When a public works department pays a contractor to repair a broken HVAC unit, a leaking roof, or a failed lift pump, the invoice usually clears before anyone confirms the work was completed to spec — and in most municipalities a signed paper form is the only artifact in the chain of evidence. This case study follows Riverside County Facilities Authority, a public-sector maintenance organization managing 214 buildings, 6,100 tracked assets, and a $9.4M annual contractor repair budget, as it deployed OxMaint AI vision inspection and CMMS work order automation to gate invoice approval on machine-verified visual proof. Over 12 months the program recovered $740K in overpayments, cut disputed invoices by 91%, and produced a tamper-evident audit record for every repair dollar. To see the verified-closeout workflow on your own assets, book a demo with our team.
Stop Paying for Repairs You Can't Verify
See how OxMaint pairs AI vision with your work orders to confirm contractor completion using geotagged photo evidence and model-based defect comparison — before the invoice is approved.
The Verification Gap in Public-Sector Maintenance
Public agencies outsource the bulk of facility repairs, yet the closeout step is the weakest control in the chain. A paper sign-off records a signature, not the post-repair condition of the asset, and captures roughly 0.3 structured data points per event — far too thin to support trend analysis, contractor scoring, or dispute resolution. By the time a defect resurfaces, the contractor is paid, the crew has demobilized, and no visual baseline exists to settle the claim.
Verification Pipeline: How AI Vision Closes the Loop
OxMaint does not replace inspectors — it produces a deterministic evidence chain. A repair advances through five gated states, and the work order remains in an open state until the vision model returns a confirmed defect-resolution match. Each transition is timestamped (server-side, UTC), GPS-stamped against the asset's geofence, and write-locked to the asset record.
The control shift is structural: payment is gated on a verifiable artifact, not a signature. Try the verified-closeout pipeline free on a single building first.
Turn Every Repair Into Audit-Ready Proof
Give your facility team geotagged, timestamped, model-verified evidence for every contractor job — and a defensible, tamper-evident record for every audit cycle.
Model & Deployment Profile
Riverside County ran on standard smartphone capture — no fixed cameras or edge hardware. Accuracy was tracked against a quarterly ground-truth audit and improved as the false-positive queue fed retraining.
| Parameter | Configuration | Operating Value |
|---|---|---|
| Capture device | Mobile app, 1080p min | No fixed hardware |
| Geofence tolerance | Per-asset radius | 30 m default |
| Auto-route threshold | Classification confidence | ≥ 0.85 |
| Resolution-match accuracy | vs ground-truth audit | 88% → 94% by Q4 |
| False positives flagged | Routed to inspector | 6% of submissions |
| Evidence retention | Immutable asset log | Full FOIA chain |
| Time-to-first-value | From go-live | 6 weeks |
Results After 12 Months
Before vs After Verification Program
| Metric | Paper Closeout | OxMaint AI Vision | Impact |
|---|---|---|---|
| Repairs with geotagged proof | 11% | 100% | Full coverage |
| Disputed contractor invoices | 340 / yr | 31 / yr | -91% |
| Invoice approval cycle | 11 days | 2 days | -82% |
| Repeat callbacks (same defect) | 29% | 7% | -76% |
| Audit-package prep time | 3 weeks | Same day | Instant pull |
| Structured data per closeout | 0.3 | 8.4 | 28x richer |
| Recovered overpayments | $0 | $740K | New recovery |
Where the $740K Came From
Recovery did not come from auditing contractors harder — it came from making completion machine-verifiable at the point of closeout. Once a confirmed match became the precondition for payment, three leakage categories closed within the first two quarters.
Each recovery category maps directly to a flag the vision pipeline raises automatically. Book a demo to map these controls onto your own contractor workflow.
Expert Review
"The verification gap in public maintenance was never a trust problem — it was an evidence problem. Agencies paid on a signature because that was the only artifact the workflow produced. The moment a before/after pair is bound to the asset with GPS, a server timestamp, and a model confidence score, the dispute economy collapses. Recovery follows automatically because the contractor and the agency are finally adjudicating against the same immutable record instead of two conflicting recollections."
Phased Rollout: Pilot to Full Coverage
Deployment was staged to validate the vision model against live contractor data before extending payment gating across all 214 buildings. Each phase added scope only after the prior phase met its match-accuracy and false-positive targets.
The phased model let the agency prove the controls on low-risk assets before tying payment to them. Book a demo to scope a pilot on your own building portfolio.
Verify Every Public Dollar You Spend on Repairs
Join public-sector teams using OxMaint to gate contractor payment on AI-verified completion — and recover the spend that paper sign-offs let slip through.
