Configure the autonomy control plane

Draw the autonomy line per channel, action, and risk tier. Every decision is deny-by-default and can only narrow what policy already allows — never widen it.

Purpose

Decide, per channel and action and risk tier, whether work runs automatically, goes to AI-augmented human review, or always requires a human — knowing the server still clamps every choice to the existing policy floor.

Who this is for

Workspace owners and admins who decide how autonomously the assistant may act, and campaign managers who need to read the current settings.

Prerequisites

A tenant owner or admin role to change a line; owners, admins, and campaign managers can read it.
A project to scope the settings to.
Agreement on which actions and risk tiers your team is comfortable letting the assistant handle.

Steps

Open Autonomy from the left navigation and choose the project to configure.
Read the three-tier model: auto-approve, AI-augmented review, and mandatory human gate.
In Draw an autonomy line, pick a channel, action, risk tier, and the disposition you want.
Set the declared class and the operational design domain bounds, then save the line.
Confirm the line appears in the configured lines table for that project.

Expected result

Your autonomy lines are saved for the project and shown in the table, while anything not configured continues to default to a mandatory human gate.

Verification

A saved line appears in the configured lines table with its disposition.
Removing or not configuring a line leaves that action at a mandatory human gate.
A declared class above the substantiated ceiling is refused rather than stored.

If something goes wrong

A line will not save with a higher autonomy class: A class above what telemetry can substantiate is refused on purpose; choose a class within the substantiated ceiling.
An action still asks for human approval after you set auto-approve: The control plane can only narrow, never widen: if existing policy gates the action, the human gate stays, by design.
Your configured lines disappeared: Settings are currently held in memory; durable storage is being added, so re-save the line if the service restarted.

Permissions

Only a tenant owner or admin can raise or lower an autonomy line. Campaign managers can read the control plane but not change it.

Support

If you are unsure which disposition fits an action, start at mandatory human gate and relax it only with evidence. Ask your workspace owner for changes you cannot make, and LumenReach support can explain the model without changing your settings.