The Challenge

Email remains a critical channel for receiving commercial requests in many organizations, but it is also one of the main sources of operational inefficiency.

In a company operating in the integrated logistics and transport organization sector, a significant portion of these requests arrived as unstructured emails, requiring manual analysis to:

• Identify relevant information;
• Create opportunities in Salesforce;
• Prepare commercial proposals.

This process involved a high level of operational effort and limited the team’s ability to scale, as it required manual analysis and handling of requests before proper routing.

The challenge was to automate the interpretation of these requests and convert them into structured, actionable data within Salesforce, while ensuring accuracy and consistency.

The Solution

To address this challenge, worldIT developed an Artificial Intelligence agent built on Salesforce's native capabilities, leveraging Prompt Builder.

Specialized prompt templates were designed to:

• Interpret the content of incoming emails;
• Extract only the relevant information;
• Structure the data required for the automatic creation of opportunities and proposals.

During the initial phase, it became evident that an approach based on a single AI interaction did not provide consistently reliable results.

As a result, a more robust architecture was designed, based on:

• Interdependent prompt chains, each with distinct responsibilities throughout the processing workflow;
• Data normalization mechanisms, ensuring compliance with Salesforce requirements;
• Embedded guiding examples within prompts, aimed at reducing ambiguity and improving response accuracy.

The focus of the solution was to ensure not only automation, but also control, consistency, and effective integration with existing business processes.

Implementation Strategy

The adopted approach was built upon four fundamental principles:

Modular Information Processing

Request interpretation is carried out in stages, with each step responsible for a specific function: analysis, extraction, normalization, and data preparation.

Normalization as an Operational Prerequisite

Dedicated prompts were implemented to normalize data, ensuring that formats, naming conventions, and data structures align with Salesforce automation requirements.

Contextual Guidance for the Model

The inclusion of examples within prompts helped guide the AI's behavior, reducing variability and improving the quality of the outputs generated.

Control Mechanisms and Exception Handling

Apex classes and triggers were developed to validate the processed information.

In cases where the AI does not achieve adequate confidence levels, requests are automatically routed to the back-office team, ensuring operational continuity.

This approach provides a balance between automation and control, mitigating the risks associated with the use of AI.

Results

The implementation of this solution delivered significant operational benefits:

• Approximately 80% of reference cases processed fully autonomously;
• Significant reduction in the manual effort associated with handling requests;
• Faster creation of commercial opportunities;
• Improved customer response times;
• Greater consistency and quality of data recorded in Salesforce.

Additionally, internal resources were effectively freed up, allowing teams to focus on higher-value activities.

Key Learnings

This project demonstrated that the true value of Artificial Intelligence lies not only in its automation capabilities, but in how it is integrated into organizational processes.

The main critical success factors included:

• Integration with well-defined business rules;
• Implementation of validation and control mechanisms;
• Clear workflows for exception handling.

The solution enabled the establishment of an effective collaboration model between Artificial Intelligence and human intervention, maximizing efficiency without compromising reliability.

Conclusion

The implementation of this AI agent within a company operating in the integrated logistics and transport organization sector demonstrates how traditionally manual processes can be transformed into intelligent, scalable, and results-driven operational workflows.

By combining Salesforce, Prompt Builder, and a carefully designed architecture, it was possible to convert unstructured requests into commercial opportunities in an automated, consistent, and controlled manner.

At worldIT, we believe that the true impact of Artificial Intelligence comes from its practical application to real business challenges, always supported by a solid foundation of processes, technology, and governance.