Municipal courts explore generative drafting of routine notice templates under human review. What correctly classifies this on the DCAI workload map?

Municipal courts exploring generative drafting of routine notice templates under human review are still operating a generative AI workload on the DCAI map. Using standard template language does not make this a rules-based automation workload; the system is still generating draft text through a model, and human review is a governance control layered on top of model serving rather than a change in workload type. Calling it rules-based automation because templates use standard language confuses template style with workload taxonomy and would incorrectly exclude generative serving from capacity planning. A discriminative classification workload sorts notices into categories, but here the model is generating notice text rather than classifying pre-written notices, so it does not match the DCAI generative drafting pattern. A retrieval search workload reuses approved notice text by finding existing matches, whereas this scenario drafts new notice text with a model, so it is not a retrieval pattern. Exam caveat: human-in-the-loop does not remove obligations for prompt and data controls, audit logs, model versioning, and retention—governance sits on the generative workload rather than erasing it from the municipal AI catalog. Operational check: document the human review gate, ensure draft notices remain clearly marked until approved, retain generation logs for audit, and capacity-plan generative serving separately from unrelated classification or retrieval workloads.