The Social Engagement Agent helps businesses automate engagement moderation across social channels. By setting up triage rules, prompts, and goals, the agent can intelligently decide how to handle incoming engagements—reducing manual effort, speeding up responses, and ensuring brand consistency.

Navigation

Click ‘Settings.’ Go to ‘AI agents’ and select ‘Social engagement agent.’

Goals

You’ll view the Goals tab. These are predefined by the system and highlight what the agent is designed to achieve:

• Reduce response times by automatically replying to public messages within minutes—not hours—across all social platforms.
• Grow the percentage of auto-handled messages to increase coverage and ensure no actionable message is missed.
• Prevent spam messages by detecting and filtering out irrelevant or harmful content, keeping your feeds clean and focused on genuine interactions.

Trigger

The Trigger tab shows when the Engagement Agent runs.

• The agent is triggered automatically whenever any public message type is received.

Tasks

Go to the ‘Tasks’ tab to configure the tasks the agent will perform.

Task 1: Analyze the message to detect spam or assess intent, sentiment, and urgency.
 

Detect Spam

• Analyze the received message to determine whether it is spam. If spam, the pipeline stops here
• If not spam, the agent continues with the next prompts, which we will explore below.

Intent Classifier (view only)

Click ‘Intent classifier’ to preview the intent types the agent considers (e.g., questions, complaints/feedback, appreciation, lead/sales inquiry, off-topic, etc.).

Sentiment Classifier (view only)

Click ‘Sentiment classifier’ to see how the agent labels sentiment (e.g., positive, negative, neutral, and unclassified).

Urgency Classifier (view only)

Click ‘Urgency classifier’ to see the urgency scale (e.g., critical, high, medium, and low).

Task 2: Triage incoming message - Route & prioritize.

Click ‘Engagement triage’ to configure.

Click ‘Add new rule.’

Enter a rule name.

Select the conditions if the rule is based on location, message type, sentiment, intent, urgency, or marked as spam. Since this rule is created based on sentiment level, select ‘Sentiment.’

Select the sentiment level from the dropdown, for example, choose ‘Unclassified.’

Select the action, for example, choose ‘Reply not required.’ In this case, if any public message is classified with sentiment as Unclassified, the agent will not respond to it.

Click ‘Save’ to apply and activate the rule.

Task 3: Generate Reply

The Generate reply task allows the Engagement agent to automatically create and send on-brand responses to public messages using prompts.

Click ‘Add prompt’ at the bottom left corner to create a new prompt.

In the Prompt text, provide clear guidance for the agent. For example:

• Thank the customer warmly and reinforce brand values.
• Apologize politely, offer to help, and direct them to DM for details.
• Answer FAQs about pricing or hours in a concise, friendly tone.

Click ‘Add’ to save the prompt.

Hover over the prompt to edit or delete the prompt.

Run Test

The Run test option allows you to simulate how the Engagement agent processes a public message and what actions it would perform based on your configured tasks, so you can validate conditions and responses before applying them broadly.

In the ‘Preview’ section, click the gear icon ⚙️

From the list, select a sample public message and click ‘Run test.’

After running the test, the agent analyzed the selected message and classified it as Compliment/Appreciation, with a Positive sentiment and Low urgency.

The message did not meet any of the defined engagement triage rules. As a result, no triage actions (such as routing or marking as spam) were applied, and the system moved directly to generating a reply.

The agent then moved to the next step — Generate reply — and crafted a response based on the detected intent (Appreciation), sentiment (Positive), and urgency (Low). The suggested reply included a brief, warm acknowledgment with an emoji, matching the positive tone of the message.

Once you’re satisfied with the test, click the ‘Publish’ button to save the configuration.

Outcomes
 

The Outcomes tab provides insights into how well the Social Engagement agent is performing. These metrics are directly tied to the goals of the agent and help track whether the outcomes are being achieved.

Average reply time – Average time it took to reply to a public message via the agent. This time will include response generation time & response publishing time (including approval delays, scheduling delays, and manual publishing for suggestions).

% Messages auto-handled by agent – % of incoming unique public messages auto-handled (responded, triaged, or closed) by the agent. Hover over it to view more details, such as total messages, reply suggested, and replied.

% Messages marked as spam – % of public messages marked spam either by spam detection tool and via mark as spam action in triage task. Hover over it to view more details, such as total messages, the number of messages detected via the spam detection tool, and the triage task.

Click the time range dropdown (default is Past 30 days). Select from preset options, such as yesterday, the past 7 days, this month, the past 3 months, the past 6 months, this year, the past 12 months, or set a custom range. The metrics in the Outcomes tab will automatically update to reflect the selected period.