Four passes between your photos and an estimate.
Mensuria is a pipeline, not a chat box. Vision reads the loss, the model drafts against a verified code base, deterministic checks demote anything it can't prove, and a senior-adjuster review hands you a clean draft to approve. Here is exactly what happens at each stage.
Vision reads the photos.
Every photo from the walkthrough is analyzed for three things at once: the materials in frame, the damage and its category, and any measurements it can read off the scene. Drywall versus plaster, a 24-inch water line, a 36-foot wall run — each observation is recorded with a confidence level before a single line item exists.
- Materials. Surfaces, finishes, and assemblies — down to baseboard profile and drywall thickness.
- Damage. Type, category, and extent — Cat-3 water, char depth, hail bruising, wind uplift.
- Measurements. Affected runs and areas read from the scene, kept as evidence for later quantities.
Drafts line items against a verified code base.
Now the model writes Xactimate-ready lines — but it can only choose from a verified selector base, not from open-ended invention. Every quantity is tied back to a specific vision observation, so an 80 SF extraction line traces to the Cat-3 water finding, and a 36 LF flood cut traces to the wall run it measured. Codes it isn't sure about are surfaced for confirmation — never guessed.
- Verified codes only. Selectors are drawn from a known-good base — the model can't mint a code that doesn't exist.
- Grounded quantities. Each amount points back to the photo and observation that justify it.
- Room-grouped output. Lines are organized the way you'd key them — by room, by trade, in order.
Deterministic grounding QA.
This pass isn't another model guessing — it's deterministic rulesthat check each drafted line against the photo evidence behind it. If a line has solid photo support, it stays high-confidence. If the support is thin, it's demoted. If there's no grounding at all, it's flagged as an open question instead of riding along silently. Nothing ungrounded ever gets a green light.
Senior-adjuster review, then your approval gate.
A second AI pass reads the draft the way a senior adjuster would — hunting for missing items, wrong units, and over-scoping. It leaves notes, not edits: add the dryer detach & reset, confirm the subfloor extent, re-check the air-mover days. Then it's yours. You open the scope editor, see every photo behind every line, resolve the flags, and only then does Approve & export unlock.
Xactimate worksheet
A clean keying worksheet — codes, quantities, descriptions, grouped by room — ready to paste or key straight in.
CSV export
The full scope as structured data for your own spreadsheet, template, or downstream tooling.
Print / PDF
A formatted, shareable copy of the approved scope — your name on the estimate, every line.
What you control.
The pipeline does the keying. You keep the judgment. Mensuria is built so the adjuster is always the last word — never the AI.
Every quantity is editable
Click any line to change the code, unit, or amount. Your edits become the source of truth — the AI never overwrites you.
Resolve or acknowledge flags
Each low-confidence line and open question is yours to clear. Fix it, or acknowledge it on the record — nothing slips by silently.
Approval is a hard gate
Export stays locked until every flag is resolved or acknowledged. There is no path to a finished estimate that skips your review.
You are the adjuster of record
Mensuria drafts and reviews; you sign off. Your name goes on the estimate, on every line, every time.
Run a claim through it.
Bring your own loss. Watch the four passes turn your photos into a draft you'd sign.