From Coding to Agent Experience: A New Definition of Engineering
Agent experience engineering is the practice of designing, supervising, and improving how humans and AI coding agents collaborate across the entire software delivery lifecycle. It treats conversation and intent as the new interface for building systems, shifting the engineer’s focus from hand-writing code toward orchestrating machine collaborators. Netlify CTO Dana Lawson argues that writing code was never the main job of engineers, and in an era of agentic AI, it is now the least strategic part. Instead, engineers become “shepherds of production,” responsible for what goes into and comes out of systems that agents help build. As agentic AI establishes a new layer of abstraction, intent expressed in natural language turns into the next programming language. This redefinition means engineering skill now centers on system context, safety, and collaboration design, not just syntax and frameworks.
Agent Experience (AX) as the Differentiating Skill
Agent experience is emerging as the skill that separates average engineers from those who can thrive in the future of software development. Lawson describes AX as the union of developer experience and user experience: it covers everything from clearer error messages for agents to build outputs structured for machines. When Netlify redesigned its platform for AI agents and new “builders” who do not know git, the same changes made life easier for professional developers, because every removed human assumption lowered friction for both. AX requires engineers to understand complex systems, the route to production, and business goals so that agents operate within safe and meaningful boundaries. In this model, engineers decide what not to build, filter noise from signal, and ensure that the new billion applications expected by 2029 create value instead of clutter.
Why Speed-Only AI Adoption Backfires
Agentic AI adoption often fails when leaders chase speed instead of management practices. Anthropic’s agentic coding forecast frames AI coding agents as junior collaborators that compress implementation, testing, documentation, and iteration into tighter loops. However, its internal research shows that developers use AI for roughly 60% of their work yet fully delegate only up to 20% of tasks, which means humans still carry most of the responsibility. Treating agents like autopilot risks confusing automation with abdication. Productivity outcomes also vary: GitHub found developers using Copilot completed a JavaScript task 55% faster, while METR reported experienced developers became 19% slower on familiar repositories with early 2025 tools. These mixed results show that workflow design, verification costs, and expertise matter. Without clear oversight, an employee who prompts an agent poorly can produce problems faster than traditional teams can review them.
The Rise of AI Supervisors and Agent-Oriented Teams
As agentic AI adoption spreads, the most constrained resource is no longer compute or headcount but reliable human oversight. Anthropic expects teams to move from reviewing everything to reviewing what matters, while agents handle routine checks and escalate boundary cases. This shift is creating a new role pattern: AI supervisors who define goals, constrain agents, inspect outputs, test outcomes, and decide when to stop. These supervisors may sit in engineering, product, legal, marketing, or operations, but they share a core responsibility for safe delegation. To support them, organizations need explicit escalation rules, human approval gates for sensitive actions, automated tests that agents cannot bypass, and audit trails for accountability. Stack Overflow’s 2025 survey found that 84% of respondents use or plan to use AI tools, yet more developers distrust than trust their accuracy, which underscores the need for structured supervision.
Designing Careers and Hiring for an Agentic Future
Agentic AI is reshaping hiring criteria and team structure. Instead of optimizing for who can write the most code, companies now need people who can design agent workflows, enforce guardrails, and align outcomes with business context. This is where agent experience engineering becomes a core competency: engineers define event-driven agent behaviors, choose what to automate, and decide where human judgment stays in the loop. Netlify’s experience shows that building for citizen developers and agents forces platforms to remove institutional and domain-knowledge barriers, broadening who can build software. In parallel, Anthropic’s forecast suggests AI coding agents will grow from one-shot helpers into deeper collaborators that handle end-to-end implementation workflows. To stay relevant, engineers should cultivate skills in prompt design, system architecture, verification, and AI supervision. The future engineer is less a solo coder and more a conductor of a fast, machine-powered team.
