AI Underwriting Pipeline — Sample Outputs

The pipeline works out what each file is before it reads it, so the file names do not matter and neither does which system exported them. What comes back is a deal summary with comps, demand and supply; a trailing-twelve categorised into a consistent profit-and-loss layout; a rent roll rolled up by floor plan; and a manifest of every field it extracted next to the model input each one fills. Enter the purchase price per unit with the files — the cap rates, the pricing ladder and the closing costs are priced off it.

Where a value could not be established the cell says MISSING rather than a plausible number, which is the point of the whole exercise — a visible hole is cheaper than a wrong figure that survives into a model. The caps below are real: this runs on one small server. Uploads and their outputs are kept so I can review runs that went wrong; they are not shared and not used to train anything.