This solution builds a complete AI-powered asset management platform inside Snowflake, featuring specialised Cortex Agents that serve different roles across front, middle, and back office functions. It demonstrates how financial firms can transform investment workflows by integrating structured portfolio data with unstructured research documents — all powered by Snowflake CoWork.

The solution creates a fictional firm called Simulated Asset Management (SAM) with:

The solution runs entirely within Snowflake using native AI features — no external infrastructure, APIs, or data movement required.

Data Foundation: A dimensional model with 48+ tables — dimensions (securities, issuers, portfolios, benchmarks, clients) and facts (positions, transactions, prices, ESG scores, attribution). Real securities sourced from 14,000+ instruments via Snowflake Public Data.

Document Corpus: Synthetic documents (broker research, press releases, NGO reports, policy documents, regulatory texts) generated via a template hydration engine. Real SEC filings (10-K, 10-Q) and earnings transcripts loaded from the marketplace share.

Cortex Search Services: 16 search services indexing documents by type with filterable attributes (ticker, company, document type, jurisdiction, severity). Enables semantic search across thousands of documents.

Cortex Analyst Semantic Views: 10+ semantic views defining relationships and metrics across portfolio, market, attribution, research, and operational data. Translates natural language into SQL for instant analytical answers.

Cortex Agents: 8 agents with role-specific instructions, tool access, and orchestration skills. Each agent intelligently combines Cortex Analyst, Cortex Search, and stored procedure tools to answer complex multi-source questions.

Snowflake CoWork: The conversational interface where users interact with agents via natural language. Agents orchestrate tools, produce visualisations, and generate PDF reports — all within a chat experience.

ML Workflows: 3 notebook-based ML demonstrations using Snowflake Feature Store, Model Registry, and Cortex ML functions for market regime detection, factor model training, and credit risk scoring.

Skill-Driven Workflows: 36 runtime skills provide structured multi-step guidance for complex tasks (investment memos, RFP responses, client letters, attribution reports). Skills include verification checkpoints and cross-references.

Multi-Tool Orchestration: A single question can trigger 5+ tools — verify a market event, calculate portfolio exposure, check policy thresholds, search research documents, and generate a committee memo.

Custom Analytical Tools: Agents have access to purpose-built stored procedures that execute complex quantitative workflows:

Portfolio Construction and Backtesting: The Portfolio Copilot can construct model portfolios, optimise allocations (max Sharpe, min variance, risk parity), run full historical backtests with custom weights, and validate with Monte Carlo simulation — enabling a complete "build, test, validate" workflow in conversation.

Real Data Integration: SEC financials, stock prices, institutional holdings (13F), insider trading, dividends, and economic indicators — all from Snowflake Public Data (Free).

Performance Attribution: Full Brinson-Fachler decomposition (sector, country, industry), factor attribution, currency effects, and linked multi-period analysis.

Proactive Insights: AI-generated morning briefings and signal extraction with urgency scoring — surfaced to agents for context-aware responses.

1. Navigate to Projects > Workspaces
2. Click + then From Git repository
3. Repository URL: https://github.com/Snowflake-Labs/sfguide-agentic-ai-for-asset-management.git
4. Authentication: Public repository (no auth needed)
5. Name the workspace (e.g., "SAM Demo")

Open scripts/setup.sql in the workspace and execute it. This creates:

1. Open python/workspace_main.py in the workspace
2. Connect a notebook service when prompted:
   • Python version: 3.11+
   • Compute pool: any available pool
   • Artifact repositories (optional): SNOWFLAKE.SNOWPARK.PYPI_SHARED_REPOSITORY
3. Open the Terminal and run: pip install -r "$PWD/requirements.txt"
4. Restart the kernel
5. Click Run

The build creates all tables, documents, search services, semantic views, agents, skills, tools, signals, and evaluation datasets.

Faster Insights: Multi-dimensional risk assessments completed in minutes instead of days. Event-driven analysis that previously required multiple analysts happens in a single conversation.

Reduced Manual Effort: Eliminates time spent searching across systems, compiling reports, and waiting for data teams. Every question gets an immediate, comprehensive answer.

Better Risk Management: Real-time monitoring of concentration limits, ESG controversies, and mandate breaches with automated policy checking.

Enhanced Client Service: Instant access to performance data, philosophy documents, and regulatory disclosures for quarterly presentations, RFP responses, and client inquiries.

Informed Decisions: Executives get firm-wide visibility with the ability to run M&A simulations, stress tests, and scenario analyses on demand.

GitHub Repository →

The repository contains setup scripts, sample data, semantic model definitions, agent skills, ML notebooks, and step-by-step instructions for deploying the complete solution.