Why legacy banking systems are becoming a competitive risk

Giovanni Gomes
Director of AI, Stefanini Group

What kinds of companies do you work with?

We have more than 2,000 active clients, and 35 to 40 percent of them are in the finance sector. We work with MasterCard, BlackRock, Citibank, and Santander, to name a few. We have an extensive range of portfolio covering applications, cybersecurity, finance, and marketing. We are proud to be a one-stop tech consulting shop.

As Director of AI, how do you help clients in the finance industry?

I’m responsible for what we call “applied AI.” Our AI specialists try to understand customer needs, then customize solutions using AI to make processes faster, better, and more efficient.

For example, we work with a global payment solutions company that wanted to update its digital solutions with a focus on user experience, emphasizing the need to simplify processes and provide predictive and proactive services.

With an AI-first approach and SophieX, our intuitive AI-powered natural language assistant, we redesigned the most critical process.

These changes delivered strong results: user satisfaction rose by 95%, ticket resolution time dropped by 67%, and support requests decreased by 34%. This directly generated more than $3.2 million in annual cost savings.

But many banks still rely on legacy applications going back to the 1960s. Are their systems even ready for AI?

AI can actually help banks and other companies modernize these systems faster. Let’s say they still use COBOL, a programming language created in 1959. We can use AI to read the source code, understand the rules and behavior of the legacy application, and then transform it into a new architecture and a new program.

Then we can move it into the cloud while preserving the business logic and behavior of the old system.

So, a single AI uploads everything …?

No. We have specific AI agents for different parts of the process. For modernization projects, we typically use seven or eight agents.

One agent reads the source code and creates documentation. Another takes that documentation and generates new code. Another handles quality control. Another manages deployment to the cloud.

Using AI, we are reducing time-to-market for these projects by 50 to 60 percent. In many cases, we can cut the timeline in half.

How long is the timeline usually?

It depends on the size of the application and the number of lines of code. We worked with Huawei to modernize systems at one of the biggest banks in Brazil. The project involved about 12 COBOL programs and 200,000 lines of code.

The client expected the project to take three months. We finished in six weeks.

What could the bank do after the conversion that it couldn’t do before?

First, they reduced their costs. Running a mainframe is expensive compared with cloud infrastructure.

Second, they gained flexibility. Once you move to a modern programming language such as Java, it becomes much easier to create new features because many more developers know the language.

Third, cloud infrastructure allows banks to scale up more easily. It provides access to security and monitoring tools that identify and resolve issues. That’s hard to do in a traditional mainframe environment.

Do you have to re-engineer all your systems at once, or can you do this step by step?

Doing a “Big Bang” upgrade to AI is not recommended. Banks want to disconnect from mainframes gradually, so they can move applications into scalable cloud environments. AI makes this transition easier and more cost-effective.

How does AI do help to speed things along?

In the past, banks often couldn’t modernize because they lacked documentation for their legacy systems. It is almost impossible to ask someone to manually read millions of lines of code and document everything. It is a huge task. Now, AI can read everything and create documentation in minutes.

Modernization is about preserving the business logic and rules while changing the technology underneath. We don’t need to throw everything away. We preserve the intelligence and change the architecture. The idea is to follow a structured roadmap until everything is migrated.

When big banks with legacy IT start competing with “digital-native” outfits, do the upstarts have the advantage?

Generally, yes. Digital-native banks have modern infrastructure, so they can release new features much faster. Today, banking is all about the customer experience.

Think about a banking app. There are always new features, new interface elements, new services. In a mainframe environment, making those changes can be slow. For a digital-native bank, it may take only a few days.

All our clients want this kind of agility.

For example, Itaú, the largest bank in Brazil by market value, bought two IT consulting companies so it could update and improve its digital services. It works with Stefanini on some projects, as well.

Customers love it when banks make their apps easier to use. To keep doing that, banks need to stay technologically current.

Do larger banks still have an advantage because of their brands?

Yes, but younger customers increasingly care more about the experience than the name of the bank.

Digital-native banks understand this very well. They invest heavily in user experience. That is why traditional banks are modernizing their infrastructure and applications: they want to compete on equal terms with digital banks.

Do bankers need to change the way they think about their industry?

Yes. They need what we call an “AI-first” mindset.

In the past, companies created products first, then looked for places where AI might help. AI-first means there is no longer any a debate about whether to use AI. It becomes central from the beginning. You put AI at the center of the strategy and design around the capabilities and benefits it provides. Instead of treating AI as an add-on, companies must think of it as a core engine of innovation and competitiveness. That changes everything.

Can that “AI-first” mindset be acquired?

The only way for a company to change is from the top down. Leaders need to understand this and keep AI on the agenda constantly.

Stefanini has been working with AI for the last 14 years. Our founder, Marco Stefanini, is still actively involved in the company. When OpenAI launched ChatGPT in late 2022, he immediately brought together our AI teams to understand what would change, and how we would respond.

Every company needs to understand this: if you don’t put AI first, your competitor will. And if that happens, you may not survive. Companies need to change. They need to adapt.