AI drives software productivity – and challenges – for Motorway | Computer Weekly

Motorway has achieved a four-fold increase in its engineering output through the adoption of an AI-first development pipeline, centered around AWS Kiro. This agentic AI system goes beyond simple autocomplete, understanding Motorway's continuous integration pipelines, infrastructure, and internal applications to guide the development process. Ryan Cormack, Motorway's principal engineer, notes that Kiro's model agnosticism is a key factor, enabling the system to utilize various large language models, including Claude's Opus 4.7 and Meta's Llama models, to select the most effective one for specific tasks, thereby reducing hallucinations. The company has shifted to a spec-driven approach where AI first conceptualizes technical designs and generates automated tests before writing any code, ensuring self-validation. While this drastically improves efficiency, it creates a 'volume crisis' in code review. Motorway has counteracted this by redirecting human effort to intensive upfront planning and final review, thus optimizing the 'manual middle' of the development process. This transformation redefines the role of developers, moving them from operational tasks to more strategic oversight, problem-solving, and enhanced communication. The new development philosophy views code as more disposable, prioritizing rapid iteration over the meticulous, long-term maintenance of every line.