LEGACY SYSTEM MODERNIZATION WITH AI: REAL PROMISE OR JUST THE LATEST HYPE?

The controversy that reignited the legacy debate

In recent weeks, the conversation around legacy system modernization has gained renewed momentum following the release of an Anthropic eBook on the use of AI for code modernization. The material highlights the potential of Claude Code to accelerate transformation initiatives involving systems written in traditional languages such as COBOL, which are still widely used by banks, insurance companies and government institutions.

The market reaction was immediate. Some interpreted the advancement of these tools as a potential threat to the traditional mainframe maintenance model, historically associated with IBM. Beyond any short-term financial impact, the discussion points to something broader: legacy system modernization is entering a new phase, driven by generative AI.

The structural challenge: legacy systems as a bottleneck to digital transformation

According to the eBook, many organizations face three structural barriers:

• Mission-critical systems built decades ago

A significant portion of the global digital infrastructure still relies on software developed between the 1970s and 1990s, systems that are highly stable and support critical operations but are difficult to evolve.

• Shortage of specialized talentT

he pool of professionals with deep expertise in these technologies continues to shrink year over year.

• High modernization risk

Rewrite and migration initiatives are known for exceeding budgets, taking years to complete and often causing operational disruptions. As a result, many organizations end up maintaining legacy systems indefinitely, even while recognizing that this limits their ability to innovate.

What changes with AI applied to software development

The central argument of Anthropic’s eBook is that AI models capable of understanding code at scale can significantly reduce the cognitive burden of modernization. Tools like Claude Code can analyze large codebases, map dependencies, explain legacy systems (often lacking proper documentation), generate refactoring suggestions and assist in migrating to modern languages through AI-assisted code translation, from COBOL to Java or Python, for example. The expected outcome is a substantial reduction in the time required to modernize critical applications.

The risk of hype: AI is not a silver bullet

Despite the enthusiasm, parts of the technical community have raised important concerns. Legacy modernization is not purely a code problem. Much of the complexity lies in embedded business logic, undocumented integrations, operational dependencies and organizational processes.

In other words, AI can accelerate technical transformation, but it does not eliminate the strategic complexity of legacy environments. Without a deep understanding of the business context, organizations risk simply recreating the same problems on a new technology stack.

As Marcos Bozza, CEO of Axoma, puts it:

“The real bottleneck lies in translating business into software. It’s becoming increasingly clear that defining requirements effectively and accurately translating business needs into user value has become one of the most critical tasks. As AI accelerates development team productivity, poorly defined requirements will only cause problems and strategic flaws to surface more quickly.”

In practical terms, the faster the technology, the more critical strategic clarity becomes. AI increases development velocity, but it also amplifies flawed assumptions.

The evolving role of technology companies

This shift is redefining the role of companies specializing in software development and technology consulting. Modernizing legacy systems with AI is not just about generating code faster, it requires orchestrating three dimensions simultaneously:

• Deep business understanding

• Technology architecture strategy

• Efficient technical execution

• 
  

Bozza adds:

“Companies like Axoma operate precisely in this intersection between technology and strategy, helping organizations translate business needs into software solutions, define sustainable architectures and use AI as an accelerator, not a shortcut.”

Modernization is no longer optional

The key takeaway from Anthropic’s eBook is straightforward: legacy system modernization is no longer a one-off initiative and has become a continuous strategic imperative. The difference now is that AI can make this process faster, more feasible, and potentially less risky.

However, technology alone does not solve the problem. The true differentiator will remain the ability to connect technology, architecture and business strategy. AI may represent the most significant advancement in software modernization to date, but its real impact will depend less on its ability to generate code and more on an organization’s ability to ask the right questions about its systems, processes, and objectives.

In an environment where development is becoming increasingly automated, understanding the problem before building the solution has never been more important.