Policies

Policies are the preventive controls the agent reads before it acts. Open Policies in the sidebar, under the Governance group. Each kind of policy has a page of its own, opened from Create policy on the Policies page. This page covers the three quality-and-compliance pages that run on your internal state; execution controls, pauses, escalation policies, and spend limits are documented separately.

Blocking Rules (Guardrails)

Blocking rules are JSONLogic guardrails that evaluate the content of every proposed action. They check things like:

• Budget limits (is the proposed spend within envelope?)
• Targeting constraints (are excluded audiences respected?)
• Brand safety rules (does the creative meet your standards?)
• Structural requirements (are required fields present?)

Each guardrail has a severity — advisory (record only), warning (record for review), or error (block and hold for review). Blocked actions appear in the Queue.

→ Guardrails

Naming Conventions

The Naming page defines the patterns your campaigns, ad sets, and ads must follow. The AI validates names before execution and flags violations.

Configure:

• Naming templates with required tokens (e.g., {market}_{product}_{date})
• Default values for optional tokens
• Which entity types (campaign, ad set, ad) each template applies to

Navigate to Policies > Naming to manage naming standards.

Launch QA

Launch QA checklists are readiness packs — structured lists of checks that must pass before a campaign goes live. Use them to enforce your team's pre-launch review process.

Configure:

• Checklist templates with required and optional items
• Scope each checklist to specific action types or platforms
• Run QA checks manually or attach them to approval gates in automations

Navigate to Policies > Launch QA to manage checklist templates and review QA runs.

Other Policy Pages

These pages also live under Policies and are documented elsewhere:

• Execution controls — how much the agent can do without asking, and the confidence minimums and data-freshness requirements → Safeguards
• Pauses and Escalation policies — emergency stops, and pulling a human into a running execution → Safeguards
• Spend limits — budget envelopes and spend caps → Budget