Sweep thousands of leases for compliance using Amazon Quick and the Adjudicated Query pattern

This article introduces the Adjudicated Query pattern for conducting compliance sweeps across thousands of leases using Amazon Quick and a deterministic rules engine.

• Bounded conversational layer translates natural-language questions into fixed operations; AI never touches compliance decisions
• Deterministic rules engine handles all pass/fail determinations with versioned rules as data, not code
• Completeness receipt proves every record was evaluated and categorized (compliant, in-breach, ambiguous, unreadable)
• Reference architecture uses Amazon Quick, Lambda MCP server, Aurora Serverless v2, API Gateway, Cognito, and QuickSight
• Six bounded operations prevent silent population narrowing: sweep_compliance, simulate_rule_change, explore_clauses, get_finding, list_rules, check_connection
• Findings are append-only with full evidence chains for defensibility in future litigation or audits
• Sample implementation available on GitHub with synthetic data, deterministic corpus generation, and acceptance tests

The pattern ensures compliance answers are provably complete and defensible by keeping AI out of the decision path while maintaining conversational access through a bounded tool surface.