Autonomy you grant, one switch at a time.
Every MarketiQ agent works at the level you set — suggest, approve or auto — inside a workspace mode, a policy and a kill switch. Publishing, sending and spending wait for a person until an owner or admin grants a named standing authorization, and even then every check still runs.
- Every agent
- War Room
- Ivy · email
- Rook · off-page
- Jett · ads
- Vera · review
How a governed action moves.
- Any agent
An agent prepares the action
A post to schedule, an email to send, a budget change. Reversible internal work such as research and drafting can run on its own.
- Policy gate
One policy point decides
The workspace mode caps the agent first (Manual means suggest only; Copilot sends high-risk actions to approval), then the level it has earned with proof. If the policy can’t be read, the action waits for approval, never runs silently.
- Each desk
The switches are read now
The kill switch, the room’s pause, per-post sign-off and standing authorizations are re-read at the moment of acting, not when a batch started.
- Approvals
A person decides the rest
Anything the policy does not allow waits in Approvals with the draft and the reason it stopped.
- Executor
Act, then record
Approved actions run through the same idempotent claims as manual ones, and governed actions are written to the audit log with who or what approved them.
The switches you hold.
Autonomy level per agent
Suggest plans only. Approve prepares real actions for one-click approval. Auto acts within cost caps and the kill switch. Set it per agent in AI Team.
Workspace mode
Manual caps every agent at suggest. Copilot sends high-risk actions to approval. Autopilot keeps each agent’s level, and the proof gate still applies.
Pause all autonomy
One switch in AI Team stops agents from acting while they keep planning. The War Room also has its own pause.
Standing authorizations
Named, revocable grants from an owner or admin: the War Room may schedule, Ivy may send, Rook may publish off-page articles. Every change is audited.
Per-post sign-off
Require sign-off before anything goes live: a post can then be scheduled or published only after its approver says yes.
Ceilings and caps
A per-run spend ceiling and daily output caps for the War Room, change limits for ad budgets, and a weekly broadcast cap and new-domain warm-up for email.
Designed to fail safe.
Fail-closed by default
If a policy can’t be evaluated, the action is prepared for approval instead of running.
Earned autonomy
For high-risk agents, the level an agent may use today depends on its proven track record, not only on the setting.
Tighten mid-flight
Tightening the War Room policy reaches runs already under way; loosening it never widens them.
Truthful outcomes
A provider answer that never came back is marked unknown and reconciled: never retried blindly, never reported as done.
An audit trail
Governed actions record their actor — a person, or an agent under a named authorization — and ad changes keep their before and after.
Roles that match the work
Owners and admins set policy and authorizations, managers and editors run the work, and viewers are read-only.
Where the controls apply.
The same policy point governs every desk; these desks add their own standing authorization or limits.
- AI TeamBuilt inLevels per agent, pause all autonomy
- War RoomBuilt inPolicy, spend ceiling, standing authorization
- Email · IvyBuilt inSend authorization, weekly broadcast cap, domain warm-up
- Off-page · RookBuilt inPublish-on-own switch
- Ads · JettBuilt inSpend approvals, change limits, rollback data
- ApprovalsBuilt inOne inbox for what waits on a person
- LinkedIn automationBuilt inDesktop opt-in: device consent, owner or admin authorization, Autopilot
- Built in
- Part of MarketiQ; nothing to connect.
What is governed autonomy in MarketiQ?
It means every agent acts only at the level you allow. Each agent is set to suggest, approve or auto; a workspace mode, policies, standing authorizations and a kill switch sit on top; and anything that publishes, sends or spends waits for a person unless you have explicitly authorized it.
Can an agent publish or spend money without me?
Not by default. Publishing, email sends and ad spend changes wait for approval. An owner or admin can grant a named standing authorization for a desk, and the agent must still pass every check at the moment it acts.
What does the kill switch do?
Pause all autonomy in AI Team stops agents from acting while they keep planning. It blocks new guarded actions; an operation already sent to a provider may still finish, and it is reconciled rather than undone.
What is the difference between Manual, Copilot and Autopilot?
Manual caps every agent at suggest, so nothing happens without you. Copilot lets low- and medium-risk work run at the agent’s level but sends high-risk actions to one-click approval. Autopilot keeps each agent’s own level, and the proof gate still decides what it has earned.
Who can change these settings?
Owners and admins set autonomy, policies and standing authorizations. Managers and editors run the work, and viewers are read-only. Changes to a standing authorization are audited.
See it work on your brand.
Book a demo and we will run MarketiQ on your own website, or choose a plan and start your workspace.