Automations

Automations are visual pipelines that automate repeatable campaign operations. They let you build multi-step processes — with AI nodes, approval gates, platform integrations, and data transformations — without writing code. Every automation execution is subject to the same governance controls as any other platform action.

Open Automations in the sidebar, under the Operate group.

What Automations Are For

While the AI assistant handles many actions through governed conversations, automations are the right tool when you need:

• Repeatability — the same steps run on a schedule or in response to an event
• Multi-step orchestration — chain AI decisions, data lookups, approvals, and platform actions together
• Human checkpoints — approval gate nodes pause execution and route to the Queue for review
• Auditability — every step is logged, with full execution history per run

The Automation Editor

Open any automation from the Automations page, or create one with Create automation.

The editor shows:

• A canvas — drag and drop nodes, connect them with edges
• A node palette — all available node types organized by category
• An execution panel — run history, logs, and step-by-step output

Node Categories

Nodes are the building blocks of an automation. Full reference: Node Reference.

Building Your First Automation: Campaign Review Pipeline

This example builds an automation that analyses campaign performance and routes a budget recommendation through an approval gate.

Step 1: Create the Automation

1. On the Automations page, click Create automation
2. The editor opens with an empty canvas

Step 2: Add an LLM Node

1. Drag LLM from the AI category onto the canvas
2. Configure the prompt: "Analyse the attached campaign performance data and recommend a budget adjustment if warranted"
3. Connect your data entry node to the LLM input

Step 3: Add an Approval Gate

1. Drag Approval Gate from the Governance category onto the canvas
2. Connect the LLM output to the Approval Gate input
3. Configure:
   • Reviewer role — who can approve (e.g., Editor and above)
   • Timeout — how long to wait before auto-rejecting

When this automation runs, execution pauses at the Approval Gate. The pending action appears in the Operations Queue with the full LLM output as the manifest.

Step 4: Add an Integration Node

1. Drag Integration from the Integrations category onto the canvas
2. Connect the Approval Gate "approved" output to the Integration node
3. Select the platform action (e.g., update campaign budget in Google Ads)
4. Map the LLM recommendation to the action parameters

Step 5: Add a Notify Node

1. Drag Notify from the Output category onto the canvas
2. Connect both the "approved" and "rejected" paths
3. Configure notification message and recipients

Step 6: Run the Automation

1. Click Run in the toolbar
2. The automation executes until it reaches the Approval Gate
3. Open the Operations Queue to review and approve
4. After approval, execution resumes and the integration action fires

Governance Controls

Every automation execution:

• Evaluates active guardrails at each integration action
• Records every step in the Audit Log
• Respects active pauses — an integration action inside a pause's scope does not run, unless the pause routes it to human review and someone approves it
• Scores confidence on AI-generated actions before presenting them for approval

Template Syntax

Automation nodes use a template syntax for variable interpolation in prompts and configurations. See Template Syntax for the full reference.