| Audience | Platform admins and data teams, marketers |
| Prerequisites | AI Decisioning overview → |
AI Decisioning (AID) helps teams automate and improve campaign decisions at the individual user level. You define agents that decide who to target and what to optimize for, then add messages so agents can test and deliver different versions of your content. Over time, agents learn which message, timing, and channel perform best for each user.
This guide covers:
- Technical setup (admins and data teams) →
Configure AI Decisioning settings, prepare data for decisioning, review data handling, and connect destinations. - Day-to-day usage for marketers →
Define agents, add and optimize messages, validate delivery, and measure performance.
Technical setup (admins and data teams)
Before marketers can create agents or messages, your workspace must be configured with the right data models and connected destinations.
Most of this setup is completed once per workspace.
Configure workspace and data
| Step | What you’ll do | Article |
|---|---|---|
| Configure settings | Define global eligibility rules, time zones, and channel defaults. | Configuration → |
| Prepare data for AID | Define audience models, structure event data, and prepare tables for decisioning. | Prepare data for AID → |
| Review data handling and privacy | Understand how AI Decisioning uses, processes, and protects customer data. | Data handling and privacy → |
Connect destinations
| Step | What you’ll do | Article |
|---|---|---|
| Connect destinations | Connect your ESP or messaging platform so AID can deliver messages. | Braze, Iterable, or Salesforce Marketing Cloud |
After technical setup is complete, marketers can manage agents and messages directly from AI Decisioning → Agents.
Define agents
Agents define who to target and what to optimize for. Each agent combines an audience, goals, and decisioning rules into a single adaptive campaign.
Create and configure an agent
| Step | What you’ll do | Article |
|---|---|---|
| Create an agent | Define the target audience and campaign goals. | Agents → |
| Configure scheduling | Set send frequency, quiet hours, and blackout dates. | Agents → Configure scheduling |
| Enable Smart Suppression (optional) | Reduce low-impact sends across the entire agent. | Smart Suppression → |
| Add collections (optional) | Define dynamic sets of products or content the agent can recommend. | Collections → |
Most teams start with a single agent focused on one goal (for example, reactivation or purchase) and expand as performance data accumulates.
Add and optimize messages
Once an agent is defined, add messages from your destination and define variants the agent can evaluate and learn from.
Create messages and variants
| Step | What you’ll do | Article |
|---|---|---|
| Add messages | Start from destination templates and connect them to your agent. | Messages → Add a message |
| Create variants | Define multiple versions of fields like subject lines or CTAs. | Messages → Add and manage variants |
| Apply tags and rules | Control variant eligibility and classify creative attributes. | Tags → |
| Review content quality | Identify overlap, gaps, and opportunities to improve creative. | Content analysis and suggestions → |
Messages are sourced from your destination. AI Decisioning does not change your base content—it evaluates variants and controls which version is sent.
Measure and learn
After your agent is running, validate delivery and review performance to refine your strategy over time.
Validate delivery and analyze results
| Step | What you’ll do | Article |
|---|---|---|
| Inspect delivery | Preview sends, confirm eligibility, and troubleshoot issues. | Inspector → |
| Analyze performance | Compare results across agents, messages, variants, and audiences. | Insights → |
Review Insights regularly to adjust goals, suppression thresholds, and creative as agents continue learning.