Equity Research Assistant

The system covers 11 GICS sectors split across a two-person research team. My coverage includes Information Technology, Healthcare, Consumer Discretionary, Consumer Staples, and Communication Services. Each sector has its own dedicated research agent prompted with sector-specific frameworks, data sourcing instructions, and output formatting requirements. Rather than a one-size-fits-all model, each agent is designed around the characteristics of its sector: what drives valuation, what risks matter most, and which metrics institutional analysts actually use when screening names.

The agents are built using the Claude Agent SDK and operated through Cowork, a desktop orchestration platform that allows agents to read and write files, run code, browse the web, and interact with connected tools within a managed, auditable workflow. This architecture lets each agent pull live market data, process it, apply a valuation framework, and produce a structured research output without requiring manual data entry at each step.

Each agent produces a structured data table for the most relevant Canadian-listed equities in its sector, supplemented by select U.S. names where the Canadian universe is too thin to be representative. All prices are denominated in CAD, with USD-listed names converted at the prevailing exchange rate.

The core data table captures the metrics most relevant to a wealth management context: current price, P/E ratio, forward P/E, EV/EBITDA, dividend yield, payout ratio, free cash flow yield, net debt to EBITDA, Morningstar fair value estimate, and analyst consensus price targets. These factors give advisors a complete picture of both valuation and financial health, distinguishing between companies that look cheap on earnings multiples but carry excessive leverage versus those trading at a premium but generating durable free cash flow.

Beyond the standardized metrics, each agent applies a sector-specific alternative valuation lens. For Information Technology, this is the Rule of 40 — a growth-plus-margin composite widely used to evaluate software and SaaS businesses. For Communication Services, the primary lens is EV/EBITDA, reflecting the capital-intensive, subscription-driven nature of telecom. Consumer Discretionary agents emphasize same-store sales growth as the key fundamental indicator of retail health. These sector-specific factors are embedded directly into each agent's analytical framework, ensuring the research reflects how practitioners in each sector actually think about value.

Each research output includes a Notes column using a consistent color-coding system across all sectors: green for positive attributes or investment thesis support, red for risks or names to avoid, and blue for key contextual or framework points. This standardization means an advisor can move between sector reports and immediately orient themselves without re-learning the format.

Each report concludes with a tiered ranking of every name by suitability for different client types, distinguishing between names appropriate for conservative portfolios prioritizing income and capital preservation, and those better suited to growth-oriented investors with longer time horizons and higher risk tolerance.

The research produced by these agents feeds directly into a standardized model portfolio construction process. The target output is 30 distinct portfolio models spanning five account types, two risk profiles, and three portfolio sizes. Each model portfolio is grounded in the scored sector research, with individual stocks selected based on the agent-generated rankings and mutual funds layered in where direct equity exposure is impractical or inappropriate for the client profile.

This project sits at the intersection of financial analysis, AI engineering, and wealth management operations, and represents a meaningful step toward making institutional-quality research infrastructure accessible at the advisory level.