AI tools are transforming software development unevenly, excelling at execution tasks 'right of code' (coding, testing, deployment) where inputs are structured and outcomes verifiable, while struggling with strategic work 'left of code' (product strategy, architecture decisions, user research) due to its ambiguous, context-dependent, and socially-driven nature. The divide isn't about model capability but the fundamental nature of each type of work. Moving AI leftward requires solving 'organizational intelligence': persistent memory, scenario-based reasoning, judgment under uncertainty, and understanding of informal power structures. Early attempts include tools like Granola for meeting intelligence and Limitless for ambient context capture. The left/right boundary will eventually shift, but until AI can navigate organizational complexity without flattening it, strategic work remains human territory.
Table of contents
The uneven AI takeover in software developmentThe axis: left and right of codeAI delivers deep value right of codeAI stutters on the left sideWhat separates valuable AI from impressive AIHow humans and AI work together todayThe organizational intelligence challenge aheadConclusionSort: