Explaining the AI Paradox

Why Your AI Strategy Is Failing: The AI Paradox of Optimizing Coding Alone - Andrew Haschka

Topic 4 of 19

Published May 18, 2026 · Clip 9:47–14:33 (4:46)

Andrew explains the AI paradox: organizations optimize AI for coding alone, but coding accounts for less than 20% of engineering time, so accelerating it pushes bottlenecks into testing, security, compliance, and validation. He describes GitLab's vision of intelligent orchestration, where developers operate above the loop guiding AI agents that perform tasks autonomously. He details the experience layer, control layer, and data layer, emphasizing that only a unified knowledge graph across all SDLC phases can solve the paradox. He cites Bendigo and Adelaide Bank consolidating 1,500 projects into one system of record in four weeks.