Before you run your first Preflight.

It is a pre-delivery review engine for completed residential appraisal packages. It compares structured report data, rendered narrative, and package evidence to surface contradictions, missing support, and review triggers before the report leaves your desk.

After you have completed the report in your appraisal software and before client delivery. Export the package, run Preflight, review the prioritized queue, record your decisions, then make any warranted changes in your appraisal software.

No. CoAppraiser works alongside the software you already use. Your appraisal software remains the place where you develop, edit, sign, and deliver the report; Preflight is the independent evidence review between completion and delivery.

Upload the completed package you intend to deliver as one exported ZIP, or select the UAD XML, rendered report PDF, and available photos or exhibits together. A complete package gives Preflight more sources to cross-check and makes missing-evidence findings more meaningful.

CoAppraiser uses three separate gates. The UAD normalizer must extract supported fields from all 12 paired PDF/XML scenarios in Fannie Mae Appendix D-1. Controlled mutations must produce exactly the expected condition, quality, and above-grade area findings. Paid agent evaluations separately score expected topics, source citations, professional boundaries, latency, and token use. These are reproducible engineering tests, not official UAD certification.

Examples include condition or quality conflicts between XML and narrative, values that differ between the structured export and rendered PDF, missing core files, incomplete comparable commentary, and visible photo evidence that may not align with the rendered report narrative.

Traditional validation is strong at required fields and valid formats. Preflight asks a different question: does the package tell one consistent, supportable story across the grid, addendum, rendered report, and available exhibits?

The multimodal Preflight agent cross-checks structured XML, the rendered report, narrative evidence, and selected appraisal photos for relationships fixed rules alone may miss. It produces structured findings with an observed issue, source location, supporting evidence, confidence, why the issue matters, and a recommended review action. Application code then verifies visual filenames, suppresses low-confidence or incomplete output, and keeps deterministic package checks separately identified.

Deterministic findings come from repeatable application rules, such as a missing XML or a direct cross-source mismatch. Preflight agent findings interpret relationships across extracted and visual evidence. The interface separates the two so you know how each item was produced.

No. It does not determine value, select or reject comparables, recommend a final adjustment, or make appraisal conclusions. It identifies evidence for you to verify; you decide what the evidence means.

No. Preflight does not declare USPAP compliance and is not a substitute for your own standards analysis, scope-of-work decisions, certification, or professional obligations.

No. It is readiness support, not an official UAD validator or an acceptance guarantee. Client overlays, automated underwriting systems, and reviewer judgment can still produce requests after delivery.

Open the finding, inspect the observed issue and cited evidence, then verify the source in your appraisal software. If a change is warranted, make it there. Preflight organizes the review; it does not silently alter your report.

Yes. You can mark an item resolved, deferred, or not applicable and add a short decision note. Keeping an item open is also valid while you investigate. Your selected status and note become part of the review history.

Yes. A revised upload creates a new immutable version inside the same review. The earlier package, findings, and decisions remain in the version history so the work can be traced.

The downloadable record preserves package and version metadata, file hashes, deterministic and model findings, evidence locations, AI execution metadata, decision statuses, and appraiser notes. It supports your workfile; it is not a replacement for the official appraisal workfile you are required to maintain.

Preflight extracts available PDF text and uses its multimodal agent to visually review the rendered report plus a controlled set of appraisal photos. Visual findings cite filenames and report pages when available. Protected files, unsupported formats, poor-quality images, and package size limits can reduce coverage. Visual findings are review prompts—not definitive property-condition assessments.

Reviews are scoped to your authenticated account, and production files are stored in private object storage. The rendered report, selected appraisal photos, extracted evidence, and related metadata may be sent to the configured AI provider to produce a review. Only upload assignments you are authorized to process and remove unnecessary sensitive data.

CoAppraiser uses the OpenAI API, not a consumer ChatGPT session. OpenAI states that API inputs and outputs are not used to train its models by default unless the customer explicitly opts in. Provider retention and legal exceptions still apply; see the OpenAI API data controls and our Terms & Conditions.

Your first completed Preflight review is free. Continued access is $59 per month through Stripe. You can cancel from the billing portal; cancellation stops the next renewal and access generally continues through the end of the current billing period.