Skip to main content

How to Create Custom Assistant

Build specialized AI assistants tailored to your specific business needs by creating a custom assistant from scratch with your preferred settings.

Overview​

Custom assistants allow you to create AI agents with specific configurations, provider services, and AI models that perfectly match your business requirements and use cases.

Video Walkthrough​

Watch a step-by-step video guide for creating a custom assistant:

When to Create Custom Assistants​

Choose custom creation when you need:

  • Specific AI model capabilities
  • Particular provider services
  • Unique configuration requirements
  • Specialized business logic
  • Custom integration needs

Step-by-Step Process​

Step 1: Start Custom Assistant Creation​

Click on the '+ Custom Assistant' button to begin creating a new assistant with personalized configurations and settings.

Before you start:

  • Define your assistant's purpose
  • Identify required capabilities
  • Consider integration requirements

Step 2: Select Assistant Type​

Choose your preferred assistant type from the available options that best matches your specific business requirements and use case.

Available types:

  • Voice Assistant: Optimized for phone and voice interactions
  • Chat Assistant: Designed for text-based conversations
  • Hybrid Assistant: Supports both voice and text communications
  • Specialized Assistant: Industry-specific configurations

Step 3: Choose Provider Services​

Select from the available provider services that will power your assistant's capabilities and determine its underlying infrastructure.

Provider considerations:

  • Performance requirements: Response time and accuracy
  • Compliance needs: Data security and privacy requirements
  • Cost factors: Pricing models and usage limits
  • Feature availability: Specific capabilities offered

Step 4: Select AI Model​

Choose the specific AI model from your selected provider that will handle your assistant's responses and reasoning capabilities.

Model selection factors:

  • Use case alignment: Match model strengths to your needs
  • Language support: Ensure proper language capabilities
  • Context handling: Consider conversation memory requirements
  • Specialization: Domain-specific model training

Step 5: Create Assistant​

Click the 'Create Assistant' button to finalize your custom assistant with all your selected configurations and bring it to life.

Interactive Demo​

Advanced Configuration Options​

Assistant Personality​

  • Tone and style: Professional, friendly, casual
  • Response length: Concise or detailed explanations
  • Language patterns: Formal or conversational

Capabilities Configuration​

  • Knowledge access: Available information sources
  • Action permissions: What the assistant can do
  • Integration scope: Connected services and APIs

Performance Settings​

  • Response time: Optimization for speed vs. accuracy
  • Conversation memory: How much context to retain
  • Error handling: Fallback behaviors and escalation

Provider and Model Guide​

ProviderStrengthsBest For
OpenAIGeneral intelligence, reasoningVersatile applications
AnthropicSafety, long conversationsDetailed interactions
GoogleMultimodal, search integrationInformation-heavy tasks
CustomSpecialized modelsIndustry-specific needs

Model Selection Tips​

  • GPT-4: Best for complex reasoning and analysis
  • GPT-3.5: Faster responses, cost-effective
  • Claude: Excellent for long-form content
  • Specialized models: Industry or task-specific optimization

Best Practices​

Planning Phase​

  1. Define clear objectives for your assistant
  2. Map out typical conversations and scenarios
  3. Identify key performance metrics to track
  4. Plan integration requirements early

Configuration Phase​

  1. Start with proven models for your use case
  2. Test different provider options if available
  3. Configure conservative permissions initially
  4. Plan for gradual capability expansion

Testing Phase​

  1. Test with diverse scenarios before deployment
  2. Verify all integrations work as expected
  3. Check performance under expected load
  4. Validate security and compliance requirements

Troubleshooting​

Common Configuration Issues​

Model not responding as expected:

  • Review selected model capabilities
  • Check configuration parameters
  • Verify provider service status

Performance issues:

  • Consider model optimization settings
  • Review provider service limits
  • Check network connectivity

Integration failures:

  • Verify API configurations
  • Check authentication settings
  • Test connection endpoints

Next Steps​

After creating your custom assistant:

  1. Create appropriate roles for your assistant
  2. Assign to channels for customer access
  3. Test thoroughly before production deployment
  4. Monitor and optimize based on real usage