Methodology
How the ask pre-send check works.
The product first checks whether the proposed ask is send-ready. Only then does it assemble the revised response artifact for the inspection window.
What this is for
Buyer agents first, then buyers under contract who need a reviewed first draft.
- Buyer agents who need a narrow, defensible first ask fast.
- Buyers under contract with a short response window.
- Inspection asks that need a send, revise, or do-not-send verdict quickly.
What this is not
Not an inspector suite, broad repair-cost blog, full renovation planner, or lender approval promise.
- The product is not trying to replace inspection report software.
- The product is not trying to estimate every cosmetic project.
- The packet is not a contractor bid or underwriting guarantee.
- The output is intentionally narrower than the total future maintenance backlog.
Inspection extraction
The first layer is the user's own report language. Pasted text and quick-add findings are normalized into a concise set of inspection items. If the user uploads a PDF, TXT, or image, the evidence layer tries to match the strongest items back to report pages.
Lead vs. verify vs. cut
Each finding is evaluated against must-request terms, verify-next terms, component signals, and verdict-plan anchors. The goal is not to preserve every detail from the report. The goal is to keep only what improves negotiation leverage and cut what weakens the ask.
Scoped exposure
Cost data is used only after the findings are scoped. Scoped exposure means lead items plus a disciplined verify reserve, so the ask reflects the live deal instead of the entire future repair backlog.
Ask range construction
The defensible, target, and stretch asks are influenced by lead-anchor strength, quote support, closing window, loan posture, lender-visible signals, and matched report evidence. If quote or evidence support is missing, the model stays more conservative and keeps the result in review mode.
Evidence and OCR
If a citation is marked (OCR), the packet is telling the user that the line was recovered from a scan or photo and should be checked against the source page before sending. OCR raises usability, but it never removes the need for user verification.
Pre-send gates
The visible verdict is based on hard gates: deadline, form path, evidence, financing boundary, stage, ownership, number basis, seller pushback, and send posture. Confidence is secondary; gates decide whether the ask is safe to send.
Limitations
- The reviewed ask is not a binding contractor quote.
- Hidden conditions, access constraints, and local contractor availability can materially change actual pricing.
- Lender-visible logic is an aid to negotiation framing, not a promise of appraisal or underwriting outcomes.
- Users should confirm OCR-marked excerpts against the source page before sending.
Method questions
Three practical questions behind the product choices.
Why seller credit first?
Because buyer agents usually need one clean, defensible ask more than they need a sprawling repair specification. Seller-credit posture can survive short response windows better, but the product still checks whether credit wording is appropriate.
Why can local context be optional?
Speed matters. The first-pass job is to judge the ask. Local market and age context improve precision, but a broad U.S. baseline is better than losing the user before the pre-send check exists.
Why is the core flow free during validation?
Because the current proof point is whether buyers and agents actually use the pre-send check, copy the revised packet, and trust it enough to review real files. Charging before that proof would slow the learning loop.