January 15, 2025
Case Study
16× Faster Utility Invoices—$22 K / mo Waste Recovered
Chesmar Homes’ Houston and DFW divisions were losing time—and money—manually keying thousands of utility invoices each month. Actify Automations delivered a custom AI-assisted automation system application that now processes the same invoices 16× faster at 99% accuracy, while flagging bills for homes that closed years ago. The first 30-day pilot freed 60+ staff hours. Initial usage during the testing & pilot period prevented over $22,000 in avoidable repeating charges—results new divisions can replicate in 14 days thanks to pre-trained models.
The Challenge
Manual bottleneck AP clerks spent 1.5 min per invoice, re-keying data into spreadsheets before BRIX upload.
Hidden leakage Utility vendors continued billing closed properties; missed flags compounded month-over-month.
Staff overload Teams were “always swamped,” leaving no bandwidth for process improvement.
The Actify Solution
Layer | What It Does | Why It Matters |
|---|---|---|
OCR + model tuning | Reads any PDF/image invoice | Expands coverage without vendor templates |
Business-rule engine | Cross-references BRIX vendor sheet; flags closed lots | Handles the bulk of the work automatically |
Custom web UI | Drag-and-drop batch upload, live status | Zero learning curve for AP staff |
Excel export & API hooks | Generates ready-to-post spreadsheets for import | If your JD Edwards version exposes REST (APIS), the system can push vouchers directly—no manual import |
Pilot Scope (first 30 days)
Division-specific invoice format AI model training
Live processing of all June invoices
Usage based progress & weekly Updates
Zero upfront fees; full data ownership
(New divisions require only 14 days—training, not rebuild.)
This past 30 Days
Metric | Manual | Automation System | Impact |
|---|---|---|---|
Invoices processed | ~2500 | 2,563 | — |
Processing speed | 1.5 min each | 0.09 min each | 16× faster |
Processing hours | 63.0 hrs | 3.8 hrs | -59.2 hrs |
Labor value (@ $30/hr) | — | $1 900 reallocated | |
Accuracy | ~95% | 99 % (≈ 26 touch-ups) | 80% reduction |
Invoices caught for listings closed over 60 days (this month) | Untracked | 153 invoices | Over $10,000 in potential savings |
Invoices caught for listings closed over 60 days (4/22-5/11) | Untracked | 257 invoices | Over $11,000 in potential savings |
Compounding Savings
Closed-listing misses snowball: when a single utility is left active, the same invoice re-appears. During initial model training we surfaced $11,000 in legacy charges; the first production month uncovered $10,000 more.
In total, the system has flagged over 300 invoices worth ≈ over $22,000 across four divisions. Each dollar saved this month would have re-billed, so catching the waste early prevents it from compounding and permanently protects builder margins—even as manual review workload drops to near-zero.