Matt Ball is a solutions architect at Augment Code who helps engineering teams apply AI coding agents to complex, established software systems. His focus is making those agents understand existing code, operate securely, and produce changes that developers can actually verify.
Ball began his career as a software engineer, led Professional Services Engineering at Qubit, and became Postman’s first solutions architect, helping shape its go-to-market strategy. At Augment, he has brought that developer-tools background to enterprise adoption of AI-assisted software development.
His central concern is deep codebase context. Because most professional software development modifies existing systems, effective agents must retrieve relevant dependencies, architectural conventions, and implementation details. At the AI Engineer World’s Fair, Ball described how agents use that context to identify missing files, plan changes, and navigate unfamiliar legacy code, while distinguishing supervised chat from autonomous and remote execution.
- Tests as agent feedback loops: Executable checks let coding agents evaluate their work and iterate toward reliable changes, especially when modernizing unfamiliar systems.
- Model flexibility without developer burden: Ball initially opposed model pickers that shift product complexity onto developers. As leading models developed different strengths, he advocated multi-model engineering supported by shared context and orchestration.
- Automated code review with measurable value: His coauthored guidance on Auggie CLI for automated code review recommends starting with one team, integrating review into continuous integration, limiting feedback to high-confidence findings, and measuring reviewer time and pull-request throughput.
Ball also emphasizes existing-IDE integration, enterprise security controls, and protecting proprietary code: requirements that determine whether coding agents can move beyond demonstrations into everyday engineering.
