Autonomy levels
You decide how much the AI is allowed to do.
Autonomy is a setting, not a plan tier. Change it at any time, scope it as narrowly as a single campaign, and vary it by team.
Full Control
Regulated & enterprise default
Agents draft everything and execute nothing. Every action waits in the approval queue until a named human releases it. Each decision is written to the audit trail with the policy that governed it.
Guided Autonomy
Most mid-market teams
Agents execute freely inside guardrails you set — budget ceilings, send caps, approved channels, brand rules. Anything exceeding a limit stops and asks. Everything else runs.
Full Automation
High-velocity teams
Agents run end to end and report outcomes. Still fully logged, still reversible, still subject to the kill switch. Autonomy is not the absence of oversight.
What ships on every plan
The controls your security review will ask for.
Mid-market security reviews routinely run six to nine months. These answers are built into the product so yours does not have to reconstruct them.
Approval gates
Human-in-the-loop boundaries scoped per campaign, per module or globally, with named approvers and escalation paths.
Evidentiary audit trail
Every agent action, proposal, approval and block — timestamped, attributed, immutable and exportable for evidence packs.
Fleet-wide kill switch
Halt every agent mid-task in one click. In-flight actions roll back. No support ticket, no waiting on a vendor.
Policy engine
Budget ceilings, daily send caps, channel allow-lists and brand rules enforced before an agent acts, not audited after.
RBAC, SSO & SCIM
SAML single sign-on, automated provisioning and de-provisioning, and permissions scoped per brand and per capability.
Multi-brand isolation
Each brand keeps its own data, voice, guardrails and approvers. One workspace scales from 1 brand to 50+ without bleed.
Role permissions
Least privilege, by default.
Roles are assigned per brand, not per workspace, so an agency operator on one client account cannot see or touch another. Autonomy level is itself a permission — only an admin can loosen the reins.
- SAML 2.0 SSO
- SCIM provisioning
- Custom roles
- Per-brand scoping
- Session policy
91% of mid-market firms now run AI, but only 12% have a governance framework. The control plane is that framework — shipped, not assembled.
| Capability | Viewer | Operator | Admin |
|---|---|---|---|
| View campaigns & reports | ✓ | ✓ | ✓ |
| Export audit trail | ✓ | ✓ | ✓ |
| Launch campaigns | — | ✓ | ✓ |
| Approve agent actions | — | ✓ | ✓ |
| Change budgets | — | — | ✓ |
| Set autonomy level | — | — | ✓ |
| Manage roles & SSO | — | — | ✓ |
| Trigger kill switch | — | ✓ | ✓ |
Rollout
Start in Full Control. Loosen only where it earns trust.
No enterprise adopts autonomous marketing on day one, and we do not ask you to. The standard rollout tightens nothing and proves everything.
| Phase | Mode | What happens |
|---|---|---|
| Weeks 1–2 | Full control | Agents draft only. Your team reviews every action and calibrates guardrails against real work. |
| Weeks 3–6 | Guided autonomy | Low-risk modules run inside guardrails. Paid spend and brand changes stay gated. |
| Week 7+ | Your choice | Expand autonomy module by module using the audit trail as evidence. Roll back any time. |
Security review questions
What your reviewers will ask.
Can we stop the AI mid-action?
Yes. The fleet-wide kill switch halts every agent immediately and rolls back in-flight actions. It is available to Operator and Admin roles, requires no support ticket, and the halt itself is written to the audit trail.
What exactly is recorded in the audit trail?
Every agent proposal, every human approval or hold with the approver's identity, every policy-engine block with the rule that fired, and every applied change with the object it touched. Records are immutable and exportable for evidence packs.
Is governance limited to higher tiers?
No. Approval gates, audit trail, RBAC and the kill switch ship on every plan including the entry tier. Charging separately for security controls stalls procurement, so we do not.
How is data isolated between brands and tenants?
The platform is multi-tenant with database-level tenant separation, and within a tenant each brand carries its own data, voice, guardrails and approvers. Roles are scoped per brand, so access to one never implies access to another.
What compliance documentation is available?
SOC 2 Type I report, GDPR documentation with a signable DPA, the sub-processor list, encryption standards (AES-256 at rest, TLS 1.2+ in transit). Request the full pack from the security page.
“The question is not whether AI can write the email. It is who signed off before it went to forty thousand people.”
Why governance is the product