Reimagining Legacy Systems Modernisation with GenAI: Accelerating Transformation and Unlocking Business Value

Legacy systems are like arteries hardened over time, vital, but brittle. They still carry the lifeblood of global business, even as their age constrains growth. For many organisations, modernisation remains a promise only partly fulfilled. After years of cloud adoption and agile transformation, most enterprises still face the same bottleneck: translating decades of undocumented logic into future-ready architecture.

Few technologists have navigated that complexity as deeply as Bhupender Saini, a Senior IEEE Member and technology leader with two decades of experience in distributed systems, scalable architectures, and AI-enabled modernisation. His work focuses on using Generative AI not just to accelerate software engineering, but to redefine what modernisation means in measurable, operational terms.

We sat down with him to discuss how AI is reshaping engineering workflows, governance, and the business value of speed.

Bhupender, Why is modernisation still a struggle for most enterprises despite years of investment?

Because modernisation was never just a technical problem. It’s a knowledge problem. Legacy systems are full of hidden dependencies, outdated business rules, and undocumented logic that only exist in people’s heads. When those people move on, understanding the system becomes a matter of archaeology.

Most modernisation programs treat migration as translation, rewriting old code in a new syntax, moving it to the cloud, and hoping for the best. That doesn’t work. You can’t scale understanding by adding more engineers. What slows organisations down isn’t computing power, it’s comprehension.

We have spent years improving deployment speed, but very little time improving discovery. Until recently, that step, understanding what the system does and why, was entirely human. That’s the part AI is finally changing. That bottleneck, comprehension at scale, is where the next breakthrough is emerging.

In your past article, you talked about Generative AI. How do you think this is changing the modernisation workflow?

GenAI is changing modernisation by giving us a way to work through expertise gaps that used to slow programs down. It does more than generate code; it helps interpret the logic behind systems where documentation is sparse or skills are limited. That becomes especially important with niche SaaS platforms like Pricefx, where the number of experienced engineers is extremely small.

In one recent modernisation effort, a large food distributor needed to move from an AS400-based pricing engine to Pricefx. The platform knowledge simply was not available at the scale the program needed. To address that, I built an agentic workflow that could generate Pricefx configuration, Groovy, XML, and JSON, directly from business requirements and design artefacts. We used ten specialised agents acting as functional owners, architects, integrators, and quality reviewers.

The workflow produced a configuration with over 95% accuracy and delivered 29 configuration stories in a week, compared to the usual nine. That reduced overall engineering effort by more than 50% and saved close to $1 million.

GenAI did not replace the engineers. It gave them room to focus on the decisions that actually push modernisation forward.

Could you elaborate on how agentic workflows redefine software delivery?

Agentic workflows change delivery by breaking the work into clear, specialised roles that AI can execute consistently. Instead of one engineer interpreting requirements, writing a configuration, reviewing it, and testing it, each agent handles a single step with predictable logic. It creates a structured assembly line rather than a set of disconnected manual tasks.

In the Pricefx program, the workflow essentially acted like a virtual engineering team. One agent refined the requirements, another translated mappings into configuration logic, another produced the Groovy or XML artefacts, and another validated them against platform rules. A separate agent generated automated tests. Each handoff produced a cleaner and more complete output for the next step.

The benefit is reliability. Humans interpret things differently; AI agents do not. Once tuned, they produce the same level of quality every time, which stabilises delivery and reduces rework.

But the intent is not to automate engineering away. The purpose is to elevate engineers, remove the mechanical layers so they can focus on architecture, design trade-offs, and domain logic. That is where the real acceleration shows up.

What role do governance, quality, and compliance play when adopting GenAI for modernisation?

Governance is the differentiator between responsible modernisation and reckless experimentation. When AI starts generating system-critical components, traceability becomes non-negotiable. You need to know which model made what decision, based on which data, and under what conditions.

Every artefact, whether it’s a generated requirement or a code block, must carry context and lineage. That’s how you build auditability into modernisation.

In my recent role as a judge for the Globee Business Awards, I’ve reviewed a couple of AI-led transformation initiatives. What separates the exceptional from the experimental is discipline, the ability to align innovation with governance. The best programs build transparency into every layer, ensuring each decision, prompt, or output can be audited.

Ultimately, AI doesn’t remove human responsibility; it expands it. We are teaching machines to understand our systems, which means we also have to teach them to respect our standards. With governance in place, the next natural question is: where does AI go from here?

Where do you see GenAI’s role in the next phase of enterprise transformation?

We are moving from automation to cognition. The first wave of AI accelerated development. The next will enable systems to evolve intelligently.

Imagine an engineering platform that documents itself, predicts code decay, and proposes architectural changes before you need to act. That’s not fiction, it’s the logical extension of context-driven AI. Models will continuously learn from telemetry, logs, and version histories, turning the development environment into a living feedback loop.

GenAI’s future is not about creating more assistants; it’s about building adaptive systems, environments that understand the intent behind design and evolve with it.

The ultimate test of modernisation won’t be speed or cost; it’ll be resilience. The systems that survive will be the ones that can think about their own evolution.