Sentiment Analysis

Understand how your customers feel. AI-powered sentiment analysis detects emotional tone in every message, helping you respond appropriately.

What Is Sentiment Analysis?

Sentiment analysis uses AI to determine the emotional tone of customer messages:

• Positive: Happy, satisfied, grateful
• Neutral: Informational, matter-of-fact
• Negative: Frustrated, angry, disappointed

Visual Indicators

Where to See Sentiment

AI Sidebar

Open any ticket to see overall sentiment in the AI Insights panel.

Per-Message Indicators

Individual messages can show sentiment badges indicating mood for that specific message.

Ticket List

Tickets with negative sentiment may show warning indicators.

Dashboard

Analytics show sentiment trends across your support.

Confidence Scores

Sentiment includes a confidence percentage:

Example: "Negative (85%)" means the AI is quite confident the customer is unhappy.

How Sentiment Is Detected

What AI Analyzes

• Word choice ("frustrated" vs "curious")
• Punctuation patterns (!!!!)
• Capitalization (ALL CAPS)
• Phrases ("I've been waiting forever")
• Context from conversation history

Per-Message vs Overall

• Per-message: Each message analyzed individually
• Overall: Latest sentiment + conversation trajectory

Sentiment Changes

Sentiment updates throughout the conversation:

Message 1: "Where's my order?" → Neutral
Message 2: "It's been 2 weeks!" → Negative
Message 3: "Thanks for the update" → Positive

Using Sentiment Effectively

Calibrate Your Tone

Match your response to customer sentiment:

Prioritize Negative Sentiment

Negative sentiment often indicates:

• Urgent issues
• Risk of churn
• Need for escalation
• Potential public complaint

Monitor Sentiment Shifts

Watch for:

• Positive → Negative: Something went wrong
• Negative → Positive: Your resolution worked
• Stable negative: May need escalation

Sentiment-Based Workflows

Automated Routing

Create automations based on sentiment:

Example automation:

• Trigger: Ticket sentiment is Negative AND confidence > 80%
• Action: Add tag "needs-attention" and notify manager

SLA Adjustments

Consider faster response for negative sentiment tickets.

Escalation Triggers

Auto-escalate when sentiment is highly negative + VIP customer.

Responding to Different Sentiments

Positive Sentiment

Customer feels good. Maintain the relationship:

• Thank them for positive feedback
• Maintain friendly tone
• Ask for reviews if appropriate
• Resolve efficiently

Neutral Sentiment

Customer is task-focused. Be helpful:

• Answer questions directly
• Provide clear information
• Don't over-apologize
• Keep it professional

Negative Sentiment

Customer is frustrated. Turn it around:

• Acknowledge their frustration
• Apologize sincerely
• Take ownership
• Provide clear resolution path
• Follow up to confirm satisfaction

Sentiment Analytics

Team Dashboard

View sentiment metrics in Analytics:

• Overall sentiment distribution
• Sentiment trends over time
• Sentiment by agent
• Sentiment by ticket type

Useful Insights

• Rising negative sentiment: Process or product issue?
• Agent sentiment scores: Training opportunities?
• Sentiment by category: Which issues cause frustration?

Accuracy & Limitations

Generally Accurate For

• Clear emotional language
• Standard support scenarios
• English language
• Written communication

May Misread

• Sarcasm ("Great, another delay")
• Technical discussions (neutral misread as negative)
• Cultural differences
• Very brief messages

When to Override

If sentiment seems wrong:

• Trust your judgment
• Consider the full context
• AI provides guidance, not certainty

Tips for Using Sentiment

1. Check sentiment first: Before responding, know the mood
2. Don't over-rely: It's a signal, not gospel
3. Track trends: Individual tickets + overall patterns
4. Respond appropriately: Match your tone to theirs
5. Celebrate improvements: Note when you turn negative to positive

Troubleshooting

"Sentiment says positive but customer seems upset"

• AI may have caught different signals
• Sarcasm can be misread
• Trust your reading of the situation

"Sentiment not showing"

• Very short messages may not analyze
• Check AI settings
• Try refreshing the ticket

"Sentiment changes unexpectedly"

• Based on latest message
• May aggregate entire conversation
• Check which message triggered change

Privacy Note

Sentiment analysis:

• Processes ticket content only
• Does not store emotional profiles
• Used only for this ticket's context
• Complies with privacy regulations

← Reply Suggestions | Priority Prediction →