Eletrobras Electrifies Innovation with AI and Automation

Eletrobras, Latin America's largest electric company, embarked on an innovative journey to automate its technical document auditing process, reducing manual effort by over 90% and saving $277,000 annually. Using Automation Anywhere's Document Automation and Google Vertex AI, Eletrobras achieved an impressive 92% accuracy in processing diverse, complex documents, with aspirations of reaching 100%. This success underscores Eletrobras' commitment to leveraging automation and generative AI to enhance efficiency and drive value across its operations.

Energized to automate

Eletrobras, Latin America's largest energy company, generated nearly 150 thousand GWh of electricity in 2023. Operating across Brazil, it focuses on the generation, transmission, and commercialization of power through hydroelectric, thermoelectric, nuclear, wind, and solar plants—highlighting its commitment to making Brazilian energy among the most renewable in the world.

Being an energy company means Eletrobras is an inherently engineering-focused organization. With engineering comes technical documentation, which requires meticulous processes and tight controls to ensure operational safety and reliability. Auditing these processes and documents is a core part of those controls, and, as such a large organization, Eletrobras reviews 65,000 technical documents annually.

Luckily, Eletrobras has an affinity for innovation and embarked on an automation and AI journey to reduce the nearly 10,000 hours its teams spent manually reviewing and approving those documents.

An innovation powerhouse focused on Intelligent Automation

Eletrobras’ focus on innovation culminates in its “Innovation Powerhouse” initiatives supported by “digital excellence centers,” which approach business growth and value creation as continuous learning processes. Its six excellence centers cover hyperautomation, AI, robotics, and other initiatives, but all work together to generate organizational value.

As the leader of Eletrobras’ hyperautomation excellence center, Vitor Paulo Moreira Correia set his sights squarely on the huge manual effort required of its document auditing process.

“The process itself is very straightforward: it’s basically identifying engineering documents for projects to improve the transmission and generation of energy,” Correia explained at Imagine Austin 2024. “But, those technical documents must be analyzed, verified visually, and approved by the engineering team.”

This manual process required thousands of hours from talented engineering resources, which made it a high-value candidate for automation. But, the team was worried that the technical and unstructured nature of the documents could prove to be a challenge for automation alone.

“We wanted a solution to not only free up the labor effort but also automate 100% of the tasks in this entire process,” added Correia. “We set an accuracy goal minimum of 60%. To get there would require using a combination of three types of technology: automation, intelligent document processing, and generative AI.”

Understanding and automating complex cognitive processes

Developing the Intelligent Automation solution for Eletrobras required first understanding the nuances of the document auditing process, which required a combination of data extraction, interpretation, and validation. Here’s how Eletrobras defined the required tasks:

1. Locate and retrieve the correct documents from the Eletrobras enterprise content management (ECM) portal.
2. Analyze each document, which could be dozens of pages long and vary widely in format, including images and engineering drawings. The extracted information must then be compared against predefined rules.
3. Finally, determine whether the documents are approved or rejected.

Across these steps and the many types and formats of documents, the existing manual effort was prone to errors and could take hours per document.

Eletrobras began automating the process using Automation Anywhere and a third-party solution for document processing but quickly found that the document processing wasn’t effective at pulling information from the technical documents that were key to the auditing process. After a demonstration of Automation Anywhere Document Automation, Eletrobras made the switch.

“The enhancements to Document Automation were a game changer,” said Correia. “We tested a few solutions, but this was the best one.”

Adding generative AI to understand highly technical documents

An initial challenge for Eletrobras was automating the processing of various document types. Many of the files were stored in formats other than PDF, including diagrams, images, and CAD files. To address this, Eletrobras implemented automation to convert these formats into PDFs and compress them when the resulting file sizes exceeded a set limit.

Once Automation Anywhere’s Document Automation successfully extracted critical information from the documents, Eletrobras needed to intelligently parse and evaluate the technical content within. This included analyzing images, graphics, and visual representations of equipment and processes to ensure accurate assessments and compliance.

To interpret this information, Eletrobras relied on Google Vertex AI as a component of the Automation Anywhere automations.

“Vertex AI would use our predefined rules to interpret and classify the data and then validate to decide if a document is approved or rejected,” Correia explained. “Generative AI would then draft and send an email to the responsible engineering team. This is all done automatically.”

Just 4 weeks from kickoff to Intelligent Automation implementation

Eletrobras' agile automation team consisted of just eight members, including a product owner who led the project, supported by business users, analysts, and two automation specialists. Despite the small team size, they delivered a comprehensive, generative AI-powered automation solution in just four weeks.

“From the construction to getting the solution running was very quick, which was an excellent opportunity to give us confidence in this approach and these technologies,” Correia said. “We faced many challenges at the beginning, one of which was not starting with Automation Anywhere Document Automation.”

Soon, Eletrobras successfully automated the entire document auditing process, with the automated process mirroring the original manual workflow. The results far surpassed their initial goal of 60% automation, ultimately reducing manual effort by over 90% while maintaining exceptional accuracy.

“Early on, people did worry about the risk of errors, especially with so many different document types,” Correia said. “But the confidence in the automation is increasing day by day. Our accuracy rates now are above 92%, and we expect it to get higher over time.”

Saving nearly $300,000 and charging towards 100% accuracy

This complex technical document auditing automation, powered by Automation Anywhere and generative AI, is already live in one of Eletrobras's five main divisions. Given that similar technical document audits are required across all divisions, the company plans to automate and standardize the process enterprise-wide. Correia anticipates “quick implementations” across the board, streamlining operations throughout the organization.

“We’re almost over 10,000 hours annually spent on this process, and we reduced it by 92%—that’s a financial savings of $277,000,” Correia added. “And, we expect to hit 100% so people can be freed up to do more intelligent, more value-added activities. This isn’t just RPA; this is Intelligent Automation and generative AI that encompasses 100% of a task by using a little creativity and innovation.”

Eletrobras continues to identify opportunities for generative AI-powered automations in other areas of the business, including financial, governance, audit, and additional engineering use cases that involve similar document processing and validation tasks. The company is also exploring Automation Anywhere’s new solutions, such as AI Agent Studio, to enhance its automation capabilities further.

“We’re looking at a lot of new products from Automation Anywhere like Co-Pilot and Autopilot to weave in more generative AI,” Correia concluded. “The future holds many promising opportunities with Automation Anywhere, so we expect to improve these results even more.”

Eletrobras isn’t alone in generating such massive savings through AI + automation. Read more customer stories to learn how organizations of all sizes, industries, and regions are using automation, generative AI, and AI agents to automate the impossible.