A summary is not the same as knowing what to do
Regulatory information rarely arrives as a clean instruction for one organisation. The difficult questions come after reading: what changed, does it apply here, when does it take effect, and what deserves action?
ARIA explores a workflow from monitoring and investigation through comparison, evidence, impact assessment and action planning. Organisation context changes the meaning of a generic update, so the system begins with both the source and the profile it may affect.
Reasoning and operations have different jobs
Model reasoning handles interpretation, comparison and uncertainty. Deterministic code handles thresholds, dates, jurisdiction matching, document versions, approvals and audit records. Typed tools and structured outputs let application code inspect the agent’s work before anything is executed.
Autonomy stops at the execution boundary. The agent can prepare a controlled plan, but a person approves consequential external action.
The larger pattern
ARIA demonstrates a pattern beyond regulation: use language models where interpretation is valuable, deterministic code where predictable behaviour matters, and explicit controls at the boundary between preparation and action.