Your resource is ready. Read it below or open the PDF.
Open the PDF

The existing process

The team reviews leases, service contracts, legal documents, loan information and other material associated with property acquisitions. Final investment summaries are typically 20–30 pages and combine internal underwriting information with external sources, broker documents and formatted Excel outputs.

Where time is spent

Initial data gathering from outside parties is a major bottleneck. The supplied notes describe another 3–4 days of summary preparation once the required information is in hand. Corrections often need to be made upstream in the underlying model rather than only in the finished report.

What is already automated

The team reported that roughly 95% of data movement within the existing workbook is dynamic, with about 5% requiring manual entry. That distinction matters: the proposed project should improve the remaining extraction, transformation and preparation work without discarding working formulas and macros.

The proposed first phase

Begin with sanitized Excel exports and representative source documents. Map the manual steps between the raw-input and preliminary-report tabs, preserve the existing model logic where possible, and identify which extraction and transformation tasks are suitable for AI-assisted processing with human review.

Later phases

Assess approved direct data connections where they are available, then consider report preparation, review workflows and an internal assistant. Access to CoStar, Argus and other sources must follow the applicable product interfaces and licensing; public or unrestricted API access is not assumed.

How success would be measured

Compare baseline and pilot time after inputs are available; assess field-extraction quality on reviewed samples; track corrections and the completeness of source references. No percentage time-saving or accuracy result is established by the supplied discovery notes.

Expected outcomes

Begin with sanitized Excel exports and representative source documents. Map the manual steps between the raw-input and preliminary-report tabs, preserve the existing model logic where possible, and identify which extraction and transformation tasks are suitable for AI-assisted processing with human review.