Crescendo AI is an intelligent segment-building assistant that interprets your business intent and converts natural language descriptions into structured filter conditions. Unlike traditional keyword searches, AI Segment accurately understands business requirements, conducts multi-turn conversations to clarify intent, and provides clear, explainable segmentation logic.
We recommend reviewing the "Tutorial｜Segment" article first to understand the basic concepts and use cases of segments.
Target Audience
- Users who need to quickly build segments based on complex business conditions
- Marketing managers who want to leverage AI to discover target audiences
- Analysts looking to uncover hidden insights within customer data
Steps
Step 1: Navigate to the Segment Creation Page
In the MAAC dashboard, go to:
Audience > Segment > [Create]
In the pop-up window, select "AI Segment" and click [Create] to enter the segment setup page.
On the "Create LINE Segment" page, you will see fields for segment name, segment criteria, and segment exclusion.
Step 2: Open the Crescendo AI Panel
On the "Create LINE Segment" page, locate the "Segment criteria" section. Under "Filter," click the purple "Describe your segment" button.
The Crescendo AI conversation panel will appear on the right side.
Step 3: Enter a Segment Description (Prompt)
When the Crescendo AI panel opens, it displays a welcome message: "Hi, what kind of segment would you like to create today?"
You can:
-
Select a suggested prompt: The panel provides 4 common segment scenarios, such as:
- "AI-predicted high-potential buyers who haven't purchased in the last 30 days"
- "Customers interested in specific products or campaigns"
- "Customers who clicked on a recent promotion"
- "Customers with no interaction in the past 90 days"
- Enter a custom description: Type your business requirements directly in the input field (supports Traditional Chinese, Japanese, English, and Thai)
Step 4: AI Intent Clarification and Multi-Turn Conversation
After you submit a segment description, the AI analyzes your requirements. Depending on complexity, the AI may ask clarifying questions, such as:
- "What do you mean by 'high-spending customers'? What spending threshold do you have in mind?"
- "Would you like to exclude customers with return records?"
- "Which tags would you like to include to identify 'cat lovers'?"
You can select from the options the AI provides or continue adding details. The AI will refine the segment definition based on your responses.
Step 5: Review and Apply the AI Filter Plan
Once the AI understands your requirements, it displays a "Segment Generator" listing all structured filter conditions, for example:
✓ Total spending > 100,000
✓ Total orders > 20
✓ No purchases in the last 30 days
Each condition is clearly readable so you can verify the AI's analysis logic. Once confirmed, click the "Apply Segment" button.
Step 6: Set Segment Exclusion (Optional)
Back on the main page, you can exclude existing segments (up to 2) under the "Segment Exclusion" section.
Click the [+ Add] button to add exclusion conditions.
Use segment exclusion to improve targeting accuracy - for example, excluding customers who have already purchased or are on a blocklist.
Step 7: Name and Create the Segment
Enter an easily identifiable name in the "Segment name" field (required).
Optionally, add a description in the "Description" field for future reference.
Once confirmed, click [Create] to finalize the segment.
AI Segment calculation is an asynchronous process that typically takes 10-15 minutes. Once complete, the segment list is a static list that you can update at any time.
Supported Filter Conditions
Crescendo AI can automatically translate your descriptions into the following filter condition types:
| Category | Supported Conditions | Notes |
|---|---|---|
| Member Attributes | Gender, age, location, membership tier, etc. | - |
| Tags | MAAC tags, CAAC tags (automatically filtered by context) | - |
| Contact Notes (New) | MAAC notes, CAAC notes keyword search | New feature |
| Purchase Metrics (Expanded) | Average order value, total spending, highest single order, order count | New: total spending, highest single order, order count |
| Product Metrics (Expanded) | Total product amount, purchase quantity, highest product price, add-to-cart count, checkout count | New: purchase quantity, highest product price, add-to-cart, checkout count |
| Engagement Behavior (New) | Message engagement: open, click on broadcast messages, coupon claim/own/redeem, across LINE, SMS, WhatsApp/Email channels | New feature supporting multi-channel message engagement tracking (90-day lookback) |
| Purchase Prediction | Predicted Purchase Probability (PPP) | - |
| Last Interaction | Last X days, X months | - |
| Reachability | LINE subscription status, push notification consent, etc. | - |
| Other | Join date, birthday, language preference, etc. | - |
New and Expanded Features
1. Contact Notes Filter (New)
The AI can now search MAAC notes and CAAC notes by keyword, helping you quickly identify specific audiences based on customer service records or internal tags.
Example: "Find customers whose notes mention 'VIP' or 'high-value'"
2. Purchase Metrics Expansion
In addition to the existing "Average order value," the AI now supports:
- Total spending: Cumulative spending amount
- Highest single order: Highest individual transaction amount
- Order count: Number of purchases
Example: "Find loyal customers with total spending over 1,000,000 and more than 10 purchases"
3. Product Metrics Expansion
The AI now includes the following product-related filters:
- Purchase quantity: Number of product units purchased
- Highest product price: Most expensive product purchased
- Add-to-cart count: Items added to cart but not purchased
- Checkout count: Number of completed orders
Example: "Find customers with cart abandonment behavior for a re-engagement campaign"
4. Engagement Behavior Tracking (New)
The AI now supports message engagement filters across 6 channel types (90-day lookback):
- LINE Broadcast: Open, click
- LINE Retargeting Messages: Open, click
- Coupons / Prizes: Claim, own, redeem
- SMS Broadcast: Click
- SMS Plus: Status update
- Omnichannel Broadcast (WhatsApp, Email): Open, click
Example: "Find customers who clicked on a promotional broadcast in the last 30 days" or "Find customers who claimed a coupon but haven't redeemed it"
Conversation Examples
Example 1: E-Commerce VIP Night Event
User input: "I want to find high-spending, loyal customers for a VIP night event"
AI clarifying questions:
- "What do you mean by 'high-spending'? What total spending threshold?"
