Skip to main content

Deepdesk CX Assistants

What are Deepdesk Assistants?​

Deepdesk Assistants are AI-powered tools that enhance customer service interactions by automating routine tasks, providing knowledge assistance, and supporting human agents with relevant information and capabilities. These assistants leverage large language models (LLMs) to understand context, interpret customer queries, and generate appropriate responses or actions.

Assistants can be configured to handle a variety of tasks, from answering common customer questions to executing complex workflows that integrate with external systems. Each assistant is designed with specific instructions, tools, and response formats to fulfill particular business needs within your customer service ecosystem.

Benefits and Use Cases​

Key Benefits​

  • Increased Efficiency: Automate routine tasks and queries, freeing human agents to focus on complex issues
  • Consistent Experience: Ensure standardized responses across all customer interactions
  • 24/7 Availability: Provide immediate responses to customers at any time
  • Scalability: Handle fluctuating volumes of customer inquiries without additional staffing
  • Knowledge Integration: Connect with your existing knowledge bases to provide accurate, up-to-date information
  • Reduced Training Time: Help new agents get up to speed faster with AI-assisted guidance
  • Data-Driven Insights: Gather consistent data on customer inquiries and pain points

Common Use Cases​

  1. Customer Query Resolution: Automatically answer common questions using knowledge base integration
  2. Agent Assistance: Provide real-time suggestions and information to human agents during conversations
  3. Data Retrieval: Pull customer account information, order status, or product details from backend systems
  4. Process Automation: Guide customers through standardized processes like returns, bookings, or account changes
  5. Multi-System Integration: Connect to various APIs to perform actions across different business systems
  6. Conversation Summarization: Create concise summaries of customer interactions for records and handoffs
  7. Sentiment Analysis: Identify customer emotions and escalate negative experiences to human agents
  8. Form Filling: Extract relevant information from conversations to complete forms or tickets

How Assistants Fit into the Deepdesk Ecosystem​

Deepdesk Assistants are a core component of the broader Deepdesk platform, which is designed to enhance customer service operations. The ecosystem includes:

  • Conversation Platform: Manages customer interactions across various channels (chat, email, voice)
  • Agent Workspace: Where human agents handle customer inquiries with AI assistance
  • Knowledge Management: Stores and organizes information that assistants can access
  • Analytics Dashboard: Tracks performance metrics for both assistants and human agents
  • Administration Portal: Configures assistants, profiles, and system settings

Within this ecosystem, assistants can be deployed in several ways:

  1. Assistants: Perform specific tasks when called directly or by operators or other assistants
  2. Operator Assistants: Orchestrate workflows between multiple specialized assistants
  3. Direct Customer-Facing Assistants: Interact directly with customers for specific use cases
  4. Agent-Facing Assistants: Provide information and suggestions to human agents

Key Concepts and Terminology​

Assistant Components​

  • Instructions: The primary guidance that defines an assistant's purpose and behavior
  • Tools: Capabilities that extend an assistant's functionality, such as API calls or knowledge base searches
  • Parameters: Variables that can be passed to assistants to modify their behavior in specific contexts
  • Response Formats: Structured ways assistants can return information (Plain Text, Assistant Cue, Knowledge Assist)

Conversation Management​

  • Threads: Persistent conversation contexts that maintain history across multiple interactions
  • Threadless Processing: Single-turn interactions where all context is provided at once
  • Conversation Events: Triggers like new messages, conversation acceptance, or conversation ending

Configuration Elements​

  • Profiles: Settings that determine which assistants are triggered for specific types of conversations
  • Routes: Rules that determine when assistants should be automatically invoked
  • Assistant User Groups: Collections of users who can access and use specific assistants

Architecture​

Deepdesk uses a hierarchical architecture to manage assistant interactions:

This architecture enables:

  1. Event-Driven Automation: Conversation events trigger operator assistants
  2. Delegation and Specialization: Operators delegate to specialized child assistants
  3. Flexibility: Child assistants can call other child assistants or tools as needed
  4. Aggregation: Results flow back up the chain, with each level processing and refining the information
  5. External Integration: Tools connect assistants to external systems and data sources

The flow typically begins when a conversation event (like a new message) triggers an operator assistant. The operator evaluates the conversation and may call one or more child assistants to handle specific aspects of the request. These child assistants may call additional assistants or use tools to access external resources. Results from these operations flow back up the chain, with each assistant processing the information before passing it to its parent. Ultimately, the operator assistant aggregates all responses and returns a unified result to the interface.

This architecture allows for highly customizable and complex workflows while maintaining a clear organization of responsibilities throughout the system.

Features​