AI Estimate Validation Agent

How Cassidy automates estimate validation using AI

Step 1: Ingest estimate and cost history

The Workflow triggers when a new estimate is uploaded or synced from your estimating platform (Trimble, QuoteSoft, FastEST). Cassidy pulls the estimate data alongside historical job-cost actuals from your ERP, normalizing everything to a unified cost-code structure.

Step 2: Normalize and enrich line items

Cassidy maps each line to your WBS and CSI MasterFormat codes, standardizes units (LF, LB, EA), and tags attributes like system type, material, size, join method, prefab vs. field, and install conditions.

Step 3: Build and apply benchmarks

The Agent computes unit rates from your current estimate and compares them against historical P25/P50/P75 bands, MCAA WebLEM labor units, and RSMeans references—adjusting for project type, location factors, and site-specific conditions.

Step 4: Detect outliers and variances

Using statistical thresholds and peer-group matching, Cassidy flags lines where MH/unit or $/unit falls outside expected ranges, providing variance magnitude and comparable historical data for each flagged item.

Step 5: Run scope completeness checks

Automated QA rules verify that standard mechanical scope elements are present—valves per LF, hangers proportional to pipe runs, seismic bracing, testing allowances, rigging, and commissioning—and flags likely omissions.

Step 6: Generate validation report

Cassidy compiles a validation package with flagged outliers, suggested adjustments, scope gap findings, and a risk register. The report routes to estimators for review, with approval workflows that log decisions and update the estimate via API.