To address this challenge, an intelligent agent was implemented using Agentforce, an autonomous and proactive solution from Salesforce. This was not a simple chatbot, but a specialized, AI-powered assistant designed to consult retirement case data and provide answers in natural language. This was the first Salesforce intelligent agent of its kind in Latin America.
The agent was built to understand the company's entire business flow, from the initial filing and normalization of a case to contribution collection and the final benefit determination. This allows human advisors to quickly and accurately verify the status of any request.
To ensure accuracy and consistency, its architecture was based on five key elements:
- Initial Context: The agent's role was defined as an expert advisor, providing it with relevant case information, including client data and request details.
- Additional Data: The agent was integrated with Data Cloud, which functions as its single source of truth. This ensures all responses are generated from verified information.
- Logical Conditions: Specific validation rules were established (e.g., "if the case meets conditions X and Y, provide response A; otherwise, respond B").
- Style and Tone Guidelines: The language and vocabulary were defined to ensure all communication aligned with the company's brand voice.
- Final Response Structure: Detailed instructions were provided on how to structure the final output, including greetings, specific data points, and next steps.
The workflow is triggered automatically when a case is marked "In Progress." The system queries the client's data, executes the prompts, and generates a response. This is then formatted into a PDF template and emailed to the client, while the case is flagged for a final review by a human advisor.
