Most people say: Legacy systems are the problem. We say: They’re the treasure.

What if everything your old systems know is exactly what your business needs for the future? What if AI is finally the key to unlocking that knowledge, and turning it into real, usable data products?

Many companies want to modernize their legacy systems, but where do you even start?

👉🏼 Here’s our 7-step approach to AI-driven, data-centric legacy modernization:

➡️ At adorsys, this is exactly how we transform legacy systems, extracting business knowledge and turning it into modern, semantic data products.

Legacy Isn’t the Problem. It’s the Opportunity!

Especially in banking and insurance, organizations are still running systems that are 20, 30, even 40 years old. Mainframes, COBOL cores, C++ modules, they still power mission-critical processes every day.

And yes, studies show that over 70% of IT budgets go into maintaining those systems instead of building new ones. 55% of financial institutions call their legacy software the main barrier to innovation.

But here’s the twist: those systems aren’t “old” — they’re full of knowledge.

They contain the very business logic, rules, and processes that made your company successful. What’s missing is the ability to understand and reuse that knowledge efficiently.

And that’s exactly where AI comes in.

Step 1 – Extracting Knowledge with LLMs & Knowledge Graphs

Our AI experts demonstrated how Large Language Models (LLMs) can act as translators between humans and legacy code. Our intelligent agents analyze old systems, from COBOL to C++, identifying modules, their relationships, business objects, and data flows.

The result is a Knowledge Graph: a structured map of what the system knows. It shows which modules interact, how data moves, and what logic drives each process.

This allows us to reconstruct lost requirements, generate arc42 documentation, and make decades of business logic visible again, all within a human-in-the-loop (HITL) process.

And here’s the magic: When connected to a Retrieval-Augmented Generation (RAG) interface, you can literally talk to your legacy system.

“Which modules handle customer master data?” “Where is the credit limit calculated?”

The answers come from your own codebase, explained in natural language by your AI assistant. That’s Talk to Your Business, not just another documentation tool. [2]

Step 2 – From Code to Data Products

While the knowledge layer is being built, our agents also analyze data flows. They identify pivotal domain events, key transactions or business actions, that can be exposed as Data Products.

Each of these events is then described semantically, using an ontology that defines meaning, relationships, and business context. This is how we create Semantic Data Products: interoperable, machine-understandable, and ready for AI.

Think of it as a semantic façade in front of your legacy systems, connecting old data to new capabilities. Behind it, your legacy environment continues to run safely; in front, new digital services can access the same knowledge via APIs and event streams.

This sets the stage for gradual modernization using the Strangler Fig Pattern: step by step, old modules are replaced by new, data-driven, semantic agents or microservices, without disruption. [1]

Why This Matters for CIOs and Enterprise Architects

According to IT-Finanzmagazin, the world still runs on over 800 billion lines of COBOL code. That’s not a burden, that’s frozen business intelligence.

With the right AI and data-driven architecture, this code becomes an asset. The payoff is immediate:

Sustainability: Reuse what exists instead of rebuilding everything.
Transparency: Know exactly where your business logic lives.
Innovation: Turn legacy data into actionable insights for AI and analytics.

Because those who truly understand their legacy are the ones who will lead the future.

The Bottom Line

AI-driven modernization isn’t about deleting the past, it’s about decoding it. It’s about taking decades of hidden knowledge and turning it into reusable, intelligent, and connected systems.

So if you’re still running COBOL… Congratulations! You did everything right.

Now’s the time to transform that code into Knowledge + Data + Future.

📎 Sources & further reading:

Keen to explore how adorsys can guide your company through legacy modernization? Reach out to us here - our team will be delighted to discuss tailored solutions for your organisation.