Skip to main content

Assistants

Assistants are the core building blocks of the Deepdesk CX Assistants platform. Each assistant is an AI-powered agent configured with specific instructions, tools, and response formats to perform particular tasks within your customer service workflows.

What is an Assistant?​

An assistant is a configurable AI agent that can:

  • Understand and process natural language input
  • Execute tasks using available tools (API calls, knowledge base searches, etc.)
  • Return responses in specific formats (plain text, structured data, etc.)
  • Maintain conversation context when configured with threads
  • Be called by other assistants or operators to handle specialized tasks

Types of Assistants​

Regular Assistants​

Standard assistants that perform specific tasks when called. Examples include:

  • Customer data lookup
  • Order status checking
  • FAQ answering
  • Sentiment analysis
  • Form filling

Operator Assistants​

Special assistants that orchestrate workflows by:

  • Evaluating incoming conversations
  • Delegating tasks to specialized child assistants
  • Aggregating results from multiple assistants
  • Managing complex multi-step processes

Assistant Roles​

Assistants can be deployed in different roles:

  1. Customer-Facing: Directly interact with customers for specific use cases
  2. Agent-Facing: Provide information and suggestions to human agents
  3. Background: Perform tasks silently without visible output (e.g., logging, analytics)
  4. Workflow: Part of a larger orchestrated process managed by an operator

Core Configuration Elements​

Every assistant requires:

  • Name & Code: Identifiers for the assistant
  • Instructions: Clear guidance on the assistant's purpose and behavior
  • Response Format: How the assistant returns information (Plain Text, Assistant Cue, Knowledge Assist)
  • Model Selection: Which LLM to use (GPT-4, Claude, etc.)

Optional but commonly used:

  • Tools: Extended capabilities like API integrations or knowledge base access
  • Parameters: Dynamic inputs that customize behavior per invocation
  • Thread Management: Whether to maintain conversation history
  • User Groups: Access control for who can use the assistant

When to Use Assistants​

Consider creating an assistant when you need to:

  • Automate a repetitive task that requires understanding natural language
  • Integrate external data sources into your customer service workflow
  • Provide consistent, standardized responses to common scenarios
  • Extract or structure information from conversations
  • Make decisions based on conversation context and customer data

Next Steps​