Amazon
Amazon uses Amazon Q Developer to upgrade its internal Java applications
Amazon pointed Amazon Q's code transformation feature at its own Java applications to move them to Java 17. CEO Andy Jassy said the average upgrade fell from about 50 developer-days to a few hours, and that Amazon estimates it saved 4,500 developer-years.
Key figures
Copied exactly from the sources. Copy any figure with its citation.
From about 50 developer-days to a few hours
Average time to upgrade an application to Java 17 (before: typical, after)1
The equivalent of 4,500 developer-years
Developer effort saved (Amazon's own estimate)1
More than 50%
Production Java systems upgraded in under six months1
79%
Auto-generated code reviews shipped without additional changes1
What it took
The problem
Updating foundational software is tedious but important work. Developers tend to put it off because it adds no visible features. Amazon had a very large base of applications written in Java (a common programming language) that needed to move to newer versions of it, with security and infrastructure cost benefits waiting on the other side.
What they built
Amazon used the code transformation capability of Amazon Q Developer on its own internal systems. The tool takes an application on an older Java version and produces the changes needed to run it on Java 17. Developers then review the output.
How they did it
Per CEO Andy Jassy's post, Amazon integrated Amazon Q into its internal systems to update its foundational Java software. Developers reviewed the generated changes, and Jassy reported that 79% of the auto-generated code reviews shipped with no further edits. The sources do not describe the review process, testing setup or team structure in more detail.
Results
Jassy said the average time to upgrade an application to Java 17 dropped from about 50 developer-days to a few hours. He said Amazon estimates this saved the equivalent of 4,500 developer-years and that more than half of its production Java systems were upgraded in under six months. He also cited an estimated $260M in annualized efficiency gains from better security and lower infrastructure costs. These are Amazon's own estimates.
What to copy
Start with a bounded, well-understood migration where success is easy to check (the application still builds and passes tests). Track how often reviewers accept the generated change as is.
All reported figures
| Measure | Figure | Citation |
|---|---|---|
| Average time to upgrade an application to Java 17 (before: typical, after) | From about 50 developer-days to a few hours1 | |
| Developer effort saved (Amazon's own estimate) | The equivalent of 4,500 developer-years1 | |
| Production Java systems upgraded in under six months | More than 50%1 | |
| Auto-generated code reviews shipped without additional changes | 79%1 | |
| Estimated annualized efficiency gains | $260M1 |
Lessons for your project
Pick maintenance work that teams keep postponing. Version upgrades add no visible features, so they are a good first target for code transformation tools.
Measure the share of AI-generated changes that ship unedited. Amazon reported 79%, which is a more honest quality signal than time saved alone.
Treat the headline savings as Amazon's own estimate, not an audited figure, when you use it in a business case.
Caveats
- All figures come from a post by Amazon's CEO, who also sells the product. No independent audit was published.
- The model behind Amazon Q Developer is not named in the sources.
Sources
- Andy Jassy LinkedIn post on Amazon Q code transformation and Java upgrades
LinkedIn (Andy Jassy) · 2024-08-22 · Social post · linkedin.com
- Amazon's AI assistant saves 4,500 years of development time, CEO Andy Jassy says
The Decoder · 2024-08-23 · News · the-decoder.com
- Andy Jassy on using generative AI in software development at Amazon
Simon Willison's Weblog · 2024-08-24 · News · simonwillison.net
Summarized in our own words from the sources above. Spotted an error? See corrections.