Getty Images
Integrate and modernize legacy systems with AI
Complete replacement of, or over-reliance on, legacy systems can be disastrous. Organizations that modernize gradually will thrive in markets where AI is vital.
For many executives considering AI's potential to modernize systems, there's an almost magnetic pull toward the new and building from scratch. But is that always wise?
McKinsey & Company warns readers that failed modernization efforts can rack up costs of hundreds of millions in its article, "AI for IT modernization: Faster, cheaper, better." This could be one of the reasons that organizations spend up to 80% of their IT budgets maintaining outdated systems, according to an August 2025 IDC report. While businesses can see a future built around AI, their existing and potential technical debts have transformed from an IT concern into a significant threat.
Technical debt isn't the only threat business leaders must consider in an increasingly modern enterprise landscape. Organizations also face the following escalating risks:
- Security vulnerabilities.
- Loss of expertise.
- Recruiting talent.
- Retaining essential business logic embedded in older code.
However, these risks cannot be the reason businesses retain and invest resources in uncompetitive legacy systems. Instead, enterprise leaders should opt for a strategy that preserves the legacy systems that work while building toward the modernized systems needed for the future.
The current landscape of legacy systems
Legacy infrastructure constrains operations across every major industry. For instance, banking platforms handle trillions of dollars in daily transactions on decades-old mainframe systems, and the U.S. financial industry still relies heavily on COBOL, a programming language written before many developers were born. Examples like this create a sense of urgency around modernization as legacy maintenance prevents investment in value-adding development.
The cost of inaction includes several disadvantages, including the following:
- Organizations cannot launch new digital products.
- Businesses struggle to meet evolving regulatory requirements.
- Top talent avoids companies running legacy systems.
Strategic advantages of gradual AI modernization
Traditional modernization projects often fail because they attempt everything at once. Organizations attempting "lift-and-shift" migrations can end up with inefficient applications running on modern infrastructure without gaining any benefits.
AI-powered modernization promises more. Cognizant's research report "AI's two-year timeline: The path to meeting the legacy modernization mandate," involved 1,000 business and technology leaders at Global 2000 organizations. They found that 66% prioritized workforce productivity improvements through modernization. One healthcare provider modernized patient records systems by connecting legacy infrastructure to cloud applications, using AI to prevent clerical errors. Staff gained more time for patient care, and operational costs declined.
Practical, real-world AI integrations
Several organizations demonstrate how AI can bridge the divide between legacy and modern systems without catastrophic disruption.
- Goldman Sachs. Deployed generative AI (GenAI) as a "developer copilot" across engineering teams. The AI assistant handles repetitive tasks such as generating boilerplate code, creating documentation, writing test cases and refactoring legacy codebases. As a result, Goldman Sachs saw an approximately 20% increase in efficiency. Working in a strictly regulated market, the firm positioned AI within private environments, incorporating code compliance checks to meet compliance and security standards.
- Airbnb. Used large language models (LLMs) to accelerate large-scale test migration. The company needed to modernize its testing infrastructure across its entire codebase. AI tools automate translation work, reducing engineering burden while maintaining test coverage. This approach enables Airbnb to preserve functionality while updating its platform's underlying architecture.
- U.S. Office of Personnel Management (OPM). Seeks to use AI to modernize COBOL-based retirement systems. The two-year project, which began in 2025, uses AI to convert code from COBOL to modern programming languages like JavaScript or Python. AI handles bulk translation while human developers validate and refine output. OPM conducted an extensive analysis, reviewing millions of lines of legacy code and categorizing them by complexity to focus modernization efforts where they matter most.
These three examples share the following common elements:
- Integrate AI to augment existing systems, not replace them.
- Pursue incremental transformation rather than wholesale replacement.
- Retain human oversight and validation over processes.
- Preserve systems and business logic that work while modernizing infrastructure.
A strategic framework to modernize with AI
Organizations approaching AI-powered legacy modernization should consider a phased approach that builds momentum over time. Since it is such an expensive undertaking, it can be tempting to address technical debt first. However, businesses must address the foundations first.
Phase one: Foundation building
Begin by focusing on immediate operational pressures while laying the groundwork for larger initiatives. Use AI to extract and document business logic from legacy systems. GenAI can effectively crawl through source code, translating it into natural language and mapping it to business specifications. This enables stakeholders to understand their legacy systems in fine detail and provides discrete, high-value integration points where AI-generated code connects the legacy system to modern interfaces.
There's strategic value in confronting what you know about your own systems. Remember that existing business logic represents decades of organizational learning, so do not discard it offhandedly.
Phase two: Systematic debt reduction
Now organizations can begin to tackle technical debt systematically. Use AI to translate legacy code to modern languages at scale, while validating each transformation before deployment.
This AI-driven code translation requires fewer specialists in obsolete languages. This removes organizational risk associated with a lack of expertise in legacy programming languages. However, businesses still require people with the skills to validate AI-generated translations. While businesses might not need the skill sets that legacy systems and languages demand, it would be incorrect to say that a skill gap does not still exist as organizations search for candidates who can effectively use these new technologies.
Despite the challenges this skill gap poses, remember the focus of this step: reducing technical debt. Each reduction in legacy maintenance costs frees resources for modernization investment. Each piece of deprecated code removed represents one less security vulnerability.
Phase three: Transformation and growth
After modernization teams have reduced technical debt and facilitated an emerging modern infrastructure, their organizations can pursue more ambitious initiatives that leaders would have considered too risky in the past.
Deploying AI capabilities across an organization can help employees respond rapidly to changing customer requirements, launch new digital services or even help a business enter adjacent markets that were previously inaccessible due to the constraints of legacy systems.
Throughout all these phases, treat modernization as a continuous process. Prioritize business outcomes over technological perfection, while maintaining a focus on governance and efficiency.
Donald Farmer is a data strategist with 30-plus years of experience, including as a product team leader at Microsoft and Qlik. He advises global clients on data, analytics, AI and innovation strategy, with expertise spanning from tech giants to startups. He lives in an experimental woodland home near Seattle.
