Overview

This article demonstrates how AI code assistants such as GitHub Copilot, Watson Code Assistant, and other AI tools can significantly enhance developer productivity. We'll explore one practical use case—upgrading a legacy Java application from Java 8 + Gradle 5.1.1 to Java 17 + Gradle 8.13—and how AI support accelerated problem resolution, reduced toil, and enabled smarter automation.

Problem Statement

Upgrading a large-scale multi-module Java application poses several challenges:

• Imagine the project consists of plenty of subprojects and huge number of Gradle tasks.
• The move from Java 8 + Gradle 5.1.1 to Java 17 + Gradle 8.13 introduces stricter dependency validation.
• Several modules failed to compile due to:
• Missing inter-project dependencies
• External library conflicts
• Implicit test class usage in production code
• Many errors that were ignored or resolved implicitly in Java 8 now fail fast in Java 17.

Manually resolving dependency issues across dozens of modules is not only tedious but also highly error-prone.

Solution: AI-Enhanced Development

Instead of slogging through manual trial-and-error, we introduced AI development agents directly into our IDEs (like IntelliJ IDEA or VS Code) using tools such as:

• GitHub Copilot
• Watson Code Assistant
• ChatGPT + JavaParser tooling

These AI agents helped:

• Detect missing dependencies
• Suggest fixes for Gradle configurations
• Identify required versions for external libraries

Automatically generate and refactor Gradle scripts

Real-World Use Case: Upgrading a Legacy Java Application

Let's walk through an actual upgrade issue encountered during the Java 17 migration.

> Task :XYZ:compileJava FAILED

/qa/sample/testbase/AbstractSystemTest.java:91: error: cannot access ProjectTestCase

import com.sample.util.junit.ProjectTestCase;                             ^

  class file for com.sample.util.junit.ProjectTestCase not found

Root Cause:

• ProjectTestCase.class exists in ABC/src/test/java.
• It is being used in XYZ/src/main/java.
• Java 17 + Gradle 8.13 doesn't allow implicit access to test classes.

Fix (with AI Guidance):

We updated XYZ/build.gradle:

implementation project(path: ':ABC', configuration: 'testArtifacts')

And added this to ABC/build.gradle:

configurations {

    testArtifacts.extendsFrom testRuntimeOnly

}

artifacts {

    testArtifacts tasks.named('testJar')

}

if (!tasks.findByName('testJar')) {

    tasks.register('testJar', Jar) {

        archiveClassifier.set('test')

        from sourceSets.test.output

    }

}

This explicitly exposes ABC's test classes as a testJar artifact and allows other modules to depend on them cleanly.

How AI Helped

Instead of troubleshooting Gradle internals manually, we leveraged AI for:

• Instant Gradle script generation
• Dependency visibility analysis
• JavaParser integration for automated dependency detection
• Accurate version suggestions for open-source libraries

Automating Dependency Discovery

In large multi-module projects, one of the main challenges is determining which module depends on which, especially when imports reference other modules' src/test/java directories.

We created a custom AI-assisted Dependency Analyzer tool that:

• Parses import statements across modules
• Detects missing project dependencies
• Classifies them as implementation, testImplementation, or testArtifacts
• Avoids introducing circular dependencies
• Updates build.gradle files automatically

This drastically reduced manual effort and build failures.

#watsonx.ai