Girish Mukim

Posted on Mar 28

BeSA Batch 09 Week6 - Supercharge Development with Kiro | Build Your AI-Enhanced SA Practice

#ai #aws #solutionarchitect #besa

BeSA Batch 09 – Week 6
AI-Driven Development with Kiro and Building an AI-Enhanced SA Practice

Disclaimer:
These notes were drafted using AI for clarity, structure, and readability. They are intended solely for learning purposes.

These are the structured notes from Week 6, focused only on the two role plays. Writing this as a quick revision for those who attended the session and a concise recap for anyone who couldn’t make it.

Role Play 1 – Supercharging Development with Kiro

Context

This conversation focused on AI-driven software development and how tools like Kiro are changing the way applications are built—from idea to production.

Getting Started with AI-Driven Development

To build AI-powered applications, a few foundational components are required:

Infrastructure

Compute layer to run AI workloads
Can include specialized AI hardware options

Foundation models

Serve as the “brain” of the system
Accessed through managed services

Supporting services

Orchestration
Memory
Knowledge bases
Security

Once these are in place, developers can start building applications using tools like Kiro or other AI-assisted development environments.

Traditional vs AI-Optimized Infrastructure

It was clarified that traditional compute (like EC2) can still be used.

However, optimized AI infrastructure provides:

Better performance
Lower cost
Improved scalability
More specialized capabilities

The shift is toward purpose-built AI compute rather than general-purpose infrastructure.

Evolution of AI Development Tools

The progression of AI-assisted development was explained in three stages:

Autocomplete phase (around 2023)

Basic code suggestions
Similar to enhanced IntelliSense

Assistant phase (around 2024)

Interactive AI assistants
Developers could ask questions and get help

Agentic development phase (current)

AI agents actively participate in development
Assist in planning, design, and execution

This represents a shift from assistance → collaboration → partial autonomy.

Benefits of AI-Driven Development

Key benefits highlighted:

Faster time to market

Rapid creation of MVPs
Faster experimentation

Increased productivity

Developers can focus on higher-level problems
AI handles repetitive tasks

Developer autonomy

Agents can take decisions and execute tasks
Developers and agents work together

Improved code quality (with proper guidance)

Structured workflows can lead to better outcomes

Challenges to Be Aware Of

Some limitations and risks were also discussed:

Scaling AI development

No centralized way to manage all knowledge and context

Black box behavior

Limited visibility into how outputs are generated

Control limitations

Hard to fully control agent behavior

Code quality concerns

Outputs need validation

These challenges reinforce the need for structured approaches.

Vibe Coding vs Structured Development

A distinction was made between quick coding approaches and structured development.

Vibe coding

Fast, iterative, “build as you go” approach
Useful for:
- Small UI changes
- Quick fixes
- Simple enhancements

Limitations:

Not scalable
Not suitable for complex systems

Structured (spec-driven) development

Starts with requirements and design
Follows a defined process
Suitable for production systems

Key idea:

Use vibe coding for prototyping
Use structured development for production

Spec-Driven Development with Kiro

Kiro introduces a specification-driven workflow.

Flow:

Requirement → Design → Tasks → Implementation

Instead of jumping directly into coding, the process ensures:

Clear requirements
Defined architecture
Predictable execution

This reduces ambiguity and improves scalability.

Core Concepts in Kiro

Specification

Defines what needs to be built
Acts as a contract

Design

Defines how the system will be built

Tasks

Break down implementation steps

Steering files

Guide the agent during development
Act like constraints or guardrails

Key insight:

Clear specifications reduce back-and-forth with AI
Provide better control over outcomes

Agentic Development Environment

Kiro acts as an agentic development environment where:

AI understands relationships between files
Changes propagate across requirements, design, and code
Context is maintained across the system

This enables:

Consistency
Better traceability
Faster iteration

Pricing Model

The pricing model consists of:

Subscription cost
Usage-based cost (based on model usage/credits)

Higher-capability models cost more, while lighter models are cheaper.

Role Play 2 – Building Your AI-Enhanced SA Practice

Context

This conversation focused on how solutions architects can systematically integrate AI into their daily workflow and build a consistent AI-enabled practice.

From Occasional Use to Systematic Practice

A key shift discussed:

From:

Using AI occasionally

To:

Building a consistent AI-driven workflow

This transition enables compounding benefits over time.

AI Toolkit Categories

A structured toolkit approach was recommended.

Four categories:

General-purpose AI tools

Used for broad tasks like research and drafting

Development tools

Integrated into IDEs
Used for coding and architecture generation

Specialized tools

Example: diagram generation tools
Convert text into structured outputs

Enterprise AI tools

Used for sensitive data
Include governance and compliance controls

Choosing the right tool depends on:

Task type
Data sensitivity
Context

Daily Workflow Integration

A structured daily workflow was outlined.

Morning briefing

Provide AI with context for the day
Set expectations

Pre-meeting preparation

Research customer
Understand industry and tech stack

During engagement

Quick lookups
Cost comparisons
Service evaluations

Post-meeting analysis

Paste notes into AI
Extract:
- Requirements
- Risks
- Action items

Periodic review

Identify patterns across engagements
Improve approach over time

This creates a continuous feedback loop.

Prompt Library

A key practice is maintaining a reusable prompt library.

Common categories:

Discovery questions
Architecture generation
Cost analysis
Security reviews
Documentation templates

Benefits:

Saves time
Standardizes quality
Enables team-wide improvement

End-to-End Use of AI in Engagements

Example workflow for a new customer:

Discovery phase

AI-assisted research
Faster understanding of customer context

Post-call analysis

Extract structured insights from notes

Design phase

Generate multiple architecture options

Refinement phase

Apply human judgment
Adapt to real constraints

AI typically gets the solution:

60–70% complete

Human expertise takes it to:

90%+ completeness

Where Human Judgment Matters

Three key areas where AI cannot replace architects:

Understanding human context

Team dynamics
Risk tolerance
Organizational constraints

Making trade-offs

Decisions under uncertainty
Balancing competing priorities

Communication strategy

Tailoring message for stakeholders
Deciding what to present and how

Key idea:
AI can generate content, but cannot decide strategy.

Responsible Use of AI

Guidelines for safe usage:

Avoid sharing sensitive or regulated data in public tools
Use anonymization where required
Prefer enterprise-approved tools for customer data

Three key checks:

Is the data public?
Is it customer-identifiable?
Is it regulated?

Transparency with Customers

Best practice:

Be transparent about using AI
Position it as part of your workflow
Emphasize validation and judgment

Focus should remain on:

Quality of recommendations
Customer outcomes

Action Plan

Simple steps to get started:

Pick one workflow and integrate AI
Build a prompt library
Engage with a learning community

Mindset Shift

Final takeaway:

AI does not replace core strengths of an architect.

Key differentiators remain:

Judgment
Customer relationships
Ability to synthesize complexity

AI amplifies these strengths rather than replacing them.

Week 6 Consolidated Takeaways

From the first role play:
AI-driven development is evolving toward agentic environments where structured, specification-driven workflows improve quality, scalability, and speed.

From the second role play:
Building a consistent AI-enabled workflow is critical for architects, with human judgment remaining the key differentiator in decision-making and communication.

This final week brought together two important themes: