Skip to main content

AI-Powered Classification

DZDESK uses artificial intelligence to automatically classify and categorize support requests.

Overview​

When a new request is created, DZDESK's AI analyzes the content and automatically suggests:

  • Category classification
  • Priority level
  • Relevant tags
  • Potential assignee group

How It Works​

Text Analysis​

The AI examines:

  • Request title
  • Description text
  • Keywords and phrases
  • Contextual patterns

Classification Process​

1. Request submitted
2. AI analyzes content
3. Suggestions generated
4. User reviews/accepts
5. Request created with classification

Category Classification​

Automatic Category Suggestions​

Based on content, AI suggests categories:

KeywordsSuggested Category
"laptop", "computer", "printer"Hardware
"outlook", "excel", "application"Software
"wifi", "internet", "VPN"Network
"password", "login", "access"Access Request

Confidence Levels​

AI provides confidence scores:

ConfidenceAction
High (>90%)Auto-apply (configurable)
Medium (60-90%)Suggest with highlight
Low (<60%)Show as option

User Override​

Users can always:

  • Accept the suggestion
  • Choose different category
  • Ignore suggestions

Priority Suggestion​

Priority Analysis​

AI considers:

  • Urgency keywords ("urgent", "critical", "asap")
  • Impact indicators ("all users", "production")
  • Business context
  • Historical patterns

Example Triggers​

ContentSuggested Priority
"System is down"Critical
"All users affected"High
"When you have time"Low
"Cannot complete my work"Medium-High

Tag Suggestions​

Automatic Tagging​

AI suggests relevant tags:

  • Based on content keywords
  • Historical similar requests
  • Common tag patterns

Example Tags​

ContentSuggested Tags
"Outlook crashes"email, outlook, crash
"New employee laptop"new-hire, hardware, setup
"VPN from home"vpn, remote, network

Group Routing​

Smart Assignment​

AI can suggest assignment to:

  • Appropriate support group
  • Specialists with relevant experience
  • Available team members

Routing Logic​

Based on:

  • Category classification
  • Historical assignment patterns
  • Current workload
  • Expertise matching

Configuration​

Enabling AI Classification​

  1. Go to Settings > AI Features
  2. Enable Auto-Classification
  3. Configure options

Configuration Options​

SettingDescription
Auto-apply thresholdConfidence level for auto-apply
Show suggestionsAlways/High confidence only
Categories enabledWhich categories to suggest
Priority suggestionsEnable/disable

Per-Category Settings​

Configure AI behavior per category:

  • Enable/disable suggestions
  • Set confidence thresholds
  • Define keyword patterns

Training the AI​

Learning from Corrections​

The AI learns when:

  • Users correct classifications
  • Patterns are reinforced
  • New categories added

Feedback Loop​

1. AI makes suggestion
2. User corrects (if needed)
3. System records correction
4. AI improves over time

Manual Training​

Administrators can:

  • Add keyword associations
  • Define category rules
  • Review classification accuracy

Accuracy and Improvement​

Measuring Accuracy​

Track classification performance:

  • Acceptance rate
  • Correction rate
  • Accuracy by category

Continuous Improvement​

AI improves through:

  • More training data
  • User feedback
  • Regular model updates

Best Practices​

For Users​

  1. Write clear titles
  2. Include relevant details
  3. Review AI suggestions
  4. Correct when needed

For Administrators​

  1. Monitor accuracy metrics
  2. Review common corrections
  3. Update category definitions
  4. Provide feedback data

Privacy Considerations​

Data Usage​

AI classification:

  • Processes text within tenant
  • Does not share across tenants
  • Follows data privacy policies

Sensitive Content​

  • AI does not expose sensitive data
  • Classifications are generic
  • Personal data not used for training

Limitations​

When AI May Struggle​

  • Very short descriptions
  • Unusual or new issues
  • Ambiguous requests
  • Multiple issues in one request

Human Oversight​

Always recommended:

  • Review suggestions
  • Verify accuracy
  • Apply human judgment