AWS re:Invent 2025 - Building software like never before with Agentic AI (DVT220)

Building Software Like Never Before with Agentic AI

Overview

• The presentation discussed how Agentic AI is reshaping software development, enabling developers to accelerate the entire software development lifecycle.
• Key focus areas included accelerating development with Kira, AWS's agentic development environment, and building custom AI agents to enhance business applications.

Accelerating Software Development with Kira

• Kira brings mature engineering practices to AI-powered coding, addressing challenges with existing AI development tools.
• Key Kira features:
  • Spec-driven development: Guides developers through requirements, design, and task planning before coding.
  • Agent hooks: Automate repetitive tasks like updating documentation and test cases.
  • Advanced context management: Allows injecting organizational coding standards, security guidelines, and other context.
• Kira helps reduce time spent on rewrites and improves code quality by aligning development to requirements upfront.

Rackspace's Kira Adoption and Results

• Rackspace, a leading AWS partner, rolled out Kira across 800 developers over 4-5 months.
• Key results:
  • 85% overall efficiency gains in software development
  • Saved over 8 FTE years of development time
  • Enabled faster modernization and migration of critical legacy applications
  • Improved security, compliance, and innovation capabilities

Building Custom AI Agents

• Enterprises are rapidly adopting Agentic AI, with Gartner predicting 33% of enterprise apps will include Agentic AI by 2028.
• Benefits of building custom AI agents:
  • Natural language as the UI, improving customer experience
  • Deeper personalization and context-aware intelligence
  • Competitive advantage through automation and autonomy
• AWS provides a comprehensive set of services to build, deploy, and operate AI agents at scale:
  • Strands Agent SDK for building agents
  • Amazon Bedrock for accessing 100+ AI models
  • Amazon Bedrock Agent Core for scalable agent runtime, memory, identity, and observability

Example: Rackspace's Drew AI

• Rackspace built a multi-agent system called Drew AI using Amazon Bedrock Agent Core.
  • Data agent analyzes security telemetry
  • Help agent troubleshoots issues
  • Action agent executes remediation tasks
• Key results:
  • 64% of customer issues resolved through direct automation
  • 58% acceleration in human support resolution
  • Scaled to 10,000 users

Key Takeaways

• Agentic AI is transforming software development, enabling accelerated delivery, higher quality, and increased autonomy.
• Kira provides a comprehensive platform to bring mature engineering practices to AI-powered coding.
• Building custom AI agents can significantly enhance business applications and customer experiences.
• AWS offers a robust set of services to empower enterprises to build, deploy, and operate AI agents at scale.