AWS published a blog post on October 2 describing a design pattern that uses its Amazon Quick generative AI service and a deterministic rules engine to check thousands of leases for compliance with state laws. AWS Machine Learning Blog
The pattern, called Adjudicated Query, aims to deliver provable completeness and defensibility in regulatory checks that can be challenged in audits or litigation, the post states. Business users can ask natural-language compliance questions in Amazon Quick, while pass/fail decisions remain in a non-AI rules engine.
The post describes a challenge of managing 50,000 leases across multiple states where landlord-tenant statutes change on the legislature’s schedule. At small scale a paralegal reviews leases; at large scale software handles the task but, the post says, can lack verifiable audit trails. AWS’s example pairs a generative AI front end with a deterministic backend to log which rule version was applied to which clause, on what date, and by whom.
The blog post describes how the pattern can also apply to other high-stakes compliance domains such as sanctions screening and insurance claims adjudication.
The blog includes a reference architecture and a working sample that users can run, but provided no independent benchmarks, no pricing for any underlying services, and no details of real-world deployments in the post itself. The publication is a first-party company statement.