Full recording of 2m2x Ep. 146, The Next Evolution of AI Development
Still using AI just to write code? That’s only the beginning. In this episode, we break down the four phases of AI-enabled software development—from coding and testing to workflow orchestration and product intelligence. The real competitive advantage comes when AI shapes the entire development lifecycle, not just the codebase.
Still Using AI Just to Write Code? That’s Only the Beginning.
Most software teams have adopted AI. Almost none of them have moved past the first phase.
There is a common and costly pattern: a team adopts an AI coding assistant, sees some productivity gains, and concludes they have an AI strategy. What they have actually done is optimize one narrow slice of a much larger system — and left the majority of the value untouched.
Real AI-enabled software development follows four phases. Most teams are stuck in Phase 1. Here is what it looks like when you move through all four.
Phase 1: Coding — The Entry Point, Not the Destination
AI-assisted coding is where almost every team begins. It is visible, measurable, and easy to justify. Developers move faster, boilerplate shrinks, and context switching improves.
But the ceiling here is easy to miss. Getting consistent quality requires well-architected context engineering and opinionated frameworks. Even when done well, you are still performing local optimization — improving one piece of the system while everything around it stays the same.
“If you are only using AI for coding, you are optimizing the smallest part of the system.”
Phase 2: Testing — AI Starts to Validate
In Phase 2, AI shifts from builder to validator. Unit test generation, test case expansion, functional test simulation. AI is no longer just writing code — it is checking it.
This reduces defects and tightens feedback loops. But testing still operates within a silo. The broader flow of how software gets planned and shipped has not changed yet.
Phase 3: Workflow Orchestration — The Stage Most Teams Never Reach
Phase 3 is where AI begins connecting the system rather than operating in isolated parts of it. In practice this looks like:
• Meeting transcripts automatically generating user stories
• Tickets written and linked without manual handoffs
• Code changes connected to test coverage
• Gaps in user journeys flagged before they become production issues
This is the shift from AI as a collection of tools to AI-enabled workflows. The connective tissue between planning, development, and testing starts to form.
Phase 4: Product Intelligence — AI Shapes What Gets Built
Phase 4 is where AI moves into product thinking itself, not just execution. At this stage AI can:
• Review usage reports and vision documents together, surfacing patterns humans miss
• Prioritize features based on data rather than gut feel
• Simulate outcomes before committing to build them
• Design validation experiments to test assumptions early
This is a qualitative shift. AI is no longer accelerating how work gets done — it is influencing what work gets done in the first place.
Competitive Advantage Is in the Span
The compounding value of AI in software development does not come from any single phase. It comes from AI spanning code quality, workflows, and product decisions — operating across the full system. Each phase builds on the last, and the organizations that move through all four will carry a structural advantage over those that do not.
Most teams are sitting on Phases 2, 3, and 4 without realizing it. The path forward is clearer than it looks. If you need help getting there, reach out to Informulate.
