AI-DLC: how AI is transforming the software development lifecycle

Software development enters the agentic AI era

The AI-Driven Development Life Cycle (AI-DLC) is an open-source methodology published by AWS that rethinks the software development lifecycle around artificial intelligence. The approach goes beyond generating code faster. It integrates AI at every stage of the lifecycle, from design to production, including testing, security and operations. Early results show productivity gains that far exceed the 10-15% typically observed with standard code assistants.

The three problems AI-DLC solves

AI-assisted development tools suffer from three recurring limitations. The first is the one-size-fits-all approach: a bug fix and an architecture overhaul go through the same rigid workflow. The second is the lack of adaptive depth: a simple utility function undergoes the same level of design as a complex distributed system. The third is excessive automation that reduces human oversight and erodes quality.

AI-DLC addresses these three problems with adaptive workflows that dynamically adjust their scope and depth based on context. AI proposes a plan, stakeholders validate it, AI executes it, humans verify the result. This collaborative cycle preserves human oversight while accelerating execution.

A paradigm shift: AI initiates, humans validate

The fundamental change in AI-DLC is the reversal of the conversation direction. In traditional development, the developer gives instructions to AI. With AI-DLC, AI initiates workflows, decomposes tasks and proposes action plans. The developer shifts from executor to decision-maker and validator.

This reversal allows developers to transcend specialization silos. A backend developer can validate AI-generated frontend code. A junior engineer can drive complex tasks with AI as a partner. Human expertise focuses on strategic decisions and quality validation.

Amazon Bedrock re-architecture: 85% time-to-market reduction

The most spectacular case is the re-architecture of the Amazon Bedrock service. The initial project estimated 18 months of work for 30 developers. With the AI-DLC approach and Amazon Q CLI, a team of 6 engineers delivered the project in 76 days. That represents 100,000 lines of production Rust code, with over 40 commits per week per developer versus an Amazon average of 2.

The team led by Anthony Liguori, Distinguished Engineer at AWS, applied several key principles. Code was organized into small, isolated modules to reduce the context needed by AI. The monorepo allowed the model to understand the entire system. Rust was chosen for its explicit error messages that help AI fix its own mistakes. Every developer reviewed every line of AI-generated code before committing.

Amazon Stores: 4.5x developer velocity with Agent Z

Amazon Stores deployed a platform called Agent Z that orchestrates Amazon Bedrock, Kiro and AgentCore. Since July 2025, teams have created over 21,000 autonomous agents to automate operational work. A delivery address optimization agent saved 20,500 hours and reduced first-delivery defects by 74.4%.

The team also created Spec Studio, a tool that converts an existing codebase into a specification, then uses Kiro to transform that spec into code. Over 15,000 specifications have been created with this tool, with adoption growth exceeding 100% month-over-month. Pilot teams measured an average 4.5x increase in developer velocity.

Blue Origin: 95% adoption among software engineers

Blue Origin built BlueGPT, an AI-native development platform on Amazon Bedrock, Amazon EKS and Strands Agents SDK. The platform enabled every employee to create and deploy specialized AI agents with access to proprietary knowledge bases. Over 2,700 agents were deployed company-wide, with 3.5 million monthly interactions.

Results are significant: 95% of software engineers use generative AI tools to write code, manufacturing teams resolve non-conformances 70% faster, and analysis workflows are accelerated by a factor of 6.

Concrete results in development and QA

Socure used Kiro to migrate a project from Scala to Go in 2 days instead of the usual 2-3 weeks, with documentation generated automatically throughout the process. Amazon Payments reduced test case generation time from one week to hours using SAARAM, a multi-agent solution built with Strands Agents SDK. Amazon Audible saved over 50 hours during a JDK 8 to JDK 17 migration by automating unit test generation with Amazon Q Developer.

Architecture, security and operations

RazorPay applied AI-DLC to modernize a PHP monolithic system to Go microservices in 2 days instead of 5-6 weeks, with zero security vulnerabilities. HENNGE reduced security testing duration by over 90% with AWS Security Agent, while discovering vulnerabilities that manual testing had missed.

Commonwealth Bank of Australia reduced incident root cause identification from hours to under 15 minutes with AWS DevOps Agent, integrated with their existing ecosystem (ServiceNow, Splunk, Grafana).

The AI-DLC tooling ecosystem

AI-DLC relies on an integrated suite of AWS services. Amazon Q Developer provides AI assistance in IDEs and the AWS console. Kiro is an agentic IDE supporting spec-driven workflows that integrates with MCP servers to contextualize agents on proprietary codebases. Frontier Agents (DevOps Agent and Security Agent) operate autonomously to investigate incidents and fix vulnerabilities. Strands Agents SDK with AgentCore enables building specialized agent fleets with identity management, memory and observability.

AI-DLC workflows are open source and available as Amazon Q Developer Rules and Kiro Steering files on GitHub.

What this means for businesses

Traditional productivity metrics (lines of code, velocity) do not capture true business value. AI-DLC measures success through concrete outcomes: deployment frequency, recovery time, customer value delivered. Organizations that succeed integrate AI across the entire development lifecycle rather than as isolated point solutions.

LCMH supports businesses in adopting these AI-native development practices on AWS, from initial assessment to implementation.

Sources

1. AWS, Open-Sourcing Adaptive Workflows for AI-Driven Development Life Cycle (AI-DLC). aws.amazon.com/blogs/devops/open-sourcing-adaptive-workflows-for-ai-driven-development-life-cycle-ai-dlc
2. AWS, AI-DLC Workflows - GitHub. github.com/awslabs/aidlc-workflows
3. AWS, How the Amazon AMET Payments Team Accelerates Test Case Generation with Strands Agents. aws.amazon.com/blogs/machine-learning/how-the-amazon-amet-payments-team-accelerates-test-case-generation-with-strands-agents
4. AWS, Boosting Unit Test Automation at Audible with Amazon Q Developer. aws.amazon.com/blogs/devops/boosting-unit-test-automation-at-audible-with-amazon-q-developer
5. AWS, Blue Origin Case Study. aws.amazon.com/solutions/case-studies/blue-origin-case-study
6. AWS, Availity Customer Story. aws.amazon.com/ai/generative-ai/customers/availity