Petrobras Embraces AI to Revolutionize Business Operations and Increases Productivity of Auditors

Challenge

Overcoming the burden of manual and costly audits

Petrobras’ strategic goals for information and communication technology aim to ensure data readiness, enabling the company to become a fully AI-driven organization and democratizing access to AI. They have developed the frameworks to prioritize data-driven and AI-assisted decisions for all workers. “We use more than 50 models and 1,000 API keys to power our AI strategy,” says Igor Rigotti, Head of Low-Code at Petrobras. “We want to lead in AI innovation and ensure our staff have the tools needed to use it in their daily work.”

One of the first strategic projects to lead in AI innovation aimed to tackle the arduous and time-consuming audit planning process.

“The audit planning process was very costly, requiring the auditor to absorb a lot of knowledge of processes, businesses, and laws, which implied a very large study time,” says Silvia Lopes Santa Isabel, Digital Audit Manager at Petrobras. “This manual process resulted in long wait times for responses.”

They immediately realized that the audit team’s consumption and generation of unstructured data could be handled by generative AI. “I think that was the big breakthrough, understanding how this technology could help us,” recalls Silvia.

Petrobras had adopted OutSystems in 2021, implementing a successful low-code Center of Excellence and building multiple critical applications. They knew OutSystems was the right fit to build their first digital audit support agent backed by AI.

Solution

Speeding up internal audits with generative AI

The Digital Audit and IT teams established a task force to develop “Raí”, an agent that would enable the audit team to access a knowledge base with multiple data sources. After a week of interviews with auditors to determine the app's requirements, the team took around a month to assemble the first MVP (minimal value product) - front-end and design the conversation structure, menu structure, and data capture framework.

Raí conducts natural language consultations and reasoning based on the databases within the audit context, providing a more accurate response than a general chatbot like ChatGPT.

It enables auditors to quickly query specific standards, determine the type of audit to perform, and follow the necessary steps.

“Raí can cross-reference information from internal technical standards, international auditing standards, business risks, and information about scopes and past audits, which are in multiple sources, including unstructured documents,” says Silvia. “This is completely changing how audit planning is conducted.”

Petrobras chose OutSystems to be a foundational part of this strategic project because it allowed the team to build a user-friendly front-end quickly, while also enabling the connection of AI models into the application.

“OutSystems gives us the flexibility to build new functions and provide a good user experience to our auditors,” says Igor. “Importantly, it also allows us to build APIs and connectors for backend orchestration.”

Results

From manual processes to connected AI agents

Raí offers real-time support to auditors, responding to specific questions and providing relevant, personalized information tailored to the type of analysis performed.

“When validating the first Raí MVP, in terms of gains in agility, responses ranged from immediate to 20 seconds, depending on the complexity of the question and the number of databases required to support the response,” says Silvia.

“After six months of use, auditors reduced the time spent on research by up to 90%,” adds Silvia. “The ROI is staggering.” What now takes five minutes to research would have taken days to complete. Even for very specific tasks, there is at least a 50% gain in time saved. Additionally, the comprehensive database of norms, standards, and internal knowledge would be impossible to consult thoroughly without Raí.

The success of Raí confirms Petrobras’ decision to make OutSystems a key platform for developing new business applications that are at same time transactional, analytical, and AI-powered projects. They have already created two additional generative AI solutions using OutSystems and plan to expand Raí’s agentic capabilities. Soon, the app will employ AI agents that will be able to execute actions based on intelligence provided by the auditor, such as creating reports.