Correlate social media sentiment and customer messages

Workflow purpose

This guide demonstrates how to use Proto's platform to correlate public social media sentiment with the messages your AI agent is already handling in the Inbox. Support volume alone only tells half the story – a spike in tickets about a specific product or policy often has a matching spike in public sentiment. Insight Advisor pulls both signals into a single workflow so analysts get a combined read without exporting data to a spreadsheet first.

Who can benefit from this guide:

Regulators and central banks monitoring public reaction to a policy change
Enterprise CX teams tracking brand sentiment alongside support volume
Communications teams preparing briefings for leadership
Analysts who currently cross-reference Perception and Inbox data by hand

Correlate sentiment and message volume

This workflow lets an analyst ask Insight Advisor for a sentiment-versus-volume view and receive a plain-language correlation summary in chat. See how to correlate social media sentiment and customer messages.

Just starting with the platform?

For initial setup of your AI agent – including workspace setup, agent training, and channel deployment – please start here. This workflow guide assumes your AI agent is set up and ready for more advanced configurations.

1. Create a trigger

Purpose: Start the flow when someone asks Insight Advisor to compare sentiment and message data.

Navigate to the Actions tab in the AI agent settings.
Click “+ Add Trigger” and select the Message Received trigger type.
Name the trigger “Sentiment correlation”.
Add a description such as “Triggers when the user asks to compare social sentiment with customer messages” so the LLM can match the intent.

2. Survey – scope the request

Purpose: Capture the time window and topic before running the correlation.

Add a Survey action titled Correlation scope.
Field date_range – Text – What time period should I look at (e.g. last 30 days)?
Field keyword – Text – Which topic or keyword should I focus on?

3. Pull the two data sources

Purpose: Retrieve sentiment data and message data for the same window.

Add a Send API request action to Perception, filtered by {keyword} and {date_range}. Map the response to sentiment_data.
Add a second Send API request action to your Inbox/Analytics export, filtered the same way. Map the response to message_data.

4. LLM action – correlate the data

Purpose: Turn two raw datasets into one plain-language summary.

Add an LLM action named Correlation summary.
Input text: {sentiment_data} {message_data}.
Output variable: correlation_summary.
Success branch: add a Set chat variable action first – correlation_summary = _.json_parse(correlation_summary) – before any field is read.
System prompt should ask for a headline (one sentence), a trend (rising/falling/flat), and a notable_spike field (string or null).

5. Send the insight

Purpose: Deliver the correlation in chat, ready to paste into a briefing.

Add a Send message action: “{correlation_summary[“headline”]} Trend: {correlation_summary[“trend”]}. {correlation_summary[“notable_spike”]}“

6. Test the flow

Purpose: Confirm the correlation holds up against a known reporting period.

Run the flow against a period you’ve already reported on manually and check the headline matches your own read of the data.
Test with a keyword that has almost no social mentions, to confirm the LLM action still returns a graceful “not enough data” response rather than guessing.
Test with a very wide date range to confirm the API requests don’t time out.

550K