Rethinking Software Development and AI
Software development is a design process not a manufacturing one. The discussion over agentic coding confuses or gets that wrong. This, in part, explains why so much investment in AI adoption fails to generate value.
When leaders think of software as a manufacturing problem, they focus AI adoption on code generation.
But software development is discovering the right solution under uncertainty. So, focusing on code generation comes at the expense of the learning that developers engage in before and while they code.
In an environment that doesn’t already have excellent user centered design, engineering, and platform discipline, automation speeds up output while degrading judgment. You get more of the wrong things faster.
Lean software development practice derives from set based design. Which uses creative problem solving within known constraints to optimize downstream cost and customer value. This insight dates back to Mary Poppendieck, who introduced lean concepts to software in the first place.
When leaders realize software is a set based design challenge then they use AI to shorten learning cycles. They reduce waste to accelerate time to value.
This means using AI to help teams gain better insight. To find and communicate the constraints. To model solutions within those constraints. To prototype multiple approaches. To give the team early and regular customer feedback. To automate guardrails and deployment processes. To increase the quality and breadth of testing. To instrument applications to track how they are used and where they are breaking down.
AI is an opportunity to radically increase team and customer collaboration. We can use speed gains not to bloat the amount of code that makes it into production but to improve the value of the code that does.
#SoftwareDevelopment #AgenticAI #LeanSoftware #ProductDiscovery #AIAdoption #SoftwareEngineering #TechLeadership
