Genesys Cloud - Main

Hi everyone,

We recently introduced a new approach to proactively identify interactions where customers had a poor experience using Genesys Cloud AI insights.

Previously, we didn't have a structured way to detect these interactions automatically and relied more on manual review.

What we implemented:

• Trigger:

• sentimentOverallScore < -0.4

• sentimentTrendScore <= 0.2

This allows us to focus on interactions that were both negative and not improving, reducing noise from recovering conversations.

• Workflow logic:

• Queue-based filtering using queueId to ensure only our queues are processed

• Ignoring empty queueIds to avoid cross-division noise

• Handling multi-division interactions by using queueId as the source of truth instead of divisionIds

• Enrichment:

• AI-generated summaries

• Agent name lookup

• Trend classification (Declining / Slightly Declining / No Change / Improving)

This has given us the ability to automatically surface interactions where customers had a genuinely poor experience, without relying on manual QA review.

Example notification:

🚨 AI Insight Triggered

Queue: Customer Service Team
Sentiment: 🔴 Negative (-0.893)

Trend: ➡️ No Change (-0.102)

Agent: <Agent Name>

Summary:
<AI insights summary>

Interaction:
https://apps.mypurecloud.com.au/directory/#/analytics/interactions/12345678/admin

Next step:

This can also be used by for evaluating a subset of these interactions using AI Supervisor Copilot to support QA, coaching, and continuous improvement.

Interested to hear how others are using sentiment and trend signals for proactive monitoring or auto-evaluation.

Hello Phaneedra, 

I love this. I would be interested to hear what the rest of the community says also. 

Cheers, 

Hi everyone,

We recently introduced a new approach to proactively identify interactions where customers had a poor experience using Genesys Cloud AI insights.

Previously, we didn't have a structured way to detect these interactions automatically and relied more on manual review.

What we implemented:

• Trigger:

• sentimentOverallScore < -0.4

• sentimentTrendScore <= 0.2

This allows us to focus on interactions that were both negative and not improving, reducing noise from recovering conversations.

• Workflow logic:

• Queue-based filtering using queueId to ensure only our queues are processed

• Ignoring empty queueIds to avoid cross-division noise

• Handling multi-division interactions by using queueId as the source of truth instead of divisionIds

• Enrichment:

• AI-generated summaries

• Agent name lookup

• Trend classification (Declining / Slightly Declining / No Change / Improving)

This has given us the ability to automatically surface interactions where customers had a genuinely poor experience, without relying on manual QA review.

Example notification:

🚨 AI Insight Triggered

Queue: Customer Service Team
Sentiment: 🔴 Negative (-0.893)

Trend: ➡️ No Change (-0.102)

Agent: <Agent Name>

Summary:
<AI insights summary>

Interaction:
https://apps.mypurecloud.com.au/directory/#/analytics/interactions/12345678/admin

Next step:

This can also be used by for evaluating a subset of these interactions using AI Supervisor Copilot to support QA, coaching, and continuous improvement.

Interested to hear how others are using sentiment and trend signals for proactive monitoring or auto-evaluation.