Lately, my day-to-day world in software engineering has shifted more toward architecture and leadership than hands-on delivery. But over the Christmas break, I had the chance to get closer to the ground again—offering some friendly guidance to help build a SaaS product from scratch. No legacy systems, no existing processes—just a blank canvas and a big idea.

A former colleague of mine, Edwin, who was deep in the build, asked a familiar question: “Where should I start? Do we do this the same way we always have—Agile, sprints, ceremonies?” I paused, thought it through, and said, “Honestly, no mate. You should try AI-DLC.” He laughed and replied, “I know it has something to do with AI… but what exactly is it?”

That question led to several long conversations, mapping AI-DLC onto his project and rethinking what modern software development could look like when AI isn’t just a supporting tool, but a first-class participant in how software is designed, built, and operated. This article is my attempt to share that thinking—for anyone ready to step outside the old playbook and explore what’s next.

What Is AWS AI-DLC?

AWS AI-DLC (AI-Driven Development Lifecycle) is an AI-native approach to software development that fundamentally rethinks how systems are designed, built, and operated. Instead of treating AI as a supporting tool used at isolated stages, AI-DLC embeds intelligence across the entire lifecycle—from ideation and requirements to implementation, testing, deployment, and operations. AI becomes an active participant in delivery, continuously assisting with decisions, execution, and optimisation.

Crucially, AI-DLC preserves a strong human-in-the-loop model. Engineers and architects define intent, apply domain context, validate outcomes, and retain ownership of production systems, while AI accelerates execution, reduces manual effort, and enforces consistency at scale. The result is a development lifecycle where humans focus on strategy and judgment, and AI handles orchestration—enabling teams to deliver faster, with higher quality and confidence.

AI-DLC deliberately keeps humans in the driver’s seat, with explicit approval gates for critical decisions, while AI accelerates the work that happens in between.
That’s why AI-DLC isn’t vibe coding. It delivers speed without sacrificing control.

How do we use AI-DLC?

Just as you interact with ChatGPT by asking a question, you can initiate an AI-DLC workflow by simply stating your intent. AI-DLC follows a simple yet powerful pattern—one that consistently repeats across every phase of the development lifecycle.

I want to start an AI-DLC project for an event management application.

From there, the AI takes over the orchestration. It automatically identifies whether the project is greenfield (new) or brownfield (existing), guides you through the appropriate lifecycle phases, asks clarifying questions to refine requirements, seeks explicit approval before advancing, and continuously tracks progress throughout the journey.

AI-DLC Core Framework

AWS has formally introduced the AI-DLC methodology and its core philosophy here. Let’s explore the key components that define it.

Inception Phase

This is the planning and definition stage where the project kicks off.

• Roles: Product Owner, Developers, and AI.

• Key Ritual: Mob Elaboration (Collaborative defining of requirements).

• Process:

  • Starts with a high-level Intent (Goal).
  • Breaks down the intent into smaller working pieces (labeled Unit 1, Unit 2... Unit n).

• Artefacts (Outputs):

  • PRFAQs (Press Release / Frequently Asked Questions)
  • User Stories
  • NFRs (Non-Functional Requirements).
  • Risks.

Construction Phase

This is the building and testing stage where the "Units" are turned into functional software components.

• Roles:

  • For Construction: Developers and AI.
  • For Testing: Product Owner, Developers, and AI.

• Key Rituals: Mob Construction and Mob Testing

• Process:

  • Involves iterative design and coding cycles.
  • Transforms "Units" into "Bolts" (Bolt 1, Bolt 2... Bolt n).

• Artefacts (Outputs):

  • Domain Design, Logical Design, and Codes.
  • Deployment Units that are: Secured, Instrumented, De-Risked, Tested, and Packaged.

Operation Phase

This is the deployment stage where the software goes live.

• Roles: Product Owner, Developers, and AI.

• Process: The finalized "Bolts" are moved into the production environment.

