Modernization at Scale: How Banco Inter Orchestrates 10 Million Accounts with Automic Automation

In the fast-paced world of digital banking, agility is a valuable currency. For Banco Inter, a Brazilian leader in the fintech sector, the strategic goal was clear: launch its own line of credit cards to accelerate time-to-market and maintain its competitive edge.

However, a significant technical obstacle stood in the way: the legacy, inflexible scheduler that came bundled with its main processing platform. To overcome this challenge, Banco Inter made a strategic decision to replace the legacy tool with Broadcom's Automic Automation.

By replacing its legacy scheduler with Automic, Banco Inter achieved rapid deployment speed and massive operational scale across three key performance metrics:

We sat down with the team at Banco Inter to learn more about their incredible growth story, how they overcame legacy roadblocks, and their vision for the future of AI-driven automation. The Banco Inter team interviewed includes Carlos Alberto (Senior Analyst), Lucas de Bessa Silva, (Executive Manager), and Ramon Costa de Paula, (Senior Analyst).

Broadcom: To start, could you describe the business scenario Banco Inter was operating in? What were the main drivers behind the decision to launch a new line of credit cards, and what were the risks of not acting quickly?

Lucas de Bessa Silva, Banco Inter: At that time, in 2017, when we were working on the processor implementation project, we had a business continuity risk: we processed our cards through a third-party processor whose system wasn't in good financial shape.

Inter was going through a period of exponential growth, and since we were Brazil's first 100% cloud-based bank, we scaled our products very quickly — but the third-party processor handling our cards didn't have that same speed. That created a lot of instability in our card product.

For a lot of people, the card is the bank itself, but in reality, cards within a bank are completely distinct. That's why one of the main reasons we built an in-house processor was to secure our growth toward our target: at that point we didn't even have a million customers, and the goal was to reach 10 million checking accounts. In other words, it was a matter of business continuity and supporting Inter's growth.

Broadcom: The main processing platform included a legacy scheduler that was identified as a major obstacle. In what concrete ways did this tool limit Banco Inter's capacity for innovation?

Lucas de Bessa Silva: The obstacle was specifically in the scheduler tool that came with the VisionPLUS solution. When First Data (the owner of VisionPLUS, now Fiserv) sold us the product, it came bundled with an open-source job scheduler to orchestrate the JCLs. When we started implementing VisionPLUS, I realized that scheduler tool wasn't going to meet our needs: it wouldn't give us the visibility or the autonomy we needed, and it didn't integrate the way we wanted — because, beyond orchestrating the JCLs of the core batch processing system, we wanted to add other jobs to run on the platform (today Automic runs both JCLs and jobs in Python, Java, and other languages, integrating with our other systems and APIs).

Ramon Costa de Paula, Banco Inter: I know a lot of things were done manually. Everything that's now part of Automic started mostly in the bank's overnight processing mesh, which was done manually — there were some automated processes, and we moved everything over to Automic.

Sometimes it was done through the system's portal; and then you had to access the machines to get the log — it was as if there were several scripts inside the server. When there was an error in that execution — resubmission, resuming — it was all manual: cron would fire off whether it succeeded or not, and you'd either handle it via script or go in by hand and trigger it.

Broadcom: You decided to replace the legacy scheduler with Automic Automation. What made Automic stand out compared to other alternatives?

Ramon Costa de Paula: I think the biggest advantage is control over the execution sequence of the meshes...Automic allows for greater control, separating the meshes of each module — one module depends on another, so being able to manage the dependency between meshes was a really good point, excellent for synchronizing the execution of these system closings and openings.

If it's a known error, we can move forward; if not, we get stuck at that point — and all the other meshes that depend on it will be left waiting. That workload control Automic brings was the best part of it.

Broadcom: The main goal was to improve time-to-market. How did this project help achieve that goal?

Lucas de Bessa Silva: Automic helped us meet that deadline: if we hadn't migrated, the scheduler tool embedded in the VisionPLUS solution had very poor usability and wouldn't have allowed us to integrate with the bank's entire legacy stack. The learning curve would have been much steeper. So, without a doubt, Automic was one of the deciding factors in delivering the card processor project within the timeline estimated by senior management.

Broadcom: Your workload has grown exponentially: processing 10 million accounts and running 120,000 jobs per day. How did Automic Automation make it possible to manage this scale efficiently?

Lucas de Bessa Silva: When we were setting up the card operation, we also set up a monitoring team to ensure the processor's availability and quality around the clock. Automic’s visualization dashboard, especially overnight when the batch processing mesh runs, lets us know exactly which jobs are running, how long they're taking, and whether they're taking longer than average.

That made it much easier for the monitoring team to anticipate problems since they already had that information visually available.

Broadcom: Beyond speed and efficiency, what was the impact on reliability and costs?

Ramon Costa de Paula: Automic is recognized as a reliable tool in the bank's audits (PCI and SOX), and because of this reliability, we have already migrated processes from other tools onto Automic. Automic rarely experiences downtime — and when it does happen, usually due to infrastructure issues rather than the tool itself, recovery typically takes around an hour, not three or five hours. Since the processing mesh runs on Automic, any downtime in the tool would prevent the bank from opening the next day — which reinforces just how safe and efficient the environment is for running new processes.

Today, issues tend to stem from business rules, not the tool; it's never an Automic problem, it's always a business rule problem.

Broadcom: What would you say were the keys to success in this migration and in the ongoing evolution of your automation environment?

Carlos Alberto, Banco Inter: I think that, in a way, from the very beginning, the partnership with ICT and the usability of Automic have been essential for us to keep expanding the range of automations we have.

Broadcom: We're continuously investing in the platform, and the latest version brings new AI-based features like Agentic AI Jobs and natural-language-to-workflow conversion. As innovation leaders, how do you believe these capabilities will impact your teams?

Lucas de Bessa Silva: Here at Inter, we breathe innovation and encourage it a lot, especially process innovation using AI. Being able to consume AI agents in a way that's integrated with the platform makes things a lot easier, because we can create agent workflows where we embed AI tasks within processes that are already defined, which streamlines the routines.

When we place a specialist agent on any given topic inside a workflow, we can use AI in an automated way, generating value — sometimes even making it easier to develop a task, or providing input on whether that process should continue or not, whether it should stop, or whether it should generate an alert.

I think this future, which is already here, makes the meshes much more adaptive: we're able to have real-time analysis of everything that's happening, with automated triggers and alerts, because we can create specialist agents for specific meshes, capable of evaluating the mesh's performance and its output.

To learn more about how Broadcom is helping organizations like Banco Inter modernize at scale and embrace the future of AI-driven orchestration, explore our Automic Automation resources at Automation by Broadcom.