Skip to main content

Managing Answer Filters

Once youโ€™ve created Answer Filters, youโ€™ll need to maintain them over time. This guide covers how to view, edit, update, and delete filters, as well as best practices for ongoing management.

Accessing Your Answer Filters

All Answer Filters for an agent are managed from the Answer Filters tab in the agent dashboard.
1

Navigate to Agent

Go to Projects โ†’ Select your project โ†’ Select your agent
2

Open Answer Filters Tab

Click on the Answer Filters tab in the agent navigation
3

View Answer Filters List

Youโ€™ll see a table listing all Answer Filters for this agent

Answer Filters List View

The filters table displays:
ColumnDescription
IDUnique identifier for the filter
QueriesPreview of query examples (truncated if long)
Created atDate and time the filter was created
ActionsDropdown menu with edit and delete options
Admin Access Required: Only project Admins and Owners can view and manage Answer Filters.

Viewing Filter Details

To see the complete details of a filter:
1

Locate the Filter

Find the filter in the filters list table
2

Open Actions Menu

Click the โ‹ฎ (three dots) icon in the Actions column
3

Select Edit

Choose โ€œEditโ€ from the dropdown menu
4

Review Details

The filter sheet opens showing all queries, good responses, and bad responses
Hover over truncated query text in the table to see the full text in a tooltip.

Editing Answer Filters

When to Edit a Filter

You should edit a filter when:
Scenario: Company policy or information has been updatedExample:
Old: "We offer a 14-day return policy"
New: "We offer a 30-day return policy"
Action: Update good responses with new information
Scenario: Users ask questions that should trigger the filter but donโ€™tExample:
Existing queries:
- "What is your return policy?"
- "Can I return items?"

Users also ask:
- "How do I send something back?"
- "What's your refund process?"
Action: Add these new query variations
Scenario: Agent makes a mistake that wasnโ€™t anticipatedExample:
Agent incorrectly says: "Refunds take 2-3 weeks"
Correct information: "Refunds take 5-7 business days"
Action: Add the incorrect response to bad responses
Scenario: Good responses arenโ€™t as effective as hopedExample:
Vague: "We have flexible return options"
Better: "Returns are accepted within 30 days for full 
        refund with original packaging and receipt"
Action: Refine good responses to be more specific

How to Edit a Filter

1

Open Edit Sheet

  1. Navigate to the Answer Filters tab
  2. Click the โ‹ฎ icon next to the filter
  3. Select โ€œEditโ€ from the dropdown
The edit sheet opens with current filter data pre-filled
2

Modify Queries

Add new queries:
  • Type in the query field
  • Click โ€+ Addโ€ or press Enter
Remove queries:
  • Click the [x] icon next to any query
Edit existing queries:
  • Currently requires removing and re-adding
  • Or use the text field to modify
3

Update Good Responses

Add new good responses:
  • Type in the good responses field
  • Click โ€+ Addโ€ or press Enter
Remove responses:
  • Click the [x] icon next to any response
Tips:
  • Be specific and actionable
  • Include all necessary details
  • Ensure information is current
4

Update Bad Responses

Add new bad responses:
  • Type responses to avoid
  • Click โ€+ Addโ€ or press Enter
Remove responses:
  • Click the [x] icon next to any response
Tips:
  • Include newly discovered mistakes
  • List any outdated information
  • Note off-brand phrasings
5

Save Changes

  1. Review all modifications
  2. Click โ€œUpdate Filterโ€ button
  3. Wait for success confirmation
Changes take effect immediately for new conversations

Editing Examples

Example 1: Adding Query Variations

Scenario: Users ask about refunds in ways not covered by current queries.
Queries:
  - "What is your refund policy?"
  - "Can I get a refund?"
  - "Do you offer refunds?"
Problem: Filter doesnโ€™t trigger for:
  • โ€œHow do I get my money back?โ€
  • โ€œCan I return this for a refund?โ€
  • โ€œWhatโ€™s your money-back guarantee?โ€

Example 2: Updating Information

Scenario: Return policy changed from 14 to 30 days.
Good Responses:
  - "We offer a 14-day return policy on all items."
  - "Returns must be requested within 14 days of purchase."
Problem: Information is outdated

Example 3: Preventing New Mistakes

Scenario: Agent started saying โ€œrefunds take 2-3 weeksโ€ when itโ€™s actually 5-7 days.
Bad Responses:
  - "We don't offer refunds."
  - "All sales are final."
Problem: Doesnโ€™t prevent the new mistake

Deleting Answer Filters

When to Delete a Filter

Delete a filter when:

Outdated Topic

The topic is no longer relevant (e.g., ended promotion, discontinued product)

Redundant

Another filter covers the same topic more effectively

No Longer Needed

The issue resolved itself or policy normalized

Causing Issues

Filter is triggering incorrectly or causing problems
Deletion is permanent: Deleted filters cannot be recovered. Consider editing instead of deleting if you might need the filter again.

How to Delete a Filter

1

Locate the Filter

Navigate to the Answer Filters tab and find the filter to delete
2

Open Delete Dialog

  1. Click the โ‹ฎ icon in the Actions column
  2. Select โ€œDeleteโ€ from the dropdown menu
A confirmation dialog appears
3

Confirm Deletion

  1. Review the filter details in the confirmation dialog
  2. Type confirmation text if required
  3. Click โ€œDelete Filterโ€ or โ€œConfirmโ€
This action cannot be undone!
4

Verify Removal

The filter disappears from the list and a success message appears
Filter is immediately removed and no longer affects responses

Maintenance Best Practices

Regular Review Schedule

Establish a routine for reviewing and updating your Answer Filters:
If you have high-volume agents:
  • โœ… Check agent conversations for new query patterns
  • โœ… Look for responses that could be improved
  • โœ… Identify new mistakes to prevent
  • โœ… Note topics that need new filters
Time needed: 15-30 minutes
Pro Tip: Set calendar reminders for your chosen review schedule to ensure filters stay current.

