Sentiment Detection uses AI to analyze the emotional tone of customer messages. When a customer is frustrated, angry, or upset, your team can be alerted immediately.
How It Works
Customer sends a message
AI analyzes the emotional tone
If a concerning sentiment is detected, staff is notified
Staff can respond appropriately
Example:
Customer: “This is ridiculous! I’ve been waiting 3 days and STILL no response!!”
AI detects: Frustrated / Angry
Staff gets pinged: “⚠️ Frustrated customer in #ticket-12345”
Detected Sentiments Sentiment Description Typical Signs Angry Strong negative emotion ALL CAPS, profanity, threats Frustrated Persistent dissatisfaction Repeated complaints, exasperation Upset General unhappiness Disappointment, sadness Confused Uncertainty or misunderstanding Questions, “I don’t understand” Happy Positive emotions Thank yous, exclamations of joy Neutral No strong emotion Factual statements
Getting Started
Enable Sentiment Detection
Go to AI Features → Sentiment Detection and toggle it on.
Set Up Actions
Configure what happens when each sentiment is detected.
Assign Roles
Choose which roles get notified for each sentiment.
Tone Actions For each sentiment, you can configure an action:
Action Description Ping Role Notify a specific role when detected Do Nothing Track the sentiment but don’t notify
Example Setup Sentiment Action Role Angry Ping Role @Senior Support Frustrated Ping Role @Support Team Upset Ping Role @Support Team Confused Do Nothing - Happy Do Nothing - Neutral Do Nothing -
Only ping for negative sentiments to avoid notification fatigue.
Settings Analysis Mode When to analyze sentiment:
Mode Description Best For First Message Analyze immediately Quick detection First 5 Messages Wait for more context Better accuracy Smart Detection AI decides timing Recommended
Smart Detection waits if:
Staff is actively responding
More context would help
Message is ambiguous
Delay Before Analysis Seconds to wait before analyzing:
0 - Immediate15-30 - Let user finish typing60+ - Wait for multi-message context
What AI Looks For Language Signals Signal Sentiment ALL CAPS Angry/Frustrated Exclamation marks!!! Strong emotion Question marks??? Confused Thank you, great! Happy ”I don’t understand” Confused Profanity Angry
Context Clues
Repeated issues → FrustratedWaiting mentions → Frustrated/UpsetThreats to leave → AngryCompliments → HappyNeutral facts → Neutral
Combination Detection AI considers multiple signals together:
"This is the THIRD time I'm asking!!" ↓ Repeated issue (frustrated) + CAPS (angry) + !! (strong emotion) ↓ Result: Frustrated (primary) + Angry (secondary)
How Staff Sees It When concerning sentiment is detected:
⚠️ Sentiment Alert: Frustrated Customer Ticket: #ticket-12345 Customer: John Smith Sentiment: Frustrated (87% confidence) "I've been waiting for 3 days with no response"
Staff can:
Click to open the ticket
See the confidence level
Review the triggering message
Best Practices Respond Appropriately When alerted to negative sentiment:
Do:
Acknowledge their frustration
Apologize for the experience
Offer a solution quickly
Escalate if needed
Don’t:
Be defensive
Ignore the sentiment
Use generic responses
Keep them waiting longer
Train Your Team Help staff handle different sentiments:
Angry → Stay calm, empathize, escalate if neededFrustrated → Acknowledge, explain, resolve quicklyConfused → Clarify, guide step-by-stepUpset → Empathize, offer solution
Combine with Other Features Sentiment Detection works well with:
Priority Detection: Angry + High Priority → Immediate escalation
Auto-Routing: Frustrated customers → Senior support
Knowledge Base: Adjust AI tone when sentiment is negative
Troubleshooting Not Detecting Sentiment
Check if enabled - Verify Sentiment Detection is onCheck analysis mode - “Smart Detection” may waitCheck message content - Very short messages may not have enough signals
Wrong Sentiment Detected
Context matters - Sarcasm and irony are hard to detectCultural differences - Some expressions vary by cultureAmbiguous messages - Short messages may be misread
Too Many Alerts
Adjust actions - Set some sentiments to “Do Nothing”Increase delay - Wait longer before analyzingUse Smart Detection - Let AI decide when to alert
Missed Angry Customers
Lower sensitivity - Some expressions are subtleCheck action config - Make sure “Angry” has an actionCheck role config - Verify the notification role exists
Privacy Note Sentiment Detection analyzes message content to detect emotional tone. It does not:
Store sentiment history long-term
Share data with third parties
Use data for purposes other than detection
Need Help? Get instant support via private DM
