Government software development pipelines are increasingly integrating AI coding assistants, but Air Force Sustainment Center Software Directorate CTO Kurt Jarvis said success depends on improving how developers use existing code, not just generating new code faster.

Speaking with GovCIO Media & Research at the Carahsoft DevSecOps Conference, Jarvis outlined an engineer-first approach that uses retrieval-augmented generation (RAG), vector databases and established DevSecOps pipelines to safely accelerate software development.

Jarvis said one of the biggest opportunities for AI is helping developers discover and reuse decades of trusted software already maintained by the Air Force. The goal, he said, is to make AI a code‑reuse engine, not a code‑generation machine.

He also emphasized that AI has not changed the Air Force's software engineering standards. Existing CI/CD pipelines are still the foundation for evaluating AI-generated code. AI can accelerate development, Jarvis said, but engineering discipline and human oversight remain essential to producing mission-ready software.