How to simplify app migration with generative AI tools

According to one report, as much as 70% of software used at Fortune 5000 companies was developed over 20 years ago. While there is a strong business rationale for upgrading legacy applications, such migrations are often risky. The high cost of migrating software, lack of knowledge about legacy technologies, and the complexity of testing older software slows progress and maintains technical debt.

Approaches to application modernization can be summed up by the so-called seven Rs of cloud migrations: retiring, replacing, relocating, re-platforming, reusing, refactoring, and rebuilding. Fortunately, generative AI can ease and accelerate many of these processes.

For this article, I define application migration as any approach that requires coding in a new platform; for example, migrating all or parts of an application or service between language and development platforms such as Java,.Net, Python, JavaScript, and PHP.

When application migration is the right move

Application migrations requiring significant recoding are often viewed as a last resort. But sometimes, migrating is the most prudent approach to modernization. Some examples include:

Applications that require significant re-engineering of the data models, business logic, or user experience may have little value in repurposing the existing code.

New capabilities through libraries and third-party services may make much of the existing code obsolete, and rewriting the code reduces significant technical debt.

APIs, services, and applications with significant scalability challenges or new security requirements may need a bottom-up architecture redesign.

Organizations running old applications on outdated platforms with little documentation and where subject matter experts are no longer with the company may have no choice but to migrate the application.

Enterprises acquiring smaller businesses may elect to migrate applications to standard platforms to reduce costs and simplify maintenance.

How generative AI can help

App migrations are expensive. If the intention is to recode without changing the architecture or user experience, code translators may help accelerate the process. But in cases where major changes are needed, the application migration could require a full code rewrite.

In some of these cases, generative AI can accelerate the process, reduce costs, or improve quality.

Gilad Shriki, co-founder of Descope, says, “GenAI is transforming application migrations by simplifying development workflows, including generating starter templates and conversion scripts that help developers maintain logic consistency while adapting to new paradigms.”

McKinsey reports that genAI can eliminate much of the manual work in app modernizations, leading to a 40% to 50% acceleration in timelines and a 40% reduction in costs. Next, we’ll look at the key considerations when planning application migrations using genAI capabilities.

Understand workflow and change requirements

Reviewing existing documentation and interviewing subject matter experts is often the best starting point to prepare for an application migration. Understanding the existing system’s business purposes, workflows, and data requirements is essential when seeking opportunities for improvement.

This outside-in review helps teams develop a checklist of which requirements are essential to the migration, where changes are needed, and where unknowns require further discovery. Furthermore, development teams should expect and plan a change management program to support end users during the migration. Even if few application changes are needed, communicating with end users helps alleviate their concerns and opens a feedback channel to report unexpected issues.

Product owners and business analysts should review emerging genAI tools to aid in writing requirements and agile user stories. Examples for the Jira platform include AI Test Case and User Story Issue Auto Generator for Jira and Agile User Story Map, Portfolio Roadmaps & Persona for Jira.

Analyze the architecture for dependencies

Technologists will also want to do an inside-out analysis, including performing a code review, diagraming the runtime infrastructure, conducting a data discovery, and analyzing log files or other observability artifacts. Even more important may be capturing the dependencies, including dependent APIs, third-party data sources, and data pipelines.

This architectural review can be time-consuming and often requires significant technical expertise. Using genAI can simplify and accelerate the process. “GenAI is impacting app migrations in several ways, including helping developers and architects answer questions quickly regarding architectural and deployment options for apps targeted for migration,” says Rob Skillington, CTO & co-founder of Chronosphere. “Additionally, genAI accelerates app migrations by helping outline what external dependencies can be used in the new proposed architecture, such as databases, load balancers, and third-party services, thereby creating a migration approach that matches the risk and availability constraints for the migration.”

GenAI-enabled ways to create architecture and flow diagrams include generating PlantUML diagrams with ChatGPT and diagraming the codebase with Eraser’s DiamondGPT.

Explain the code without losing your mind

Reviewing someone else’s code ranks high among the most loathed activities for software developers. However, code reviews are essential for finding defects, identifying security vulnerabilities, generating documentation for pull requests, and complying with coding standards.

Analyzing legacy code is of even greater importance when preparing for a migration. Generative AI tools like OpenAI Codex (now part of ChatGPT) and GitHub Copilot can accelerate tasks such as translating code into descriptions, extracting business rules, diagraming function call sequences, identifying data validations, and separating testing functions.

