AWS re:Invent 2025 - Transforming FinServ AI DLC with Kiro Agentic IDE (DEV203)

Transforming FinServ AI DLC with Kiro Agentic IDE

Overview

• Presentation by Vivec Wilso, an AWS AI hero, on transforming the financial services AI development lifecycle (DLC) using the Kiro Agentic IDE.
• Kiro is a new category that replaces the previous "AWS Machine Learning Heroes" program.
• Wilso has 27 years of industry experience and will share practical examples and best practices for using Kiro to address key challenges in FinServ AI development.

Challenges in FinServ AI Development

1. Technical Debt: Over 20% of IT budgets are spent on technical debt issues, such as:
   • Lambda runtime issues
   • End-of-life support
   • Container-based image validations
2. Compliance Burden: Heavy penalties and fines for non-compliant applications, requiring lengthy pipeline runs for even minor code changes.
3. Security Debt: Piling up of incidents and unpatched open-source vulnerabilities leading to security debt.

Kiro's Spec-Driven Workflow

1. Kiro Specs: A declarative way of producing code, consisting of requirements, design, architecture, and tasks.
2. Agent Steering: Enforcing compliance and best practices through a global agent steering file with rules defined in plain English.
3. Agent Hooks: Automating mundane tasks like documentation updates and implementing policy-as-code using tools like CFNGuard.
4. MCP Servers: Leveraging MCP (Model Composition Protocol) servers to speed up development and access organizational data, improving code accuracy.
5. Kiro CLI: Using custom agents and the Kiro CLI to improve productivity and context engineering.

Agent Steering

• Kiro supports global agent steering, allowing developers to define a set of rules in a steering file to be applied across all projects.
• The steering file can enforce the use of stable versions, prevent hardcoding of secrets, and implement organizational guidelines and standards.
• Using keywords like "must" and "critical" in the steering file helps the cloud models respect the defined rules.

Agent Hooks and Policy-as-Code

• Leveraging the CFNGuard open-source tool to implement policy-as-code for cloud formation templates.
• Automating the execution of CFNGuard rules using Kiro agent hooks, triggered on file save actions.
• Integrating CFNGuard rules within the AWS Toolkit, providing inline feedback on cloud formation stack errors.

MCP Servers

• Kiro makes it easy to install and use various MCP servers, such as the Diagrams MCP server for generating architecture diagrams in under a minute.
• Other useful MCP servers include the Headless Chrome Dev Tools browser for front-end application debugging.
• Kiro supports environment variables for sharing MCP files across teams, simplifying team development.

Kiro CLI and Custom Agents

• Using the Kiro CLI to create custom agents for specialized tasks, such as code review, architecture, and compliance.
• Custom agents help maintain context and optimize productivity by focusing on specific domains.
• Launching custom agents using the delegate launch command, or swapping the active agent using the agent swap command.

Key Takeaways

• Kiro's spec-driven workflow and agent-based approach help address key challenges in FinServ AI development, including technical debt, compliance burden, and security debt.
• Agent steering, agent hooks, MCP servers, and the Kiro CLI provide a comprehensive set of tools to improve productivity, enforce best practices, and automate repetitive tasks.
• The integration of policy-as-code and the ability to create custom agents tailored to specific domains are particularly powerful features of the Kiro Agentic IDE.

Conclusion

• Wilso concludes with a quote from Henry Ford, highlighting how Kiro is transforming the financial services industry and beyond through its agentic AI coding capabilities.
• The presentation encourages the audience to explore the Kiro platform and share their experiences, as well as to complete the session survey to provide feedback to the Kiro team.