What is included in every plan
Every plan is built around reviewed operations, not unsupervised automation. The limits change by plan, but the operating model stays the same.
Review before action
Draft replies, workflow changes, and policy-sensitive decisions stay reviewable before anything customer-facing or external runs.
Structured AI outputs
Reasoning steps return structured data that can be checked, displayed, and audited instead of free-form automation instructions.
Deterministic policy checks
Rules and policy checks remain the final authority for whether work can proceed, needs edits, or should be held for a person.
Execution traces
Each run records the trigger, selected context, missing information, policy result, tool calls, and final decision.
Permissioned connectors
Inbox, document, and future system connections are scoped to the work being reviewed and remain auditable.
Human override
Teams can edit drafts, add context, reject proposed work, and keep uncertain cases out of automated handling.