Powdered Drink City — Financial Model AI Agent

Powdered Drink City is a direct-to-consumer ecommerce brand selling powdered drink mixes across Shopify, Amazon, and wholesale channels. Like most emerging CPG brands, the business relies on a complex Google Sheets financial model to plan ad spend, forecast revenue, manage cash flow, and optimize for profitability.

The goal of this project is to build an AI-powered multi-agent system that can deeply analyze these financial models using natural language. Instead of manually tracing formulas across dozens of spreadsheet tabs, an FP&A analyst can ask the agent questions like “What happens to my EBITDA if I raise prices by $5 in Q3?” or “Find me three scenarios to hit 10% gross sales growth while keeping cash above $1M.”

Multi-Agent Architecture

The system uses a supervisor/planner agent that routes user requests to seven specialist sub-agents, each with a dedicated playbook:

Each agent has access to specific tools for reading and writing to Google Sheets, tracing formula chains, performing sensitivity analyses, and retrieving strategic business knowledge from a RAG-powered knowledge base.

Key Capabilities

• Read and understand complex multi-tab financial models with 170+ rows
• Trace formula chains to explain how any Strategic Outcome is calculated
• Run goal-seek optimization using Latin hypercube sampling
• Perform 1- and 2-variable sensitivity analyses
• Provide strategic guidance via RAG over business knowledge base
• Search current trends using Tavily web search
• Maintain conversation context using 5 memory types
• Log all actions to an audit trail in the spreadsheet