- User selects: "Over 100,000"
- AI: "Would you like to exclude customers who had returns or complaints in the past 3 months?"
- User selects: "Yes"
AI-generated segment conditions:
✓ Total spending ≥ 100,000
✓ Order count ≥ 5
✓ Exclude customers with returns in the past 3 months
Example 2: Pet Supplies Brand
User input: "Find customers interested in cat products"
AI clarifying questions:
- AI searches relevant tags and suggests: "I found the tags 'Cat Lover' and 'Pet Supplies Buyer.' Would you like to include them?"
- User selects: "Include all"
- AI: "Would you like to add 'cart abandonment' behavior? This can help re-engage unconverted prospects."
- User selects: "Yes, add it"
AI-generated segment conditions:
✓ Includes tags: 'Cat Lover' or 'Pet Supplies Buyer'
✓ Last purchase: Within the past 180 days
✓ Add-to-cart count > 0 (cart abandonment behavior)
Example 3: Churned Customer Re-Engagement
User input: "Find churned high-value customers"
AI clarifying questions:
- "How do you define 'high-value'? Please provide a spending threshold."
- User: "Past spending over 500,000"
- AI: "How long without a purchase counts as 'churned'?"
- User: "No purchases for over 6 months"
AI-generated segment conditions:
✓ Historical total spending ≥ 500,000
✓ Last purchase: 6 months ago or earlier
✓ No interaction in the past 30 days
Example 4: Out-of-Scope Request
User input: "I want to find customers who frequently shop on Instagram and browse via web"
AI response: "Sorry, I'm currently unable to integrate third-party platforms (such as Instagram shopping) or web browsing behavior, as this data is not linked to your Crescendo system. However, I suggest:
- Using 'clicked on an Instagram post' as an engagement signal instead
- Combining this with 'web visitor' behavioral filters
Certain external data sources (such as third-party social media shopping platforms) may not be directly integrable. The AI will proactively notify you and suggest alternatives.
Important Notes
- Multi-turn conversation: The AI may ask multiple clarifying questions. Please answer carefully to ensure accurate segment definitions.
- Asynchronous calculation: After segment creation, the AI calculates matching contacts in the background. This typically takes 10-15 minutes.
- Static list updates: AI Segments are created as static lists. To update, you can re-edit or delete and recreate the segment.
- Language support: Currently supports Traditional Chinese, Japanese, English, and Thai. If you use an unsupported language, the AI will respond in English.
- Automatic tag filtering: The AI automatically determines MAAC or CAAC tags based on context. You can manually adjust in the "Segment Generator."
- Segment exclusion limit: You can set up to 2 segment exclusions.
- Ad platform sync: AI Segments support syncing to ad platforms. After enabling "Sync to ad platforms," the segment will be available in your ad platform.
FAQ
Q1: What is the difference between AI Segment and manual filtering?
A: AI Segment uses intent recognition technology to understand complex business requirements and automatically convert them into structured filter conditions. Manual filtering requires you to select each condition individually. AI Segment is faster and more intuitive, especially for complex segmentation scenarios.
Q2: How long does it take to create an AI Segment?
A: After segment creation, the system calculates the number of matching contacts. This typically takes 10-15 minutes, depending on your data volume.
Q3: How can I verify that the AI understood my requirements correctly?
A: In the "Segment Generator," you can view all filter conditions the AI has generated. Review each condition to confirm it matches your intent. If any condition is incorrect, click [Edit] to adjust.
Q4: Can I modify an existing AI Segment?
A: Yes. On the segment management page, find the segment and click [Edit]. You can reopen the Crescendo AI panel to make adjustments or manually edit the filter conditions.
Q5: What should I do if the AI cannot understand my request?
A: If the AI cannot understand or the request is out of scope, it will proactively notify you and suggest alternatives. You can:
- Rephrase your description using more specific numbers or conditions
- Break complex requests into multiple simpler segments
- Switch to manual filtering mode
Q6: Does AI Segment support multiple languages?
A: Yes. It currently supports Traditional Chinese, Japanese, English, and Thai. If you use an unsupported language, the AI will respond in English.
Q7: Can I sync a segment to ad platforms after creation?
A: Yes. On the segment creation page, enable the "Sync to ad platforms" option. The segment list will automatically sync to your connected ad platform.
Related Articles
- Tutorial｜Segment - Learn the basic concepts, use cases, and manual filtering methods for segments
- Tutorial｜Ad Platform Integration - Learn how to sync segments to your ad platform
Article version: 26.Q1 release
Last updated: April 2026
Applies to: Crescendo v3.0 and above