What Happens When Every Developer Has 20 AI Agents?

The rapid integration of AI into software development is fundamentally altering productivity while simultaneously creating significant infrastructure strain. Although some suggest that SaaS is obsolete, the high costs of maintaining bespoke internal systems make established platforms essential for operational stability. The Model Context Protocol (MCP) has emerged as a critical standard, achieving meteoric industry adoption by enabling AI models to securely access private data sources. However, the ability to generate code at high velocity creates massive downstream bottlenecks in code review, security, and deployment. As developers shift toward building numerous individual AI agents, organizations face a looming crisis in governance and maintenance, forcing a transition from an initial burst of experimental creativity toward more sustainable, centralized management of these new software paradigms. Stephen O'Grady, co-founder of RedMonk, provides these insights based on his research into developer tooling and industry trends.