• Artefacts (Outputs):

  • Deployment Units active in the Production Environment.

The Five Critical Artefacts of AI-DLC

Imagine you’re building a modern event management platform—one that allows organisers to create events, sell tickets, manage capacity in real time, and control entry seamlessly on event day. This is how AI-DLC’s five core artefacts come together in practice.

Intent: The North Star

What it is
Intent captures the why—the business goal or outcome that anchors everything that follows. It is the single source of truth that AI uses to guide decomposition, prioritisation, and decision-making.

Example Intent

“Build a scalable event management application that enables organisers to create and manage events, sell tickets securely, control venue capacity, and handle fast, fraud-free entry on event day.”

This Intent becomes the guiding compass for AI throughout the lifecycle, ensuring every feature, design choice, and deployment decision traces back to this core objective.

Unit: The Self-Contained Value Block

What it is
A Unit is a cohesive, independently valuable slice of functionality. AI decomposes the Intent into Units that deliver measurable outcomes and can be built, tested, and deployed in isolation.

Example Units
From the event management Intent, AI may derive Units such as:

• Event Creation & Configuration

• Ticket Sales & Payments

• Capacity & Availability Management

• Event Day Entry & Check-In

• Reporting & Analytics

Each Unit is loosely coupled, allowing teams to move fast without waiting on other components—perfect for parallel development and incremental releases.

Bolt: The High-Velocity Execution Cycle

What it is
Bolts represent the smallest execution loop in AI-DLC. They focus on rapid, tangible progress—measured in hours or days rather than traditional multi-week sprints.

Example Bolts (for the “Event Day Entry & Check-In” Unit)

• Day 1: Implement QR code–based ticket validation

• Day 2: Add offline check-in support for poor connectivity

• Day 3: Introduce real-time entry count and capacity alerts

AI plans these Bolts to maximise delivery speed while keeping scope sharply focused, with humans validating quality and user experience.

Domain Design: The Business Logic Blueprint

What it is
Domain Design captures the heart of your business logic using domain-driven design principles—independent of infrastructure or deployment concerns.

Example Domain Model (Event Management)

• Entities: Event, Ticket, Attendee, Venue

• Value Objects: TicketType, CapacityLimit, EntryStatus

• Aggregates: EventSession (enforcing capacity rules)

• Domain Events: TicketPurchased, CapacityReached, AttendeeCheckedIn

• Repositories: EventRepository, TicketRepository

AI then extends this into Logical Design, recommending architectural patterns (for example, event-driven processing for entry scans) and documenting decisions in Architecture Decision Records (ADRs) for human review and approval.

Deployment Units: Ready-to-Ship Deliverables

What it is
Deployment Units are fully operational, production-ready packages—combining application code, configuration, infrastructure definitions, and automated tests.

For the event platform, this could include:

• Ticketing and payment services

• Entry scanning APIs

• Real-time capacity monitoring components

• CI/CD pipelines and infrastructure templates

Each Deployment Unit is validated for scalability, reliability, and security—ensuring the system performs flawlessly even during peak event-day traffic.

Closing Thoughts

Edwin and I caught up again on a warm midsummer evening, beers in hand, somewhere in town. By then, his project was already past the halfway mark. What stood out wasn’t just the progress, but how smoothly it was moving. No long pauses, no painful rework—just steady momentum. Choosing AI-DLC over the usual Waterfall or even Agile approach paid off. Features were landing fast, working software was in users’ hands, and milestone after milestone quietly ticked itself off.

If I were starting a new software product today, I wouldn’t think twice—I’d use AI-DLC. Not because it’s trendy, but because it matches where the industry is clearly heading. Humans still set the vision, apply judgment, and take ownership of outcomes. AI simply accelerates the journey from idea to reality. To me, AI-DLC feels like the natural evolution of SDLC—one that leans into the future instead of trying to hold it back.