Skip to main contentAI Agents Overview
Agents are specialized AI assistants configured for specific purposes.
Agent Types
Department Assistants
- Quality Assistant: Helps quality team
- Safety Assistant: Supports safety incidents
- Maintenance Assistant: Equipment troubleshooting
- Logistics Assistant: Supply chain support
Role-Specific Assistants
- Manager Assistant: Dashboard and insights
- Executive Assistant: Strategic summaries
- Operator Assistant: On-the-floor support
Creating Agents
- Admin Console → AI Agents
- Click “Create Agent”
- Configure:
- Name: What department/role
- Purpose: What this agent does
- System Prompt: Agent personality/behavior
- LLM Provider: Which AI model
- Available Tools: What can agent access
- Knowledge Base: Documents to reference
- Test with sample queries
- Enable for team
Agent Configuration
System Prompt
Instructions for how agent should behave:
Example: “You are a Quality Assistant helping manufacturing teams. Focus on defect patterns, root causes, and prevention. Be direct and actionable.”
Choose what agent can do:
- Search claims: Query claims with filters
- Get claim details: Read full claim info
- View analytics: Access metrics
- Check SLA status: Query deadlines
- Access knowledge base: Reference procedures
- Query team workload: See capacity
Knowledge Base Link
Attach relevant documentation:
- Standard procedures
- Troubleshooting guides
- Historical resolutions
- Best practices
- Equipment manuals
Managing Agents
Enabling/Disabling
- Enable for specific departments
- Disable if not useful
- Archive unused agents
Monitoring Usage
- How often used
- User satisfaction
- Common queries
- Knowledge base hits
Updating Agents
- Refine system prompt
- Add new tools
- Update knowledge base
- Change LLM provider
Agent Permissions
Control what agents can access:
- Own department claims only
- All claims in organization
- Specific categories
- Historical data only
Best Practices
- Start with one agent: Test before expanding
- Clear purpose: Each agent has specific role
- Good documentation: Rich knowledge base
- Regular testing: Verify responses accurate
- User feedback: Adjust based on usage
- Periodic reviews: Update when processes change
Next Steps
