Extraction is AI-assisted, verification is human. AI reads the contract package — MSA, order forms, amendments, side letters — and extracts the commercial terms, resolving which document governs when they conflict. A person verifies every extracted term and every finding against the source documents before anything reaches your report. A false positive in front of a controller is a failure; we treat it that way.
The math is deterministic. Once terms are confirmed, expected billing and recognition are calculated, not guessed. Every dollar figure in the report is reproducible from the confirmed terms and your exports.
Entity mapping is proposed, you confirm. Matching the customer in a contract to the customer in Stripe and the row in your revenue schedule is genuinely error-prone work. We propose the mappings; you confirm them before checks run against them.
Judgment areas are flagged, not resolved unilaterally. Standalone selling price allocation, bundled performance obligations, variable-consideration estimates, and unusual termination or acceptance provisions are identified and framed as questions for your revenue accountant or auditor — with the relevant clauses assembled — rather than answered by the scan.
Findings are classified by what they are. Dollar-quantified billing variance, revenue-treatment variance, and control exceptions are reported separately. Missing-document and hygiene issues are listed, but never dressed up as dollar findings.