Overview

By using a dedicated AI to track 'digital body language,' you can eliminate guesswork and allow MoEngage to automatically optimize communications for the best results. This establishes the specific business goals and operational guardrails the AI needs to autonomously determine the most effective message and timing for every individual in your audience.

This article guides you through creating a new Campaign Decisioning Agent, managing existing agents, and understanding how they optimize your user communications.

Terms to Know

Refer to the glossary below for information on the various Decisioning Agent-related entities:

Create

Creating an Agent is the process of configuring a specialized AI "brain" by defining its business objectives, success metrics (Rewards), target audience, and operational constraints.

To create an agent, perform the following steps:

• On the sidebar menu in MoEngage, hover over the Decisioning menu item . The Decisioning menu appears.
• Click Campaign Decisioning.
  
• On the Campaign Decisioning page, click + Create agent.
  
  

The process includes two steps:

• Step 1: Basic Details and Reward Classification
• Step 2: Content (Targeting and Optimization)

Step 1: Basic Details and Reward Classification

In this step, you define the Agent's identity and its success metrics (Rewards).

1. In the Basic details section:
   

   • Agent name: Enter a unique name to identify the business objective (for example, "Resell Agent").

   • Purpose: Briefly explain what this Agent is intended to achieve (for example, "Maximize engagement for dormant users").  

2. In the Reward classification section, define the Intent Signals the AI uses to optimize its decisions.
3. To modify these signals, click the Pencil icon in the upper-right corner.
   
   
4. In the Edit Event Group dialog box, perform the following:

   • Switch between tabs: Click the Positive events or Negative events tab to manage specific signals.

   • Select an event: In the events list, click the specific user action you want the AI to track (for example, WhatsApp Message Clicked).

   • Define severity: In the Classification list, click the severity level to define the event's importance (options include Good, Very Good, Bad, or Very Bad).

   • Add events: Click + Add Event to include additional signals. You can add up to 10 events.

   • Remove events: Click the Trash icon  next to any event row to delete it from the classification logic.

5. Click Save to update the logic.
6. Click Next to proceed to Step 2. 
   

Step 2: Content

Define the operational boundaries and the AI algorithm's behavior.

1. In the Eligibility and Exclusion section, you can select the target audience for your offering using the following segmentation filters:
   All Users Filter by User Property Filter by User Behavior Filter by User Affinity Filter by Custom Segment

   This filter option makes the offering eligible for all users. No further filtering can be applied.

   

   You can target users based on their properties (attributes) such as first seen, city, or subscription plan. Click the User property tab and perform the following:

   1. Select the desired attribute in the list to enable the relevant filters. For example, numeric attributes can have filter conditions such as 'is equal to', 'is not equal to', 'is between', and so on, while string attributes have different filters.
   2. Enter the filter value(s) to create the desired user segment.
   3. Click + Nested Filter to add a filter to create nested AND or OR conditions. (This is optional)

   For more information, refer to Create Segments.

   

   You can target users based on actions they have performed (events), such as app opened or viewed product. Click the User Behavior tab and perform the following:

   1. Select the execution condition and the desired user activity (event).
   2. Enter the event frequency and time duration for the event execution.
   3. Click + Attributes to add an event attribute with a filter condition.
   4. Enter the filter value(s) to create a rule.

   5. Use the + Aggregation button to add the aggregation functionality. For more information, refer to Aggregation Operations.

   6. Click + Nested Filter to add a filter to create nested AND or OR conditions. (This is optional)

   For more information, refer to Create Segments.

   

   You can target groups of users based on their affinity for certain attribute values. For more information, refer to User Affinity. Click the User Affinity tab and perform the following:

   1. Select the desired user event or attribute.
   2. Select the user affinity parameter from the following:
      • Predominantly: Select an affinity attribute with a filter condition and provide the affinity values.
      • For a minimum of: Select an affinity attribute with a filter condition and provide the affinity values.
      • Most no. of times: Enter a value or use the arrows to provide the top percentage of users.
      • Least no. of times: Enter a value or use the arrows to provide the bottom percentage of users.
   3. Enter or use the arrows to provide the time duration for the event execution.
   4. Click + Attribute to add more attributes to the filter.
   5. Click + Nested Filter to add a filter to create nested AND or OR conditions. (This is optional)

   For more information, refer to Create Segments.

   

   You can target a pre-built custom segment.

   1. Select a custom segment as the target audience from the list of custom segments available.
   2. Click + Nested Filter to add a filter to create nested AND or OR conditions. (This is optional)

   For more information, refer to Custom Segments.

   

2. Optimization Preference: Adjust the slider to define how the AI balances learning versus performance:

   • High Exploration: The AI tries many options to learn what works best. This is ideal for new campaigns.

   • Low/No Exploration: The AI relies on existing data to maximize results. This is ideal for established campaigns.

3. Agent Distribution Control: Set frequency guardrails to prevent over-messaging (for example, "Send a maximum of 1 communication in 1 Day").
4. Agent Control Group: Turn this toggle on to set aside a random sample of users (for example, 10%) who will not receive communications. This serves as a baseline to measure the AI's actual impact.
5. Click Create to publish your Agent.

Manage Your Agents

The Campaign Decisioning page lists all agents you have created. After creating the agent, navigate to Decisioning > Campaign Decisioning.

On this page, you can:

• Search and filter: Use the search by agent name box to find an agent by name or use the Select Status list to filter by Active, Paused, or Stopped.

• Pause and resume: Click the ellipsis (⋮) icon and select Pause to stop the Agent's decisioning logic temporarily.

• Edit: Modify the Agent's configuration.