What It Does
AI Sentiment Analysis automatically reads the tone of incoming conversations and applies sentiment tags in Help Scout. This helps your team quickly spot upset customers and prioritize accordingly.
How to Enable
- Make sure your AI provider is configured at DeskPress > AI Agent > AI API (you need a valid API key).
- Go to DeskPress > AI Agent > AI Drafts and scroll to the Sentiment Analysis section.
- Toggle Enable Sentiment to on.
Sentiment analysis will now run automatically on new conversations processed by the AI pipeline.
Configuration Walkthrough
Step 1: Turn It On
Navigate to DeskPress > AI Agent > AI Drafts and find the Sentiment Analysis section. Flip the Enable Sentiment toggle to on.
Step 2: Customize Your Tag Names (Optional)
By default, the system uses these tag names:
sentiment-angrysentiment-frustratedsentiment-happysentiment-neutral
You can change these to match your existing tag structure in Help Scout. For example, if your team already uses tags like priority-high or mood-positive, you could align the sentiment tags to follow the same naming convention.
Step 3: Create Views in Help Scout (Recommended)
To get the most out of sentiment analysis, set up custom views in Help Scout that filter by these tags. For example:
- Create a view called "Urgent - Angry Customers" filtered by the
sentiment-angrytag - Create a view called "Needs Attention" filtered by the
sentiment-frustratedtag
This way, your team can quickly see which conversations need immediate attention.
How It Works
Here's what happens behind the scenes:
- When a new conversation arrives (or is processed by the AI pipeline), the system takes the first 1,000 characters of the message.
- The AI analyzes the text and classifies it into one of four sentiment categories: Angry, Frustrated, Happy, or Neutral.
- The corresponding tag is automatically added to the conversation in Help Scout.
- If the conversation already has a sentiment tag from a previous analysis, the old tag is removed before the new one is applied. This keeps things clean -- each conversation only ever has one sentiment tag.
Sentiment analysis runs alongside AI Draft generation. Both features work together seamlessly.
Settings Reference
| Setting | Default | Description |
|---|---|---|
| Enable Sentiment | Off | Master toggle to activate sentiment analysis |
| Angry Tag | sentiment-angry | Tag name applied to conversations with an angry tone |
| Frustrated Tag | sentiment-frustrated | Tag name applied to conversations with a frustrated tone |
| Happy Tag | sentiment-happy | Tag name applied to conversations with a happy tone |
| Neutral Tag | sentiment-neutral | Tag name applied to conversations with a neutral tone |
Tips and Common Questions
Does this use a lot of AI tokens?
No. Sentiment analysis is very lightweight. It only processes the first 1,000 characters of a conversation, so the token usage per analysis is minimal.
Can I disable specific sentiment categories?
Not individually -- sentiment analysis is either on or off. However, you can ignore specific tags in Help Scout by simply not creating views or workflows for them.
What if a conversation changes tone over time?
Sentiment is analyzed when the conversation is first processed. It reflects the tone of the initial message. If a customer's tone changes during the conversation, the tag won't automatically update.
Does it work without AI Draft Answers?
Sentiment analysis runs as part of the same AI processing pipeline. It requires the AI provider to be configured and conversations to be processed through the pipeline. It works alongside AI Drafts but analyzes sentiment independently.
How accurate is it?
The AI does a good job distinguishing between clearly angry, frustrated, happy, and neutral messages. Edge cases (sarcasm, mixed emotions) may occasionally be categorized as neutral. Overall, it's a reliable triage tool for prioritizing your queue.
Can I use sentiment tags in Help Scout workflows?
Yes. Since sentiment tags are standard Help Scout tags, you can use them in any Help Scout workflow, auto-reply, or notification rule. For example, you could set up a workflow that assigns conversations tagged sentiment-angry to a senior team member.