For a decade, the gold standard of a skilled software engineer was clean, efficient code. But according to Netlify Chief Technology Officer Dana Lawson, that benchmark is becoming obsolete. Writing code, she argues, is no longer the job — and the sooner engineers internalize that shift, the better positioned they will be for what comes next.
Lawson described a fundamental transition underway in how software gets built. As AI-powered coding agents become capable of generating, testing, and deploying code autonomously, the engineer's primary value proposition moves upstream: from execution to judgment, from writing to orchestration.
She introduced the concept of "agent experience" (AX) as the new core competency for developers. Where user experience (UX) describes how humans interact with software, AX refers to how AI agents interact with the systems and tools that developers build and manage. Lawson's argument is that the next generation of impactful engineers will be those who understand how to design environments in which AI agents can operate effectively — defining goals, constraints, feedback loops, and fallback logic.
The underlying premise is that we are entering what Lawson calls a "billion-agent economy" — a world where AI agents handle vast amounts of software development work autonomously, coordinating across APIs, repositories, test suites, and deployment pipelines. Engineers who can architect and oversee these agent-driven workflows will command outsized influence in that future.
This framing aligns with a broader pattern visible at leading technology organizations. Anthropic recently reported that approximately 80 percent of production code at certain organizations is now generated by Claude AI. OpenAI has expanded Codex into an enterprise-grade agent platform. GitHub Copilot is evolving from an autocomplete tool into an agentic pair programmer capable of handling full tasks end-to-end.
Critics have pushed back, noting that AI agents still hallucinate, make architecture mistakes, and require significant human oversight — especially for complex, high-stakes systems. Debugging AI-generated code can sometimes be harder than writing correct code from scratch. Still, the trajectory is unmistakable: the share of routine coding handled by AI is growing rapidly.
For engineers feeling uncertain about this transition, Lawson's message is less about displacement and more about elevation. The skills that remain irreplaceable — systems thinking, domain expertise, understanding user needs, navigating organizational constraints — are exactly what makes a strong human overseer of AI agent pipelines.
Why It Matters
Companies building developer tooling, hiring engineering talent, and setting product roadmaps are actively restructuring roles around AI agent workflows. Organizations that reorient their engineering teams as agent orchestrators rather than line-by-line coders will likely move faster with leaner headcount. For individual engineers, the practical implication is clear: investing in understanding how to evaluate, direct, and improve AI-generated work is now as strategically important as deep technical expertise in any single language or framework.
Source: The New Stack