From Assembly Lines to AI Agents: Rethinking Software Delivery Organizations
Former T-Bank technical director Alexander Polomodov breaks down how AI agents are dismantling the Taylorist, pipeline-style organization of software delivery — collapsing specialized roles like frontend, backend, QA, and analyst into leaner "product engineer" and "discovery" pairings supported by agents. He explores the organizational friction this creates in large companies (platform teams vs. business verticals, legacy tooling built for humans not agents), why spec-driven development and domain-precise language dramatically improve agent output, and why early-adopter advantages in custom internal tooling are now becoming liabilities. The conversation closes on harder open questions: how to train juniors and hire engineers when old signals of competence are easily faked by agents, why agents excel at well-specified stateless systems but struggle with stateful legacy and physical-world systems, and how to measure whether AI adoption is actually creating business value versus just generating more code.