Browse 4 articles tagged risk-controls across approvals, auditability, plugins, and controlled AI operations.
The configuration underneath an approval gate - durable resume and revalidation, role separation as policy, building gated actions, and the problems still open.
Read Approval Design: Durability, Policy, and Plugins →A single AI-recommended customer refund moves through a two-person review gate, from recommendation to execution record, on real screens inside Latch.
Read Approval Design in Practice: One Refund, End to End →How AI triage should use confidence thresholds, missing context, and deliberate silence to preserve operator trust in high-risk workflows.
Read When AI Triage Should Recommend the Next Step and When It Should Stay Silent →How to design approval gates for high-risk operations that shrink blast radius, enforce two-person review, and keep the evidence on the case.
Read Approval Design for High-Risk Operations Workflows →