Salesforce Service Copilot

About the Agent

The ZBrain Service Copilot for Salesforce is an AI-powered assistant designed to streamline customer support operations. Seamlessly integrated into the Salesforce Service Console, the agent automates case analysis, retrieves relevant historical data, and generates intelligent responses. By classifying queries into general inquiries, follow-ups, or knowledge article creation requests, it ensures efficient routing and faster resolutions while minimizing manual effort. The agent also dynamically creates and references knowledge articles, enhancing consistency and efficiency across support interactions.

Challenges the Agent Addresses

• Slow Case Resolution: Eliminates the need for manual data retrieval by instantly surfacing relevant case details and historical insights.
• Fragmented Information: Consolidates data from multiple sources, providing customer support teams with a complete and contextual view.
• Inefficient Knowledge Management: Automates the creation and retrieval of knowledge articles, ensuring quick access to accurate information and improving decision-making efficiency.
• Manual Query Handling: Classifies and processes queries intelligently, ensuring accurate and efficient responses.
• Repetitive Troubleshooting: Retrieves prior case details and responses for follow-up queries, reducing redundant efforts and improving resolution consistency.

How the Agent Works

The ZBrain Service Copilot follows a multi-step process to deliver real-time, data-driven insights:

Step 1: Webhook Trigger

The agent is immediately activated when a query is entered into the Salesforce Service Console chat interface.

Key Tasks:

• Monitors the chat interface for incoming queries.
• Capture query details along with the associated case context.

Outcome:

• A comprehensive case dataset is created, enabling accurate assessment and efficient resolution.

Step 2: Input Collection & Parsing

The agent collects all relevant inputs to build a comprehensive understanding of the query context.

Key Tasks:

• Gather JSON-formatted case details, user text inputs, and any previous conversation history.
• Utilize a prompt instructing the LLM to analyze the incoming query and return a JSON response to categorize and route the query accurately.

Outcome:

• The system secures a structured and complete dataset that accurately represents the current query and its context, enabling precise processing.

Step 3: Query Type Determination & Routing

The agent classifies the incoming query to determine the appropriate handling route.

Key Tasks:

• Use custom JavaScript to process the JSON response from the LLM and extract the value.
• Route the query based on its type:
  • Follow Up: Append the previous conversation context to provide continuity.
  • General/Normal Query: Process the request using current case details and data fetched from the knowledge base.
  • KB Query: Trigger the knowledge article creation branch for documenting the case resolution.

Outcome:

• Queries are accurately categorized and efficiently routed, ensuring that each query follows the correct processing pathway for optimal response generation.

Step 4: Application of Conversational Guardrails

The agent applies predefined conversational guidelines to maintain response quality and consistency.

Key Tasks:

• Leverage guardrails to ensure responses meet quality, compliance, and clarity standards.
• Reference the existing knowledge base as needed to support accurate and contextually relevant answers.

Outcome:

• The agent delivers responses that are not only accurate but also adhere to set communication standards, enhancing overall user experience.

Step 5: Branch Execution

Distinct processing flows are executed based on the determined query type.

Key Tasks:

• For follow-up queries, use complete historical conversation data for coherent follow-up responses.
• For general queries, analyze current case details and reference the knowledge base to generate accurate, context-aware responses.
• For KB creation requests, the agent first checks the existing knowledge base. If an article already exists, it retrieves and shares the link in the chat. If not, it triggers an HTTP call to generate a new knowledge article in the preferred document editing tool (Google Docs, Word, or Confluence), incorporating complete case details and resolution steps. The knowledge base is then updated, and the newly created article’s URL is instantly provided in the chat.

Outcome:

• Each query is processed through its specific branch, ensuring a tailored, effective resolution, whether it’s a follow-up, general inquiry, or a request to create a knowledge article.

Step 6: Response Composition & Delivery

The agent finalizes the process by formatting and delivering the response back to the user.

Key Tasks:

• Compile the processed information from the executed branch.
• Format the final response for clarity and coherence.
• Deliver the response via an HTTP call to the Salesforce Service Console chat interface.

Outcome:

• The support agent receives a well-structured, context-aware response promptly, significantly enhancing the efficiency and effectiveness of case resolution.

Why use ZBrain Service Copilot for Salesforce?

• Enhanced Efficiency: Real-time case summaries and actionable insights reduce the time spent manually retrieving and analyzing data.
• Consistent Resolutions: By referencing existing knowledge articles and automatically generating new ones, the agent promotes uniformity in case resolutions.
• Context-aware Support: The ability to process follow-up queries with historical context leads to more coherent and accurate responses.
• Reduced Manual Effort: Automated parsing, routing, and knowledge article creation minimize repetitive tasks, allowing agents to focus on resolving cases.
• Seamless Salesforce Integration: Operates directly within the Salesforce Service Console, eliminating the need to switch between multiple systems.
• Scalability: The AI-driven workflow adapts to high case volumes and evolving support requirements.