“A common challenge with app migration is understanding legacy code and application logic, particularly knowledge lost due to time or engineer attrition,” says Chad Johnson, director of artificial intelligence at SADA. “Generative AI can help decipher and explain legacy code without engineers needing to know old languages and nuances. Deep research agents can broadly scan the web for archived documentation, user forums, and blogs to craft solutions to challenging bugs and migration issues.”

Developers can prompt genAI tools with a code segment and receive a natural language response explaining its purpose and describing its functionalities.

Translate utility code to other languages

When reviewing an app for migration, some areas of code need to be rewritten, especially if the user experience or business logic requires changes. Other utility code areas may only need to be translated from one language to another while improving performance and avoiding defects. Code translators are one option to automate parts of this transition.

“GenAI simplifies application porting by automating code translation, optimizing performance, and identifying errors during migration,” says Shreyas Agnihotri, vice president of growth at Yugabyte. “For example, when transitioning applications to leverage newer technologies, genAI can refactor legacy code for compatibility, generate test cases, and streamline dependency mapping, significantly reducing manual effort and risks.”

Developers can use tools like the Java Language Conversion Assistant (JLCA) to translate Java to.Net, or CodeConvert to convert C# to Java or Python. Explaining and translating code from legacy languages like COBOL has additional challenges, as skilled experts are hard to find. GenAI and other tools for migrating Cobol projects are making it easier for developers without COBOL expertise to contribute to these projects.

“In mainframe-to-modern platform migrations, genAI’s ability to analyze COBOL and other legacy codebases becomes particularly valuable. It can parse complex business rules embedded in decades-old code, automatically generate equivalent test scenarios in modern testing frameworks, and validate that critical business logic remains unchanged through the migration,” says Sterling Chin, senior developer advocate at Postman.

Generate test cases on legacy code

Translating, refactoring, and improving code might be the easier aspects of an application migration. Testing can often be the more difficult challenge, especially because many legacy applications lack unit testing and automated regression tests.

“GenAI’s ability to comprehensively analyze existing applications enables a test-driven migration approach,” says Chin of Postman. “Automated test suites can be generated upfront to validate functional parity throughout the migration process, ensuring each migrated component maintains the original system’s behavior while enabling incremental modernization.”

When possible, a best practice is to generate test cases and a testing framework on the legacy application before migrating it. Then, equivalent test cases on the new application can be used to benchmark results across the two application versions.

“AI-powered testing frameworks further accelerate the process by auto-generating test cases, integrating security checks earlier in the software development lifecycle, and optimizing performance benchmarks for the new platform,” adds Shriki of Descope. “These advancements not only reduce technical debt but also allow teams to modernize applications faster without compromising quality, security, or developer effectiveness.”

Plan the data migration

Even after testing the application and its dependencies, a data migration may be necessary. This migration may be simple if the database technology and schema don’t require changes; for example, translating a small, on-premise SQL database to a cloud database service. However, if the schema changes significantly or data quality issues need fixes, the migration could be a project in itself.

“Generative AI can address both of the major pain points associated with app migration: extensive planning to account for various dependencies, and repetitive data migration with tedious checks to prevent data loss or corruption,” says Lou Bachenheimer, CTO of Americas at SS&C Blue Prism. “AI can understand the requisite steps and trigger pre-built governed automations to execute the data migration, leveraging traditional automation technologies to mitigate the risks associated with generative AI touching potentially sensitive data while significantly accelerating and streamlining the migration process.”

An incremental migration where data segments are migrated iteratively may be necessary for large-scale databases, if the app has high-velocity transactions, or if the business can’t sign off on lengthy downtimes to cutover from the legacy app to the migrated one.

Manage the risks of AI-generated code

While many tools aid in app migration and generate code, experts warn that development teams should conduct detailed code and security reviews as part of testing, even if the application is undergoing minimal functional changes. Development teams should communicate to stakeholders that testing and security reviews are still required for app migrations.

“The adoption of genAI tools in software development is driving unprecedented developer productivity, yet the rapid growth in code volume is outpacing key manual application security controls, like security reviews or threat modeling, highlighting the need to automate these processes,” says Moti Gindi, chief product officer of Apiiro. “While efficient, genAI-generated code often lacks awareness of company-specific policies and compliance requirements and thus can introduce new business risks. Moreover, the integration of genAI into the application stack of every business introduces new potential attack surfaces, like prompt injection, necessitating the implementation and enforcement of new code security controls tailored to this emerging threat.”

The key to successful migrations is understanding requirements, communicating with end users, selecting appropriate tools to aid migration, and developing a thorough testing program. Enterprises have a backlog of apps requiring modernization, and IT organizations that develop migration practices can accelerate, reduce costs, and minimize risks.