Bart Labs · Milwaukee, WI
AI that hunts. Algorithms that catch.
Accountability infrastructure for public and private financial data.
Our Mission
Public money is public record. But public records alone don't tell you where the problems are — that takes systematic analysis at a scale no human team can sustain.
CivicSentinel exists to close that gap. We apply 15 deterministic fraud detection algorithms — built to professional audit standards — against 100% of financial transactions. Every payment, every vendor, every contract. No sampling, no guesswork.
The AI's job is to gather the data and investigate what the algorithms flag. The algorithms do the detecting. That distinction is deliberate: every finding we surface can be traced back to math, not a model's opinion.
Serving public, private, and non-profit clients — target launch 2026
Who We Serve
CivicSentinel is designed to serve different kinds of accountability needs — from open civic access to deep-dive private engagements.
Public Data
Analysis of publicly available government financial records — spending, contracts, payroll, and vendor data — surfaced through the open CivicSentinel dashboard.
Private Data
Detection runs against proprietary or non-public financial data — for organizations that need the same rigor applied to their own records before someone else does.
Non-Profit & Civic
Subsidized access for non-profit advocacy organizations, community groups, and public interest journalism outlets working on government accountability.
The Process
AI crawlers collect and normalize financial data from public portals and client-provided sources — transactions, contracts, payroll, vendor registrations — covering 100% of records in scope.
Fifteen deterministic algorithms — Benford's Law, duplicate detection, bid-rigging patterns, payroll anomalies, vendor clustering, timing analysis, and more — run against every record.
Flagged findings are investigated by AI models that cross-reference public records, identify relationships, and draft preliminary findings in GAO 5-Element format.
Every finding includes the condition, criteria, cause, effect, and recommendation — traceable to source data, auditable, and ready for human review or publication.
Our Standards
No finding is generated by a language model alone. Every anomaly surfaces from deterministic rules — the same methodology used by professional auditors. AI does research and drafting, not detection.
Statistical sampling misses things. CivicSentinel runs every detector against every transaction in scope. If a pattern exists in the data, it will be found — not estimated.
Every finding cites the specific transaction IDs, dates, vendors, and amounts that triggered it. Nothing is published that can't be independently verified against public records.
Methodology conforms to GAO Yellow Book 2024, ACFE 2024, COSO 2nd Edition, and Daubert evidentiary standards — the same frameworks used in federal audits and litigation.
CivicSentinel works only with data we have clear authorization to analyze — public records that are legally available, or private data provided by the client. We do not solicit leaked documents or work with data of uncertain provenance.
Findings are published with methodology, not just conclusions. Anyone can review how an anomaly was detected, what criteria it was measured against, and what the underlying data shows.
CivicSentinel is targeting a 2026 launch. Public dashboard, private audit services, and civic access program — all from a single platform.