Actually Implementing AI

This article discusses the practical implementation of AI in software development, moving beyond hype and marketing to focus on real-world delivery. It highlights insights from consultant Enrico Teotti, emphasizing that the most useful AI work is often mundane, like combining application code, database structure, analytics, and performance data to answer complex questions. The article stresses the importance of read-only access, narrow permissions, and human oversight to avoid risks. It covers AI agentic workflows, software engineering, product management, and security, offering grounded advice for teams adopting AI responsibly.