EY hit 4x coding productivity by connecting AI agents to engineering standards

Coding agents offer rapid code generation, but often produce unusable code that needs extensive rework. EY addressed this by integrating agents with their engineering standards and compliance frameworks. This resulted in a significant productivity increase of 4x to 5x across their product development teams. This success involved an 18-24 month process focusing on cultural shifts and technical integrations to enable semi-autonomous coding. Initially, EY promoted adoption with assistive tools, empowering engineers through organic adoption rather than top-down mandates. Realizing agents needed access to their code repositories and standards, EY created a comprehensive "context universe". Evaluating multiple agent platforms, they allowed developers to choose their preferred tool, with Factory emerging as the popular choice. The team then implemented a workload classification framework, dividing tasks into those suitable for high autonomy vs. those needing human oversight. EY shifted developer roles, transforming them into orchestrators guiding the agents. The integration of security and compliance allowed EY to measure early efficiency gains varying from 15% to 60% across different user roles. The horizon model development with semi-autonomous agent execution and orchestrator teams was a key component of their success. Ultimately, the productivity gains resulted from a combination of experimentation, cultural shifts, and behavioral changes in how developers approach their work.