CRE Document Ingestion

Provide the rent roll as Excel, CSV, or PDF and the parser maps varied column headers, normalizes unit types like 2x2 to 2BR/2BA, and outputs unit mix, occupancy, rents, and lease expirations as JSON. It validates unit counts and flags rent or square footage outliers.

The financials parser handles monthly-column, annual summary, and budget-comparison T-12 formats, extracting revenue lines, expense categories, and NOI with per-unit metrics. It excludes below-the-line items like CapEx and debt service and validates that revenue minus expenses equals NOI.

The classifier identifies rent rolls, T-12/T-3 operating statements, offering memoranda, pro formas, PSAs, surveys, inspection/PCA reports, Phase I ESAs, title commitments, and lease abstracts. Each classification includes a confidence level, and ambiguous files are flagged for user confirmation.

Yes, it accepts PDFs, PowerPoint, Word documents, and image-based property flyers, but OCR quality directly affects extraction confidence for scanned files. Key fields like asking price, unit count, and cap rate are located via phrase recognition and section headers.

Conflicts are never silently resolved; each is flagged with both values and a recommended resolution based on source authority. For example, the rent roll is treated as more authoritative for occupancy while the OM is authoritative for asking price.

Confidence drops below 0.70 for partial-year financials, heavily combined expense categories, pro forma mixed with actuals, or PDFs with ambiguous table structure. Missing critical fields like address, unit count, or price block underwriting and require manual input.