What Happens When You Submit an AI Inspection? A Step-by-Step Walkthrough
Upload home photos and receive a full findings report with severity ratings, repair cost estimates, UAD condition mapping, and auto-routed risk factors — in under 90 seconds.
Published 2025-02-28 · Updated 2026-05-15 · 5 min read
What is an AI inspection?
An AI inspection takes a set of photos of any part of your home (HVAC, roof, plumbing fixture, foundation, etc.) and runs them through a multimodal vision model. The model identifies visible issues, classifies their severity, and produces an itemized repair estimate. It is not a substitute for a licensed inspector for legal disclosures — it is an early-warning layer between professional visits.
How long does an AI inspection take?
End-to-end: 30–90 seconds for 4–8 photos. The vision call itself runs 8–20 seconds; the rest is image compression, validation, and database write.
What does the output look like?
- Findings list — each tagged with category, severity (ok / monitor / schedule / urgent), and a short description
- Cost estimate — low / high range based on local labor and materials
- UAD condition rating — overall C1–C6 grade aligned with Fannie Mae appraisal vocabulary
- Risk auto-routing — urgent items can be one-tap dispatched to a vetted technician
How accurate is the cost estimate?
Cost ranges are calibrated against ProActive's vendor-bill database for Central Texas (Bexar, Travis, Comal, Hays, Williamson counties). Median accuracy in 2024 was ±18% of the final invoiced amount — comparable to a contractor's verbal estimate but available in 90 seconds instead of 3 days.
Does the AI replace a licensed inspector?
No. For mortgage, insurance, or legal disclosure purposes you still need a licensed Texas Real Estate Commission (TREC) inspector. The AI inspection is an interstitial layer — useful between professional visits, before a contractor call, or to triage what to fix first.
Are my photos private?
Photos are stored against your property record, are accessible only to your account, and are not used to train external models. Server-side compression reduces them to ~150–300 KB for long-term storage; the original full-resolution copy is held only during the AI analysis call.
Frequently asked questions
How many photos can I submit at once?
Up to 12 photos per inspection submission. Larger photo sets are batched automatically.
Which AI model is used?
GPT-4o vision (OpenAI). Some workflows use Claude Sonnet for cross-validation on high-severity findings.
Can the AI catch foundation issues?
It can surface visible indicators (stair-step brick cracks, sloped floors visible in photo, drywall separation) and flag them — but foundation diagnosis requires a structural engineer.