Chasing Jarvis: Can Technically Sophisticated Non-Programmers Deploy SaaS Applications Using AI Coding Agents?

This empirical study investigates whether technically sophisticated non-programmers—domain experts with substantial technology experience but without traditional programming backgrounds—can successfully build and deploy production-grade Software-as-a Service (SaaS) applications using AI coding agents, addressing a central question in the ongoing debate about AI's impact on software development accessibility. Through a two-year participatory action research approach (2023–2025), combining systematic tool experimentation with the development and deployment of Lumina AI—a production RAG chatbot widget platform built entirely through AI-assisted development—this study provides empirical evidence on the capabilities and limitations of contemporary coding agents. The research synthesizes findings from 80+ peer-reviewed sources, industry reports, and technical documentation alongside first-hand development experience across multiple AI coding platforms (Cline, Claude Code, Augment Code, Google AI Studio, Cursor). Results demonstrate that technically sophisticated non-programmers can indeed achieve production deployment, though success depends critically on mastering 'context engineering'—the systematic structuring of information environments for AI systems—rather than traditional programming skills.

Keywords: artificial intelligence, coding agents, context engineering, software development democratization, entrepreneurship, SaaS deployment