We’ve all heard the phrase “AI won’t take your job, but someone using AI will.” Strangely, though, we seem to forget this when it comes to legacy modernization. People are hoping AI will finally deliver the “easy button” for these projects and do the whole thing itself.
I understand why. Legacy modernization is some of the hardest work in software. It’s technical. It’s architectural. It’s organizational. It often requires shifting long-standing processes, rules, and expectations.
But as much as I love where AI is heading - and as excited as I am about how we’re using it at nvisia - it’s important to keep perspective:
AI tools are currently operating at the level of junior and mid-level engineers. More skill than that is required to modernize a complex legacy system. Architects are still necessary, but AI tools are a big help.
Legacy systems have their advantages. They can be impressively robust and feature-rich, enhanced over years to provide an amount of features no MVP will match, but there’s always reasons you want to replace them, too.
In the typical nvisia modernization project, our clients want new systems that will:
Improve customer and user experience
Implement a new data strategy
Provide additional flexibility and configurability for the business
Automate manual workflows
Improve maintainability and lower TCO
And so much more!
With so many changes planned for the system, modernization is about way more than porting code to a new language. You cannot simply say “Hey AI, can you write this in .NET for me?” That’s not how it works.
When you break down legacy modernization, the work divides naturally into two roles:
Senior engineers and architects who determine strategy, boundaries, modernization patterns, business rules, and the sequencing of change.
AI-enabled tooling that accelerates the hands-on execution of certain tasks — code generation, scaffolding, wrapper creation, unit test generation, data transformation, UI generation, and more.
This is where the real opportunity is today.
Strangler Fig is a common pattern architects use to support iterative modernization roadmaps. You still need a senior person deciding where to cut, what to wrap, and how to sequence the evolution. But from there…
AI can generate REST APIs from legacy code
AI can scaffold wrappers
AI can generate new UI components from mockups
Tools like OpenLegacy can automatically surface mainframe functions as callable services
AI doesn’t remove the architect; it makes the architect more effective.
Another common pattern architects use for modernization is Branch by Abstraction. Once you define the abstraction layer, AI can:
Generate both sides of the interface
Scaffold calling code
Build unit tests to support migration
Convert business rules into new implementations
Again - strategy stays human. Execution becomes shared.
When we break systems into layers, the opportunities become clearer:
Your UI/UX will be changing. Don’t attempt to convert legacy front-end code. Instead, have a UX designer build your new workflow using AI-enabled design tools, then generate front-end code from the mockups.
Use AI to understand the old system, not replicate it. Generate new data access code from a clean data model.
Business rules are better candidates for conversion. Use AI to identify them. Once identified, they are perfect candidates for model-to-model conversion.
This is where senior engineers + AI really shine together.
Modernization projects carry familiar risks. Time saved in creating code can be focused on mitigating challenges. AI may be able to help.
SMEs know the old system, not the future one. AI can brainstorm features, competitive comparisons, and user expectations - but leaders must still observe the real work happening today and envision where it needs to go.
Legacy modernization projects have typically moved batch to real-time. New AI techniques may push the opposite way: putting more work into automated backgrounds. Tread carefully and design for human-in-the-loop interventions.
Data migration has traditionally been a major hurdle for legacy modernization projects. AI will help tremendously, but you still need to start early.
Please don’t wait until the end. Parallel run, testing automation, and early migration strategy are non-negotiable. AI testing tools are ready. Teams should be using them now.
As leaders, we shouldn’t be asking:
“Are you using AI?”
That’s too surface-level. Instead, ask:
How are we breaking down the problem?
What modernization pattern are we using, and why?
Where does AI strategically accelerate this pattern?
What velocity gains are we seeing from code generation?
How are we validating AI output?
When those answers are clear, modernization becomes more achievable - and far less overwhelming.
Legacy modernization has never been easy work. It still isn’t. And it still requires the skill and leadership of senior engineers, architects, and thoughtful product owners.
But there is a tremendous opportunity right now to reshape how we approach it:
AI isn’t the easy button. But it is the accelerator modern teams have been waiting for.
And the teams who learn how to pair senior strategy with AI-powered execution? They’re the ones who will modernize faster, reduce risk, and deliver systems that truly move organizations forward.