A generic LLM can describe a standard. Auditus.ai is being built to track what auditors actually flag, how often, and what resolved it — stored in an engagement vault that strengthens with every audit processed through the platform.
Illustrative figures. The numbers on this page describe the shape of the engagement-vault network effect as the platform scales. They are not historical performance, achieved savings, or commitments. Auditus.ai is pre-launch.
audits the vault is designed to ingest in Year 1. Each anonymized engagement enriches the checklist and benchmarks. We do not train on any auditor’s or competitor’s work.
distinct deficiency patterns the checklist is being designed to cover at launch. New patterns are reviewed by practicing CPAs and former CPAB inspectors before they join the library.
illustrative audit-fee rework that a cohort of this size could address, plus 49,300 preparer hours. Actual outcomes vary by engagement.
A version-controlled, paragraph-mapped checklist tied to 49 standards across CAS, IFRS, NI, CPAB, CSQM, and Rule 204. Maintained by practicing CPAs — not by prompt engineers. 31 CPAB inspection themes are explicitly calibrated against. Last updated 2026-05-14.
See the standards library →Every score is benchmarked against a peer cohort drawn from comparable engagements in the vault. A new entrant cannot compute these percentiles without comparable engagement volume — and we add to the vault every week.
See the scorecard →Auditors mark each finding as Confirmed, Disputed, or Not raised. 89% of our findings are auditor-confirmed today. Every signal retrains the weights. The model becomes more precise with every engagement — and more wrong only on edges we then explicitly fix.
See the findings + feedback loop →Every finding cites a specific standard paragraph, a specific source document, and links to the cohort of prior engagements where the same pattern appeared. Auditors and audit committees can independently verify any conclusion. The audit trail is the product.
Open an example finding →Each anonymized engagement is designed to sharpen the checklist, scorecard benchmarks, and risk model. Illustrative — actual vault size at launch will be reported separately.
Every finding will be grounded, cited, and reviewed. Patterns surface from the corpus, not from prompt engineering. Pre-launch figure.
The patterns become the proprietary checklist. Illustrative — the live count will be published once the platform is in production.
Illustrative cohort-level modelling, not historical client savings. Actual outcomes vary by engagement.
The platform is designed to invite auditor confirmation on every finding so the model calibrates. Target operating range, not historical performance.
CAS, IFRS, IAS, NI, CPAB, CSQM, Rule 204 — fully indexed and version-controlled. See the Standards page.
Each finding in the current engagement is connected to a cohort of prior engagements where the same deficiency pattern appeared. The higher the count, the more we know about how it remediates.
“MD&A vs. financials inconsistency — near-universal pattern”
“Bank confirmation completeness gap”
“Post-ERP-cutover ITGC gap — extremely common in mid-year migrations”
“New customer contract — missing IFRS 15 memo”
“Prior-year significant deficiency carry-forward”
“Going-concern forecast lacking stress scenarios”
Audit working papers are confidential to the firm and the client. No new entrant can scrape this data; it has to be earned, one engagement at a time. We’ve already earned 247 of them.
Every pattern in the library is reviewed by a practicing CPA and mapped to specific standard paragraphs. Building this team in parallel — while shipping a product — is a separate moat.
Auditors confirming or disputing findings is a labeling signal competitors can’t fake without comparable trust and access. That trust compounds with every cycle.
Peer cohort percentiles need volume. A new entrant cannot tell an issuer where they sit relative to peers; we already can.
CPAB priorities and IFRS standards change every year. Our checklist is version-controlled and explicitly mapped to the active period’s priorities. Staying calibrated is daily work.
Auditors trust Auditus.ai because every output is cited. That trust is the moat — and the citation infrastructure took years to get right.