The Agent Copilot is built on a modular and scalable architecture. It is designed to manage user interactions and deliver intelligent responses using a combination of AI services and Sprinklr capabilities. This section provides a detailed overview of the core components and how they work together to power the Agent Copilot. 

1. Support Agent Interaction 

When a support agent interacts with the Agent Copilot, the input is routed to the Orchestrator, which serves as the central processing unit. It receives queries or bot-triggered messages and initiates the task-handling workflow. 

2. Orchestrator 

The Orchestrator is the brain that coordinates the Agent Copilot’s behavior. It is responsible for interpreting the input, breaking it down into actionable tasks, and managing the entire workflow from start to finish. It acts as the control hub that coordinates all other components, including the Data Layer, Skills Engine, Unified Data Layer, and AI providers. 

It performs the following actions in real-time: 

• Receives and parses incoming agent or system-generated messages. 

• Matches tasks with appropriate skills, based on request context. 

• Coordinates with the Data Layer to extract context and apply masking. 

• Executes functions using skills and references data from the Unified Data Layer. 

• Adapts by integrating with various AI providers and custom models for maximum flexibility. 

3. Skills Engine 

Skills are modular micro-capabilities used to execute tasks. Each task can invoke one or more skills—such as updating a case, generating a summary, or triggering an API call—based on the user’s intent and task logic. 

4. Data Layer 

The Data Layer manages all incoming request data. It extracts relevant context, applies PII (Personally Identifiable Information) masking where needed, and prepares the input for task execution. This promotes secure and context-aware processing of every request. 

5. Unified Data Layer 

• The Unified Data Layer provides the Orchestrator with structured access to Sprinklr’s capabilities and user data such as Knowledge Base articles and Universal Case data. Integrates Universal Case functionality for customer support use cases. Supplies user-specific information for personalized interactions. 

6. Integration with AI Providers 

Agent Copilot supports seamless integration with available third-party AI providers (see Sprinklr’s Subprocessor List) which enables enhanced natural language understanding, reasoning, and content generation capabilities. To connect with these providers, you can use Sprinklr-provided keys or your own keys using BYOK (Bring Your Own Key) enablement type. You can also integrate custom AI models using BYOM (Bring Your Own Model) enablement type that supports function calling. This allows you to build domain-specific intelligence into your Agent Copilot for highly tailored use cases. 

System Workflow: From Interaction to Response 

The architecture follows a streamlined flow for handling every user query: 

1. Agent Interaction: The request is received by the Orchestrator. 

2. Task Management: Each query is routed to a task based on query and task matching.   

3. Skills Execution: Copilot skills can be executed within a task to generate the best suited response. 

4. Data Processing: Contextual information is extracted, and personal data is masked in accordance with the PII masking template configured. 

5. Response Generation: AI providers or custom models are invoked to formulate intelligent responses. 

6. Unified Data Access: Data is retrieved from Sprinklr data layer for enriched and accurate output. 

The Agent Copilot architecture is a flexible, modular system designed to deliver intelligent, secure, and efficient task execution. By organizing skills, using contextual data, and integrating with industry-leading AI models, the platform provides high-quality user experiences tailored to your organization’s needs.