It extracts documented knowledge, infers recurring practices, observes work and asks targeted questions only when context is missing. Your team’s deliberate effort is confirming what matters.
Statistical detection of regularities in your journals, ledger and bank activity: recurring postings, supplier-to-account patterns, timing cycles.
Reading invoices, purchase orders and delivery notes at line-item level—the nature of a transaction, not just its pattern. Client-authored files rank highest: your mapping sheets and policies are read as ground truth.
Calls, email threads and shared Slack channels carry the tacit knowledge no system holds: reasoning, exceptions, who decides what. Capture starts at the first demo.
The one deliberate ask. Centinel ranks every open question by materiality and confidence, and brings your team only the ones worth a human minute—in Slack or in the product.
After onboarding, Centinel watches the ledger: corrections, rule violations, new suppliers and patterns. What it observes becomes proposals—the freshness engine that keeps the Constitution alive.
Public signals and patterns learned from companies like yours—same industry, size, geography, ERP—pre-populate the workspace before the first call. Priors, never policy: everything is corroborated against your real data.
Every inferred rule carries two scores: how much money it governs, and how sure Centinel is. Only high-materiality, low-confidence knowledge generates a question. Everything else is confirmed by evidence—or simply doesn’t matter enough to interrupt anyone.
Nothing governs until it is confirmed. Centinel keeps what it has observed strictly separate from what your team has validated.
Each dossier shows what Centinel has observed across your history, what has been confirmed as policy, and exactly what needs attention—with the euros at stake attached to every open question.
And it runs on data you already have. See what Centinel reads.