Is Rook just a ChatGPT wrapper?

No. The financial math — DSCR, IRR, cap rate, waterfall, valuation, forecast — runs in dedicated code with fixed inputs and outputs. AI orchestrates the workflow; it never produces the numbers. Every figure on a deliverable traces back to a source: a comp pulled from Redfin or Zillow, a record from the county assessor, or an explicit formula in the proforma.

Where do your comps come from?

Named sources, not scraped from the open web. Redfin and Zillow are the screening layer. County recorded deeds are the closing-grade source. Census ACS provides demographics. Every comp on a deliverable shows where it came from and when it was pulled. For deals entering due diligence, MLS-driven comps are crosswalked against county records before they enter the proforma.

Will an appraiser respect this work?

Yes — because we don’t compete with the appraisal, we feed it. Rook produces Broker Opinion of Value (BOV) tier work product: sourced comps, paired-sales reasoning, defensible cap-rate selection, and a Highest-and-Best-Use narrative as discrete deliverables. The appraiser vets our work; they don’t rebuild it from a blank page. Rook does not issue USPAP-compliant appraisals — that remains the appraiser’s scope.

Will a bank flag this as high-risk?

No. The lender package is its own deliverable, not a repackaged proforma. It includes T-12, rent roll, sources and uses, DSCR projections, stress sensitivities (rate +200 bps, vacancy +500 bps, expense +10%), and sponsor bio. Every figure ties back to the proforma with a clear source-of-input. If a specific lender uses their own submission template, we intake it once and reuse it on subsequent submissions at no recurring cost.

Will a CFO have to rebuild the model to survive an audit?

No. The proforma is built for review in both directions: top-down through the NOI cascade, bottom-up by drilling into any cell. Every dollar cites a source or a formula. The Assumptions tab centralizes inputs, and the audit-trail tab logs every change between versions. A skeptical CFO audits assumptions in thirty minutes — not thirty hours. The math remains theirs to challenge, but not to retype.

Do you have real underwriting discipline, or is this dressed-up document reading?

Real discipline. Rook treats real estate as structured data — properties, parcels, owners, leases, comps, jurisdictions — with provenance attached, not text extracted from PDFs. File-naming, version control, audit trails, and approval gates on any destructive action are enforced by a checked-in governance file, not by good intentions.

Are your outputs sourced, scored, traceable, and defensible?

Yes — that is the architecture, not a slogan.

Sourced. Every market-data figure shows where it came from and when.

Scored. Comps, valuation, and forecast outputs carry confidence ratings. Low-confidence figures flag for analyst review rather than entering the model silently.

Traceable. Version history is preserved end-to-end, with changes between versions logged.

Defensible. A reviewer can reconstruct any number on any deliverable.

Other AI tools skip data verification. Do you?

No — verification is where Rook spends its time. The flow is pull, store, verify, score, integrate. Speed comes from never re-running verification that’s already complete, not from skipping it. Where public records are thin or unavailable, the deal carries a “manual verification required” flag, and an analyst clears it before the proforma locks. We don’t paper over a weak data source.

Tools running autonomously compound errors. Do you run autonomously?

No. Every delete, overwrite, bulk move, publish, or transmit follows an explicit Plan → Approve → Execute → Verify → Summarize sequence. On critical errors, the system stops and reports — no silent retries. For tokenized deals, deal-specific approval gates can be added so that any artifact heading on-chain requires an explicit sign-off recorded in the audit trail.

Aren’t you just another part of the AI bubble in real estate?

The bubble label fits tools that built fast UIs over unverified data. Rook starts at the data layer — provenance, source records, audit trail, version discipline. We agree with the critique and built the workspace around it deliberately. Speed is downstream of trust, not the other way around.

How does Rook handle tokenized issuances?

Tokenization compounds data risk: an error in the offering memo propagates to every wallet holder once the issuance settles on-chain. Rook produces the underwriting artifacts a tokenization platform’s broker-dealer, transfer agent, and Reg D+S compliance review require — BOV, investor materials prep (not a PPM), lender package, waterfall — with provenance preserved end to end. Dedicated workflows map to standard institutional capital structures (closed-end preferred, evergreen open-end, open-end bridge, hybrid dual-tranche, co-GP) so the integration is engineered, not retrofitted.

Contact Rook.

Send a message