The business behind the platform.

External confirmations are core audit evidence under ISA 505 — and presumptively required for receivables under PCAOB AS 2310 as revised, effective for fiscal years ending on or after June 15, 2025 — and at most firms they are weeks of manual labor: choosing samples defensibly, chasing signatures, mailing requests, re-typing responses into Excel, and reconstructing what happened for the file. Incumbent platforms are built as delivery systems — the request gets there, but the last mile from response back to reconciled, documented evidence stays manual. Confirmatica's thesis is to own that last mile: extraction, verification, reconciliation, and the working paper, in one system with one audit trail.

The initial market is Czech and Slovak mid-tier audit firms and international-network member firms — the segment incumbent platforms serve least well. The local reality is engineered in, not localized after: ledger imports that understand Pohoda, Money S3, Helios and SAP-style exports and Czech number formats; per-institution workflow types for banks that run their own confirmation processes; a paper-with-QR fallback for offline counterparties; bilingual Czech/English portals for signers and responders; a curated institutional directory seeded with CZ/SK banks and institutions, where every role address must carry a cited published source. Sales are direct and relationship-based; access is manual-approval only — no self-serve tier exists by design. Volume sizing basis (design assumption, not a forecast): 100 firms × ~30 clients × ~40 confirmations ≈ 120,000 confirmations a year — a small-data, high-value-transaction product.

One annual plan per firm with a fair-use confirmation-volume cap per tier — no per-seat pricing, no per-confirmation billing. Sales are direct and relationship-based; platform access is manual-approval only, which matches how audit firms actually buy. The cost mechanics favor software margins: marginal cost per confirmation is measured in cents (Document AI pages plus Gemini tokens — an engineering estimate from metered cloud pricing, not audited COGS), the native e-sign flow has no per-envelope vendor cost, and fixed infrastructure idles at low cost on scale-to-zero. No revenue actuals exist yet; we publish the mechanics rather than a projected margin.

1. CI-executed cross-tenant isolation proof — a probe suite attempts cross-tenant reads and writes against real Postgres with RLS enforced, and fails the build on any leak (tests/test_isolation.py).
2. Inspector-runnable evidence-integrity verifier — a standalone read-only script re-walks every hash chain (apps/api/scripts/verify_chain.py); write-once anchors mean even a database administrator cannot rewrite history undetected.
3. A model governance log — every AI call recorded append-only with model id, prompt/response digests, and latency (model_calls). Inspectors and DD teams ask; few vendors can answer.
4. Controls enforced in the data model — unverified-contact dispatch refusal (test_dispatch_blocked_for_unverified_contact), authorization-before-dispatch in the state machine, human-conclusion gates — each with a named test in a CI suite of 180+ backend tests.
5. One-artifact close-out — the engagement evidence pack: manifest, fresh chain-verification verdict, signed authorizations, letters, response documents, working papers, in one ZIP.
6. Session and MFA hardening under test — revocation across five triggers, TOTP with single-use backup codes probed for concurrent double-spend, IP-independent account lockout.
7. Adversarial audit discipline — a 66-finding multi-model audit, all findings resolved with regression tests, then re-verified in a second, separate multi-model audit pass.

Investor and partner enquiries proceed through existing relationships. There is no data room, deck, or offering on this page — and platform access for firms remains approval-gated at Request access.