Documentation

Keep a separate document tracking your filters: What to Document:
Filter Documentation Template

Filter ID: [Auto-generated ID]
Topic: Return & Refund Policy
Created: 2024-01-15
Last Updated: 2024-03-20
Owner: Customer Support Team

Purpose:
  Ensure consistent communication of our 30-day return policy

Business Context:
  - Policy changed from 14 to 30 days on Jan 1, 2024
  - High-priority topic for customer satisfaction
  - Legal requirement in some jurisdictions

Update Triggers:
  - Policy changes (review quarterly)
  - Customer complaints about clarity
  - New regional requirements

Related Answer Filters:
  - Shipping Returns (ID: abc123)
  - Warranty Policy (ID: def456)

Notes:
  - Updated on 3/20 to clarify international returns
  - Consider splitting domestic/international in future

Organizing Answer Filters

As your filter library grows, organization becomes important:

Naming Conventions

While filters donโ€™t have visible names in the UI, use consistent patterns in your documentation:
Good Naming:
  - "Returns - General Policy"
  - "Returns - International"
  - "Returns - Damaged Items"
  - "Pricing - Basic Plan"
  - "Pricing - Enterprise"

Bad Naming:
  - "Filter 1"
  - "New filter"
  - "Refunds thing"

Consider how filters relate to each other:
Customer Support:
  - Returns & Refunds
  - Shipping Questions
  - Account Access
  - Order Tracking

Sales:
  - Pricing Information
  - Plan Comparisons
  - Enterprise Inquiries
  - Trial Questions

Technical:
  - Password Reset
  - Integration Help
  - API Questions
  - Troubleshooting

Bulk Management Strategies

Audit Checklist

Use this checklist when reviewing multiple filters:
## Filter Audit Checklist

For each filter, verify:

### Accuracy
- [ ] All information is current and correct
- [ ] Good responses reflect latest policies
- [ ] Bad responses cover known mistakes
- [ ] No contradictions with other filters

### Completeness
- [ ] 3-5 query examples provided
- [ ] Query examples cover main variations
- [ ] Good responses are specific and actionable
- [ ] Bad responses prevent key mistakes

### Relevance
- [ ] Topic is still applicable
- [ ] Filter is still triggering
- [ ] User queries match examples
- [ ] Responses remain helpful

### Quality
- [ ] Good responses are clear and specific
- [ ] Language is brand-aligned
- [ ] Information is complete
- [ ] Tone is appropriate

### Technical
- [ ] Filter triggers at correct threshold
- [ ] No unintended triggering
- [ ] Works across languages (if applicable)
- [ ] Integrates well with other filters

### Actions
- [ ] Update: [What needs updating]
- [ ] Monitor: [What to watch]
- [ ] Delete: [Reason if deleting]
- [ ] Keep as-is: [Confirmed working well]

Consolidation Strategy

When you have too many filters, consider consolidating:
Filter 1: "Returns - Within 14 days"
Queries:
  - "Can I return within 14 days?"
  - "What if it's been less than 2 weeks?"

Filter 2: "Returns - Within 30 days"
Queries:
  - "Can I return within 30 days?"
  - "What if it's been less than a month?"

Filter 3: "Returns - General"
Queries:
  - "What's your return policy?"
  - "Do you accept returns?"
Problems:
  • Redundant coverage
  • Potential conflicts
  • Hard to maintain

Integration with Other Features

Answer Filters work alongside other PLai Framework features:

With System Prompts

Relationship: System prompts provide general behavior, filters provide specific responses
System Prompt:
  "You are a helpful customer service agent. Always be polite
   and professional."

Answer Filter (Refunds):
  "We offer a 30-day return policy..."

Result:
  Agent maintains polite tone (from prompt) while giving
  specific refund information (from filter)

With Datasources

Relationship: Answer Filters provide specific answers, datasources provide broader knowledge
Best practice:
  • Use filters for FAQs and policies
  • Use datasources for detailed documentation
  • Answer Filters take precedence for specific questions
  • Datasources fill gaps filters donโ€™t cover

With Analytics

Relationship: Analytics help you improve filters over time
What to track:
  • Which filters trigger most often
  • User satisfaction after filtered responses
  • Topics that need new filters
  • Filter performance over time

Next Steps


Quick Reference

Filter Management Checklist

Daily (High-Volume Agents):
- [ ] Review flagged conversations
- [ ] Note new query patterns

Weekly:
- [ ] Check for new mistakes to prevent
- [ ] Add query variations as needed
- [ ] Update 1-2 filters based on usage

Monthly:
- [ ] Complete filter audit
- [ ] Update all outdated information
- [ ] Consolidate redundant filters
- [ ] Create new filters for trends

Quarterly:
- [ ] Comprehensive filter review
- [ ] Update documentation
- [ ] Plan filter strategy for next quarter based on agent analytics

Common Actions Quick Guide

TaskSteps
EditAnswer Filters tab โ†’ โ‹ฎ โ†’ Edit โ†’ Modify โ†’ Update Filter
DeleteAnswer Filters tab โ†’ โ‹ฎ โ†’ Delete โ†’ Confirm
ViewAnswer Filters tab โ†’ Find filter in list
TestEdit filter โ†’ Note queries โ†’ Test in chat
DocumentExport filter details โ†’ Add to docs