Overview

Every conversation in Sparrow Intel comes with a free analysis layer — sentiment, action items, detected issues, intent category, and relevant knowledge — generated by Chirp AI as messages arrive. All of it lives in the right-hand context panel.

This lesson covers what each section means, how to use it, and how it powers the rest of the platform (Predictions, Rules, Tasks, Autopilot).

Why this matters

Conversations is the highest-signal place to learn what's actually happening in your operation. A guest mentioning "the AC has been off all night" is more important — and more actionable — than any review or survey response, if you can act on it before it ends the stay badly.

The AI sidebar is what makes that signal visible. Without it, you'd have to read every message looking for clues. With it, the clues come pre-tagged.

Sentiment

Two sentiment values appear:

• Post sentiment — the latest message read in isolation
• Conversation sentiment — the trend across the whole thread

Both are color-coded (green for positive, yellow for neutral, red for negative, grey for mixed). The conversation-level value is what most rules and Autopilot decisions use, because it reflects how the relationship is trending — not just whether the latest message was nice.

A conversation can have a positive most-recent message but a negative overall sentiment if the thread had problems earlier. That's the right behavior — guests remember the whole stay.

Action items

Action items are specific things the guest is asking you to do (or asking about):

"Can we get extra towels?"

"Is late check-out possible?"

"What's the wifi password?"

Each one shows up as a discrete bullet on the conversation. Action items help you avoid the most common reply mistake: answering one question and missing the second.

If you see three action items in the sidebar, your reply should address three things — or explicitly defer one.

Cleaning issues and maintenance issues

These are operational, not conversational. Chirp AI extracts them when a guest mentions:

• Cleaning issues: "the bathroom wasn't cleaned," "we found hair in the bed," "the kitchen was sticky"
• Maintenance issues: "the dishwasher won't run," "AC is making a clicking noise," "the porch light is out"

Each detected issue becomes an actionable item that you can:

• Read and address in your reply
• Convert into a task in Breezeway, Guesty, Hostaway, Track, Asana, or via webhook
• Auto-route to the right team via a Rule

We cover the task creation flow in Creating Tasks from Conversations. The point here is just: those issues are already extracted for you, so you don't have to spot them by hand.

Intent category

Every guest post is classified into a category — what the message is fundamentally about. Default categories include:

• Routine Inquiry
• Check-in Logistics
• Issue Report
• Complaint
• Booking Modification
• Emergency
• Billing
• Compliment

Categories are useful for rules and Autopilot. For example, you can configure Autopilot to handle Routine Inquiry and Compliment automatically, while always leaving Complaints and Emergencies for a human.

Admins can add custom categories in Settings — see Admin Setup Checklist.

Knowledge snippets

In the same panel you'll often see relevant knowledge snippets — short, reusable pieces of property or policy information that Chirp AI surfaces because they're relevant to the conversation.

Snippets answer questions like:

• "What's the parking policy at the Lakehouse?"
• "Is the hot tub available year-round?"
• "What's the late-checkout fee structure?"

When a guest asks something a snippet covers, you'll see the snippet appear in the sidebar — and Chirp AI will use it when drafting any reply suggestion. That's how AI replies stay grounded in your facts instead of being generic.

How the sidebar drives the rest of the platform

Everything in the sidebar is also rule-eligible:

• A rule can fire when conversation sentiment turns negative
• A rule can create a task whenever a maintenance issue is detected
• A rule can auto-assign all "Emergency" intent posts to a supervisor

It's also the foundation of Predictions — Chirp's prediction of how a guest is likely to review the stay relies heavily on the conversation analysis. The same sidebar data that helps you reply better also feeds the system that protects your public reputation.

What if the sidebar got something wrong?

Chirp is good but not infallible. Occasional misclassification happens, especially on short or ambiguous messages.

If you notice a pattern (e.g., a particular type of message consistently misclassified), let our support team know via the in-product chat. We use real conversations to tune the model.

In the meantime, you can:

• Override sentiment by adding a label
• Add a knowledge snippet that disambiguates
• Edit the AI's reply suggestion before sending

Up next

You've finished the Cruising Altitude phase. Time to move into Power Maneuvers — features that turn an organized inbox into a system that works for you.

Next up: Knowledge Snippets: Teaching Sparrow Your Way.