Securities | All securities are authentic from SEC Filings (OpenFIGI) | | Multi-Tool Orchestration | Agents dynamically combine 3-14 tools per query | | Industry-Standard Model | Dimension/fact architecture with audit trails | | Complete Workflow Coverage | Agents covering front, middle, back office |

---

What You'll Learn

By exploring this guide, you'll understand how to:

1. Build Multi-Tool Cortex Agents - Combine Cortex Analyst (structured data) with Cortex Search (documents) in a single conversational interface

2. Design Semantic Views - Create business-friendly data models that translate natural language to SQL automatically

3. Implement RAG at Scale - Index thousands of documents across multiple corpus types with Cortex Search

4. Create Industry-Specific Agents - Configure agents with role-specific instructions, tool access, and business context

5. Generate Realistic Demo Data - Use template-based document generation with placeholder hydration for authentic content

6. Integrate Real Data Sources - Leverage Snowflake Marketplace for SEC filings, stock prices, and financial data

---

Repository Structure

sfguide-agentic-ai-for-asset-management/
├── scripts/
│ ├── setup.sql # Automated setup (~15-20 min)
│ └── teardown.sql # Complete cleanup script
├── python/ # Python modules
│ ├── config.py # Central configuration
│ ├── generate_structured.py # Dimension/fact table generation
│ ├── generate_unstructured.py # Document generation
│ ├── generate_market_data.py # Real market data
│ ├── create_agents.py # Agent definitions
│ ├── create_semantic_views.py # Semantic view definitions
│ ├── create_cortex_search.py # Search service definitions
│ └── ...
├── content_library/ # Document templates
│ ├── _rules/ # YAML config
│ ├── security/ # Broker research, press releases
│ ├── issuer/ # NGO reports, engagement notes
│ ├── portfolio/ # IPS, portfolio reviews
│ └── global/ # Policy docs, procedures
└── docs/ # Demo scenario documentation

---

Getting Started

Prerequisites

• Snowflake account with Cortex features enabled
• ACCOUNTADMIN role (for setup)
• Snowflake Intelligence available

Step 1: Run Setup Script

Copy and paste [scripts/setup.sql](scripts/setup.sql) into a Snowflake SQL Worksheet and run it.

The setup script automatically:

• Installs Snowflake Public Data Free from the Marketplace
• Creates CURATED dimension/fact tables and views
• Creates RAW corpus tables from document templates
• Creates MARKET_DATA tables with real SEC filings
• Creates Cortex Search services for document retrieval
• Creates Cortex Analyst semantic views for structured queries
• Creates Snowflake Intelligence agents for each business role

Step 2: Access Snowflake Intelligence

1. Switch to role SAM_DEMO_ROLE and warehouse SAM_DEMO_WH 2. Navigate to AI & ML → Snowflake Intelligence in the left sidebar 3. Choose an agent → Start asking questions!

---

Cortex Agents

| Agent | Role | Key Capabilities | |-------|------|-----------------| | Portfolio Copilot | Portfolio Manager | Holdings, risk, event impact, supply chain, SEC financials | | Research Copilot | Research Analyst | Multi-source research, earnings, investment memos | | Thematic Macro Advisor | Thematic PM | Theme positioning, macro events, allocation | | Quant Analyst | Quantitative Analyst | Factor screening, attribution, validation | | Sales Advisor | Client Relations | Client reporting, RFP, onboarding | | ESG Guardian | ESG Officer | NGO reports, controversies, engagement | | Compliance Advisor | Compliance Officer | Mandate monitoring, breach detection | | Middle Office Copilot | Operations Manager | Settlement, reconciliation, NAV, corporate actions | | Executive Command Center | C-Suite | Firm KPIs, competitor intel, M&A simulation |

---

Multi-Tool Demo Prompts

These signature prompts showcase each agent's full orchestration capabilities. For complete step-by-step scenarios with talking points and expected responses, see the linked documentation.

---

Portfolio Copilot

> Anna, Senior PM | [Full Scenarios →](docs/demo_scenarios_portfolio_manager.md)

Event-Driven Risk Assessment:

I just heard about a major earthquake in Taiwan affecting semiconductor production. Can you:
1. Verify this event and identify affected sectors
2. Show my direct exposure to Taiwan-based technology holdings across all portfolios
3. Calculate indirect exposure through supply chain dependencies (especially for companies like Apple, NVIDIA)
4. Check if any positions breach our concentration limits
5. Analyze the financial health of my most exposed companies using SEC data
6. Find what management is saying in recent earnings calls about supply chain resilience
7. See what analysts are recommending about semiconductor exposure
8. Review our firm's policy on geographic concentration and event response
9. Generate an Investment Committee memo documenting this risk assessment with specific portfolio actions and timeline

---

Research Copilot

> David, Research Analyst | [Full Scenarios →](docs/demo_scenarios_research_analyst.md)

Investment Memo Generation:

Generate a comprehensive investment research report for NVIDIA covering financial health, management outlook, analyst views, competitive position,…