Turn borrower documents into structured credit evidence.

Reading the document is table stakes. Knowing what it is, is the work.

Any modern tool can pull text and numbers out of a PDF. The questions that consume analyst time come after: which entity does this statement belong to, which fiscal year, is it the audit or the internal draft, and where should its contents be used downstream. CORE's classification assigns document type, entity, fiscal year, program use, and downstream workflow before an analyst has to hunt.

That is the difference between loan document extraction and loan document intelligence: the file arrives organized into the borrower's structure, not into a generic folder.

Source lineage down to the page and span

Every extracted fact keeps its source document, location, confidence, related notes, and review state — with a split view that shows the PDF beside the fact. When a number reaches a spread, a policy test, or a memo claim, the path back to the page it came from is preserved.

Entity context assembled for review, not hidden in a second system

Borrowers, guarantors, affiliates, owners, projects, and aliases that appear across the document set are assembled as reviewable context. The analyst sees the structure CORE inferred and corrects it before it shapes underwriting — multi-entity complexity is handled in the open.

Analyst gates before anything becomes credit evidence

Extraction proposes; the analyst confirms, supersedes, re-extracts, or rejects. Only accepted evidence feeds spreading, policy analysis, risk work, and memo claims — which is what makes AI-assisted document processing defensible in a regulated or committee-reviewed credit file.