Amazon Uses Aza and Kiro to Build Team Competition App

Key Takeaways

• An Amazon employee used the internal Aza assistant and Kiro to move from an initial idea to a working submission app.
• The example shows how AI tools can lower the starting barrier for employees who are not working as full-time software developers.
• Wider access to code generation still calls for testing, human review, security controls, and clear ownership.

An Amazon employee’s decision to build a submission app for a team competition offers a small but telling example of how generative AI is changing workplace development. The employee did not begin with a detailed architecture or even a clear understanding of which tools to use. Instead, the first step was asking Aza, Amazon’s internal AI workplace assistant, for direction.

Aza pointed the employee toward relevant tools and helped turn an uncertain starting point into a practical course of action. The employee then opened Kiro and began building the application. That sequence matters. Rather than generating an entire finished system from one prompt, AI served as a guide into the development process, helping the user find an entry point and continue working.

The account reflects a broader shift in enterprise software creation. Employees who understand a business problem can increasingly experiment with applications without waiting for a conventional development project to be approved, staffed, and scheduled. What happens when the person experiencing a workflow problem can also prototype the fix? In some cases, the gap between an idea and a testable application becomes considerably smaller.

Evidence from the developer market supports that direction. The IDC 2024 Generative AI Developer Survey found that developers using AI coding assistants reported a mean productivity boost of 35%, while 78% of respondents saw measurable gains in software delivery speed and output. Those results do not suggest identical outcomes for every task, but they indicate that AI assistance can reduce time spent on routine work and initial implementation.

Adoption is also moving beyond isolated pilots. The Stack Overflow Developer Survey reported in 2025 that 84% of developers use or plan to use AI tools in their workflows, with roughly half of professional developers using them daily. Tools such as Microsoft’s GitHub Copilot and OpenAI’s ChatGPT have helped normalize the idea of consulting an AI system while writing, reviewing, or understanding code.

Productivity extends beyond typing code faster. For someone returning to hands-on development, or an employee attempting a first internal application, the bigger benefit may be reduced hesitation. AI can explain unfamiliar concepts, suggest a project structure, identify tools, and help troubleshoot errors. That can make experimentation feel more approachable, even when the user starts without a polished plan.

Controlled research points to real but context-dependent gains. A 2026 meta-analysis published on arXiv found a moderate positive effect on developer productivity, measured at g = 0.33. The finding supports the view that generative AI generally accelerates coding tasks, while also showing why employers should avoid treating every assistant, employee, and project as interchangeable.

Speed introduces another issue. Code that runs is not automatically code that is ready for production. Employee-built applications can touch company data, identity systems, access permissions, or operational workflows. Git-based processes using services such as GitHub or GitLab can provide review histories, approval steps, and a record of AI-assisted changes. Automated testing and security scanning can add further checks before an application reaches a broader audience.

That said, excessive controls can undermine the experimentation these tools are intended to encourage. A practical enterprise model can separate prototypes from production systems, giving employees room to create while applying stronger review to applications that handle sensitive data or become operationally important. Clear ownership also helps. Someone still needs to maintain the application, evaluate generated dependencies, and respond when business requirements change.

Amazon’s example is modest, which is precisely why it is useful. The submission app did not begin as a major transformation program. It began with uncertainty, a question to Aza, and a move into Kiro. Give employees accessible tools, sensible boundaries, and a route from prototype to review, and they can often produce solutions that conventional project pipelines might not have surfaced.