GitHub Copilot app modernization
Assistant-driven upgrade help inside the developer workflow.
AI-assisted modernization
AI does the repetitive rewrites. Engineers own the result.
Automated code transformation is fast. It is not accountable. Here is how we use it, where it fails and what stands between its output and your production.
The short answer
AI-assisted modernization uses automated code transformation tools to do the repetitive rewrites in a legacy codebase: upgrading Java versions, swapping frameworks, moving to new APIs. Tools produce the change, tests prove it, and senior engineers review and sign every diff. The result is faster delivery without handing judgment to a model.
Strengths
What the tools do well
Modernization contains a large amount of work that is repetitive and rule-shaped. That is where tools help.
- Renaming packages and imports across thousands of files, such as javax to jakarta.
- Replacing removed or deprecated APIs with their successors.
- Updating dependency versions and build configuration.
- Applying one framework change consistently everywhere it occurs.
- Producing a report of what changed, which helps the reviewer.
Limits
Where they fail
A tool changes code. It does not know what the code is for.
- Behavior that depends on runtime configuration, such as app-server settings, JNDI names and security realms.
- Business logic that a test does not cover. The tool cannot see a rule that lives in someone's head.
- Libraries with no automated migration path.
- Subtle changes in semantics, such as how a newer framework version handles a default.
- Generative output that looks right, compiles, and is wrong.
Tools
Tools we run
We choose the tool that fits the codebase. Often that is a combination: a deterministic recipe for the bulk edit, an AI-assisted tool for the cases a recipe does not cover.
Accountability
The review step
- 1
Small slices
Tools run on a bounded part of the code so a reviewer can read the whole diff.
- 2
Automated gates
The build and the test suite run before a person looks.
- 3
Engineer review
A senior engineer reads the diff, checks edge cases and runs it against the tests.
- 4
Signature
The engineer signs the change. Failed output is rewritten by hand, and the failure is recorded.
Safety net
Tests as the safety net
The quality of AI-assisted modernization is capped by the quality of the tests. If you cannot test it, you cannot safely change it, by hand or by AI. So the first step in every engagement is to measure coverage and write characterization tests where it is thin. See the method.
Evidence
Results from our sample-repo run
We are running OpenRewrite, AWS Transform and GitHub Copilot app modernization on the same open-source Java 8 and Spring Boot 2 application. For each we will record the files changed, the test results before and after, the manual fixes needed and the minutes of engineer review. Until the run is published, we claim no outcomes. The method above is how we work today.
Judgment
When AI is the wrong tool
- The code has no tests and no one can describe its behavior. Write the tests first.
- The change is a design decision, not a rewrite, such as how to split a monolith.
- The application is small and a person can do the change in an afternoon.
- The target needs vendor-specific knowledge the tool does not have.
Go deeper
AI tools, compared and explained
How each tool works, where it stops, and how we combine them.
It is application modernization where automated tools such as OpenRewrite and AWS Transform do the repetitive code rewrites. Tests prove each change works, and senior engineers review and sign every change before it ships.
Not without checks. Our process sends every change through automated tests and a named senior engineer. Output that fails review is rewritten by hand.
OpenRewrite, AWS Transform and GitHub Copilot app modernization, chosen per codebase. We are not affiliated with AWS.
It reduces the cost of repetitive edits. It does not remove the cost of testing, review and cutover, which are most of the risk. We will show the measured effect once our sample-repo run is published.
An AI-ready application runs on a current Java version, exposes clean APIs, has tests you trust and keeps its data reachable. The assessment scores all four.
Know the cost before anything moves.
A fixed-price assessment gives you the inventory, the risk score and a phased plan. Scope and cost are fixed before the migration starts.
Book a Fixed-Price AssessmentGet in touch
Talk to a modernization engineer
Tell us what you run today. An engineer, not a sales script, replies within one business day.