Executable orchestrator that turns tokenized/extracted CRE document content (rent rolls, T-12s, operating statements, Prose Frontier narrative artifacts) into validated, typed, auditable, target-model-ready database payloads. Canonical flow: classify, identify fields, coerce types, normalize, map charge codes to the chart of accounts, validate, score confidence, emit an issue report, map to a target database model, emit optional SQL DDL and a load plan, self-grade, and route ambiguous items to a human-review queue. Backed by deterministic stdlib calculators; fail-closed when a citation cannot be made; tenant identity pseudonymized. Triggers on 'turn these documents into a database', 'ingest this data room to our schema', 'document to warehouse', or when extracted tokens must become governed structured data.
Transforms a tokenized/extracted operating statement (any line type: actual, budget, reforecast, prior-year; any period grain; monthly, quarterly, annual-summary, or multi-scenario) into validated, account-level, period-level database records mapped to the canonical chart of accounts. Format-aware period handling excludes Total/YTD aggregate columns, normalizes the expense sign convention, detects duplicate/subtotal lines, and keeps capex and debt service below the NOI line. Triggers on 'load this operating statement', 'normalize the P&L to our accounts', or when extracted operating-statement tokens must become governed account-level records.
Reconciles a normalized rent roll against a normalized T-12 on a stated, consistent basis (annualized contractual vs recognized accrual; collected cash out of scope) and never forces a tie. Reconciles base rent, recoveries and other income (jointly, per the canonical chart), occupancy, and the EGI / NOI-revenue bridge; classifies each gap as mapping, timing, or missing on a deterministic signature; surfaces residual_unexplained; and routes every untied dimension to human review. Triggers on 'tie out the rent roll to the T-12', 'reconcile contractual rent to actuals', 'NOI bridge from the rent roll', or 'revenue leakage check'.
Transforms a tokenized/extracted rent roll into validated, typed, auditable, database-ready records: a multi-line charge schedule (base rent, CAM/tax/insurance recoveries, percentage rent, parking, storage) mapped to the canonical chart of accounts, lease- and unit-level facts, GPR and occupancy aggregates, a data-quality grade, and a target-model load plan. Tenant identity is pseudonymized; per-unit natural-person data never leaves the boundary. Triggers on 'load this rent roll into the database', 'normalize the rent roll to our schema', 'rent roll to warehouse', or when extracted rent-roll tokens need to become structured records before underwriting or tie-out.
Transforms a tokenized/extracted trailing-twelve-month operating statement into validated account-level monthly records. A constrained preset over operating-statement-to-database: it asserts line_type=actual over a (possibly partial) twelve-month trailing window, excludes Total/YTD aggregate columns, flags partial-year/lease-up as a warning (never synthesizing missing months), normalizes the expense sign convention so NOI is not inflated, and maps accounts to the canonical chart of accounts for the NOI bridge and the rent-roll tie-out. Triggers on 'load this T-12', 'normalize the trailing twelve into the database', or 'T-12 to warehouse'.
Orchestration skill that assembles the OUTPUT of single-document extractors into validated, warehouse-ready tabular datasets. It does not re-extract individual documents; it sits above the extractors and turns their per-document fact tables into multi-document datasets with declared extraction schemas, data-quality and validation rules, standardized provenance columns, warehouse table naming, and a deck-readiness gate. Triggers on 'build the warehouse dataset', 'assemble these extractions', 'validate the data room for the model/deck', or when several extracted documents must become one queryable table. Failing rows are surfaced for review, never silently dropped. Output is the validated dataset that warehouse-to-exhibit-mapper consumes.
Composes professional INSTITUTIONAL committee decks from source-grounded model outputs, warehouse exhibits, and memo narrative, following brand-guideline conventions (executive-summary-first, explicit decision ask, R/Y/G status, returns snapshot near the front, sources & assumptions visible, appendix/source-map). Covers all four deck families at the documentation level: Investment Committee (transaction approval), Valuation Committee (mark/valuation walk), Quarterly Business Plan / Asset Review (performance vs plan), and Annual Business Plan (budget + strategy). Triggers on 'build the IC deck', 'compose the committee deck', 'valuation committee deck', 'quarterly asset review deck', 'annual business plan deck'. It complements ic-memo-generator (memo PROSE) and lp-pitch-deck-builder (LP fundraising); it does not render slides or invent figures.
Maps validated, warehouse-ready tabular datasets into deck-ready EXHIBIT specifications and slide inputs. Selects table vs. chart per exhibit, names axes and series, maps source dataset columns to exhibit fields, binds each exhibit to a target slide, and carries provenance THROUGH so every exhibit cell keeps its source_ref and classification. Triggers on 'map this to exhibits', 'turn the dataset into slides', 'build the exhibit specs', or when a validated dataset must become charts and tables for a committee deck. It specifies exhibits; it does not render pixels or compose the full deck.