Models assist with structure. People decide what matters.

AI accelerates the mechanical work of triage, extraction and drafting so reviewers spend their time verifying substance instead of typing. It never independently issues a conclusion that requires a license, and every output it produces retains its source provenance and review status.

What AI may assist with

Tasks where model assistance speeds up structuring evidence.

Every item below produces a draft or candidate for a human reviewer — none of them is released to a client without verification.

Lead triage
Scope drafting
Document request generation
Metadata extraction
Document indexing
OCR drafting
Image quality screening
Asset suggestions
Nameplate extraction
Quantity candidates
Duplicate detection
Evidence completeness checks
Reconciliation checks
Report drafting
Workflow monitoring
Annual recapture reminders

Draft, always

An AI-generated report section, metadata field or asset suggestion is stored and displayed as a draft or candidate until a qualified reviewer confirms it. Confirmation changes its status in the record; it does not disappear into a final document unmarked.

What AI may not independently approve

Conclusions that require a licensed, qualified human — always.

These determinations sit outside what any model, however capable, is permitted to finalize on our platform. They require the applicable licensed or qualified professional.

Requires a qualified professional — not AI-approved

Legal boundaries
Structural safety
Final tax classifications
Licensed engineering conclusions
Official valuations
Thermal diagnoses
Coverage determinations
Legal opinions
Safety-critical flight decisions
Final issue approval

Human review gates

Where a person confirms before anything moves forward.

Gate 1 — Extraction review

After AI drafts metadata, quantities or asset candidates from a capture, a reviewer checks the draft against the source imagery or document before it enters the structured record.

Gate 2 — Finding review

Candidate findings surfaced by reconciliation or completeness checks are confirmed, corrected or rejected by a qualified reviewer before they inform a deliverable.

Gate 3 — Report review

AI-drafted report language is checked for accuracy, boundary language and traceability to evidence before a deliverable is released.

Gate 4 — Final issue approval

A qualified person signs off before anything leaves the workspace as a delivered artifact. This gate is never automated.

Provenance travels with every AI output

Whether a metadata field, a report paragraph or a quantity candidate, every AI-generated element in the record carries a marker showing it was model-assisted, along with the source capture it was derived from and its current review status. Nothing produced by a model is presented as human-confirmed until it actually is.

FAQ

The AI operating model, answered.

Next step

Start with the decision you need to make.

Tell us the property and the decision. You get a written scope with an exact price before anything is scheduled.

Get my property plan See how evidence is tracked