Modernizing Fuel Access in Underserved Communities

Overview

Puma Energy invested in the Databricks Data + AI Platform with Agent Bricks to predict fuel stockouts, generate safety audit reports, and personalize loyalty offers across Latin America, Africa and Asia-Pacific. The company decreased data latency by 98-99% (from 4-5 hours to under 5 minutes), reduced fuel stockouts by more than 20%, and cut data costs by 15%.

The challenge

Puma Energy sought to predict fuel stockouts, generate safety audit reports and personalize loyalty offers, but fragmented data, slow dashboards and manual workarounds meant decisions lagged by hours. 95% of dashboard use involved raw data exports, and insights took 4–5 hours to reach decision-makers.

The solution

Puma Energy chose the Databricks Data + AI Platform with Agent Bricks, beginning with Delta Lake as the foundation for unified data storage and real-time analytics. Unity Catalog governed all structured and unstructured data with fine-grained access controls, and Databricks SQL let teams query live data and monitor performance in near real time. Databricks Agent Bricks powers GenAI use cases including retrieval-augmented generation to extract insights from unstructured feedback, while Lakeflow Jobs, MLflow and MLOps pipelines automate model development and deployment across a mix of models including GPT, Llama and Claude, including a multimodal GenAI tool (Guardian Vision) that analyzes CCTV footage for policy violations.

Reported business value

Puma Energy cut data latency by 98–99%, from 4–5 hours to under 5 minutes, reduced fuel stockouts by more than 20%, cut safety and audit preparation time from two weeks to one hour, and reduced data costs by 15%.