Deterministic Data Room Audits.

AI data room analysis reads across every document in a data room at once and reports where they disagree with each other. Search finds a document; summarisation compresses one. Analysis is the cross-document step: the same number appearing three ways in three files becomes a finding rather than an inconsistency nobody indexed.

Upload the documents you have, as they stand, including the gaps; there is no data-room connector to set up. Missing documents are reported as unverifiable rather than skipped, so an absent cohort file is visible to the committee instead of disappearing into a confident summary. Findings arrive traced to page and line, which is what makes them usable during exclusivity rather than after.

Neither. It does not produce a QoE, and it does not summarize. It cross-examines the claims in one document against the numbers in the others and returns each contradiction as a citable finding, tied to both sources. Your deal team and your QoE provider adjudicate.

Not as an accounting reconciliation. It does not test add-backs against the general ledger or rebuild the working-capital peg; that is the QoE provider's work. What it tests is whether the CIM's narrative survives the financial statements supplied with it. The engine's published record is on pitch decks and public filings (see the WeWork S-1 audit); on CIMs it runs as a scoped pilot on your own documents.

Where the contradiction sits between documents you supply, yes: a concentration or revenue-durability claim stated one way in the CIM and contradicted by the model is the kind of finding the engine is built to hold, tied to both sources. It does not audit contracts or cash receipts.

Only as far as the documents supplied allow. Where the carve-out financials and the parent-level disclosures contradict each other, that surfaces as a finding with both sources cited. Untangling the carve-out itself remains the accountants' work and the deal team's call.

Those retrieve and summarize within a document. askOdin triangulates across documents and preserves contradictions instead of reconciling them away. We do not compete; we consume — keep the room, add the audit.

No. Extraction runs against an external model API under a paid-tier agreement that prohibits training use, and the deterministic engine does the analysis. Processing runs under Stateless API Orchestration and Strict Data Sovereignty; raw documents are held no longer than 30 days under a documented retention ceiling and never enter any training corpus. What persists beyond that is the derived record (the verdict and the analysis), and deletion requests reach it too. The ceiling is documented policy; until its automated enforcement ships, deletion is executed on request. The detail, control by control, is on the security page.

Yes, on the same terms. The test does not change with the buyer: do the target's claims hold against the documents supplied with them? In a bilateral deal there may be no CIM; the management presentation, the business plan and the financial statements you receive are the documents it reads against each other. Findings carry both sources, so they can go into a board paper the way they go into an investment committee memo. It does not replace financial, tax or legal diligence, and on target materials it runs as a scoped pilot.

Yes, as a consistency pre-check. Before the CIM goes out, a sell-side adviser can upload the draft CIM with the financial statements and model behind it and see where they disagree — an adjusted EBITDA stated two ways, a one-time cost that recurs in the prior years — with both sources cited, so the contradiction is fixed before a buyer finds it. It is not a QoE or a vendor due diligence report, and it does not test the numbers against the ledger. The seller's documents are handled exactly as any customer's, as set out on the security page. The sub-processors are named publicly in Privacy Policy §5; the supporting security documentation is provided under NDA. Terms are agreed per deal.