Skip to content

Data Intelligence

This document specifies the requirements for CalcBridge's data intelligence capabilities, including anomaly detection, insights, and AI-powered analysis.


Overview

CalcBridge's intelligence layer provides AI-powered analysis of portfolio data, surfacing patterns, anomalies, and actionable recommendations.


Requirements

Functional Requirements

ID Requirement Priority Status
DI-001 System shall generate AI-powered insights from nominated workbooks Must Have Implemented
DI-002 Insights shall be tailored by stakeholder type Must Have Implemented
DI-003 Users shall ask natural language questions about data Must Have Implemented
DI-004 PII shall be scrubbed before AI processing Must Have Implemented
DI-005 Session management with regeneration support Should Have Implemented
DI-006 Predictive compliance warnings Could Have Planned
DI-007 Trade recommendations Could Have Planned

Insight Categories

Category Description
Financial Health Portfolio metrics, exposure trends, concentration analysis
Operational Efficiency Process optimization, resource allocation insights
Risk & Compliance Risk exposures, compliance warnings, threshold proximity
Trends & Forecasts Historical patterns, forward-looking projections
Recommendations Actionable improvement suggestions

Stakeholder Views

Stakeholder Focus Key Metrics
Executive Strategic overview KPIs, trends, board-level metrics
Finance Financial detail Variances, P&L, accounting
Operations Efficiency Throughput, resource allocation
Analyst Statistical depth Patterns, distributions, correlations

AI Chat

  • Natural language Q&A about portfolio data
  • Context-aware responses based on workbook structure
  • PII scrubbing on all inputs and outputs
  • Response sanitization to prevent data leakage

User Stories

US-DI-001: Generate Portfolio Insights

As an Executive I want to see AI-generated insights about my portfolio So that I can quickly understand portfolio health and trends

US-DI-002: Ask Data Questions

As a CLO Analyst I want to ask questions about my data in plain English So that I can get answers without writing formulas