PayKeeper’s verification system is designed to support—not replace—sound fund control decisions. The system does not release funds based solely on a single photograph, metadata element, or document, nor does it operate under the assumption that automated analysis is infallible. Critical inspections involving structural integrity, safety compliance, electrical work, or mechanical systems continue to require manual verification by qualified professionals.

What the system does provide is a structured framework for gathering and evaluating multiple independent signals. By combining photographic evidence, metadata analysis, supporting documentation, and historical project data, PayKeeper builds a comprehensive risk profile for each draw request. When submissions present uncertainty or raise concerns, they are routed to manual review, and payment is withheld until discrepancies are resolved. Importantly, the system enforces contractual escrow logic throughout: funds are released only when all predetermined conditions have been satisfied, regardless of any individual verification score. A complete audit trail with detailed logging ensures transparency and supports human oversight at every stage of the process.

For each draw milestone, contractors typically submit 8-15 photos showing: overall property progress, close-ups of completed work areas, installed material labels/packaging, and context shots proving location. Our mobile app provides photo checklists specific to each trade and milestone.

GPS metadata accuracy is typically within 15-30 feet for modern smartphones. For properties where this creates ambiguity, our Street View matching provides secondary confirmation. 

For installed materials where packaging isn’t visible, contractors can submit invoice documentation alongside photos. Our AI performs OCR on invoices and cross-references model numbers against photos showing installed products. For critical items like HVAC units, we require photos of the serial number plate.

Every AI verification includes a confidence score. Items flagged with moderate confidence go to our expert review team for human validation within 4 business hours. Contractors can also contest flags through the portal by submitting additional evidence. False positive rates are under 2% and declining as our models improve.

No. AI verification is ideal for routine draws on straightforward projects. We still recommend traditional inspections for: foundation work, framing inspections, mechanical/electrical rough-ins before drywall, and final certificate of occupancy walkthroughs. AI handles 70-80% of mid-project draws, eliminating delays where inspection adds the least value.

Yes. Submitted photos and documents are encrypted at rest and in transit. Data is only shared with the parties to the specific construction contract (borrower, lender, contractor). We don’t sell or monetize contractor data, and we comply with all state privacy regulations.

Absolutely. For interior work, baseline photos from appraisal or pre-construction walkthrough become the comparison point. Our AI tracks progression of interior finishes across draws and focuses heavily on material spec verification and quantity estimation.

Yes. Our AI scales from residential renovations to multi-million dollar commercial projects. Commercial projects often benefit even more due to higher draw frequencies and the cost of delays. We support all construction types: ground-up, renovation, tenant improvement, and infrastructure.