As part of its modernization and continuous improvement process, ANCAP explored the incorporation of artificial intelligence to enhance the management of case files and administrative documentation.
Through a pilot project, we analyzed different use cases related to document classification, search, and generation. This initiative made it possible not only to measure the technical impact of the solution, but also to estimate its operational benefits, costs, and feasibility within a complex organizational context, with high requirements for traceability, confidentiality, and regulatory compliance.
Technical overview
Client: ANCAP (Administración Nacional de Combustibles, Alcohol y Portland)
Size/location: Public company in Uruguay that operates as a national reference in the production and distribution of alcohol, Portland cement, and fuels, providing services across all regions of the country.
Main objectives:
- Explore the use of artificial intelligence in case management, applying Azure OpenAI models across a set of use cases.
- Evaluate the benefits and costs of incorporating these technologies to support classification, search, and document generation tasks.
Objectives and challenges
The problem faced by the client before adopting the solution:
- Case file headers were created manually, indexing cases by topic, subtopic, or patronymic criteria, identifying key entities such as companies, individuals, agencies, or institutions. The project aims to automate identification and classification using artificial intelligence to improve process reliability.
- The project also seeks to enable data cleansing and normalization of the database.
- Information classification (public, restricted, confidential) was performed by users. Delegating this task to artificial intelligence ensures that decisions are based on content and attachments analysis, increasing traceability and reinforcing compliance with confidentiality policies.
- Users did not use the advanced search functionality due to its technical complexity. To address this, artificial intelligence is proposed to allow queries in natural language, interpreted by the tool to deliver more comprehensive results.
- Resolutions were manually generated from draft resolutions. By automating this process with artificial intelligence, business rules and predefined formats would be applied, reducing manual tasks to result validation only.
Client goals
Through the implementation of artificial intelligence, the client aims to:
- Automate and increase the reliability of processes.
- Improve efficiency and document traceability.
- Reduce manual workload for employees.
- Strengthen confidentiality and policy compliance processes.
- Simplify tools so all users can operate them.
- Reduce validation times.
The solution
The project was planned in three stages. During the preparation phase, the scope was agreed upon, timelines, costs, and efforts were estimated, and the framework was defined.
In the second stage, a pilot project was built and presented to the client for validation.
Based on the pilot results, AI models are refined for progressive improvements, working jointly between the development team and the functional area.
Integration with an MCP Server is proposed.
The tool becomes more accessible by enabling interaction in natural language.
Results achieved
- For thematic and patronymic classification, efficiency improved through the generation of an initial summary.
- Subsequent classification and analysis tasks were simplified.
- Topics, people, companies, and names related to the nature of the case file were correctly identified by the AI, improving the existing database with reasonable suggestions.
- In confidentiality and restricted classification, AI showed strong performance.
- Greater rigor was observed in applying legal criteria compared to usual practices.
- In advanced case search, flexibility for users was maintained without compromising the technical accuracy required by system filters.
- The generated information has a structure adaptable to thematic and patronymic filters within the system.
- In automatic resolution generation, the results correctly respected established rules, optimizing validation times.
Conclusion
The experience confirmed that applying artificial intelligence to case management delivers concrete improvements in efficiency, traceability, and task simplification for users.
The pilot results provide a solid foundation for future evolution stages, consolidating AI as a strategic tool to strengthen document management in large-scale organizations.