Conversational AI Agent Turns Refinery Data into Instant Answers

The customer is a large, integrated private energy company headquartered in South America, operating across oil, gas, fuels, refining, chemicals, transportation, distribution, and renewable energy. With more than 21,000 direct and indirect employees and operations across multiple Latin American markets, the organization has significant regional scale and complex downstream analytics requirements.

The customer operates a Lakehouse data platform on AWS. Within it, the Downstream Business Line had built multiple data models calculating a wide range of business-critical KPIs across its units. The problem: making that modeled information accessible to business users required building a custom, ad-hoc dashboard for every individual use case. That meant significant investment indesign, development, and implementation time before any value could be extracted from the data.

This dashboard-per-use-case approach meant that unlocking insights was slow and resource-intensive. Business users without direct technical support often had to wait hours, sometimes days, for a dashboard to be built or for the technical team to run a query and interpret the results on their behalf. As the number of KPIs and use cases grew, this model of data access became a bottleneck rather than an enabler for decision-making.

Netrix Global implemented a Generative AI conversational agent powered by Amazon Bedrock, connected directly to the client's existing Lakehouse data platform on AWS. The agent allows business users to query production, consumption, crude oil runs, materials profit and loss, planning metrics, and other refinery indicators entirely in natural language.

• Business-aware reasoning: The agent reasons over each user's question, mapping it to the underlying business context, logic, and rules embedded in the data model, not just the raw schema.
• Governed SQL generation and execution: It translates natural-language questions into SQL queries executed directly against the Lakehouse platform, returning accurate answers in the same vocabulary the business already uses.
• Dynamic visualization: When useful, the agent generates charts on demand to enrich its responses, giving users visual context without any manual dashboard work.
• Delivered through Microsoft Teams: To maximize adoption, Netrix integrated the agent as a Microsoft Teams bot, meeting users inside the collaboration tool they already use every day. AWS-native AI infrastructure paired with a Microsoft-native user experience.