From Coding to Curating: A New Definition of Engineering
The engineer’s new job is to design, supervise, and refine networks of AI agents so that intent expressed in natural language becomes reliable, safe software behavior at scale. This shift means writing code is no longer the main activity, even though it remains a useful support skill. Netlify CTO Dana Lawson points out that coding was always a small slice of the work, often less than a quarter of the role, and rarely the most strategic part. Instead, engineers now curate systems: they define constraints, clarify architecture, and decide which problems are worth solving in the first place. As agentic AI takes over routine tasks like scaffolding features or generating tests, career security depends less on syntax expertise and more on AI management skills, systems judgment, and the ability to turn fuzzy human intent into clear, auditable workflows for machines and people.
Agent Experience Engineering Becomes the Core Skill
Agent experience engineering (AX) treats AI agents as first-class users of your platform alongside humans. Lawson describes AX as the practice of designing where humans and agents collaborate seamlessly across the entire software delivery lifecycle. It blends developer experience and user experience: clearer error messages, machine-structured build output, and fewer hidden assumptions all make systems more agent-legible and easier for people, too. Netlify’s own rebuild to serve citizen developers and AI agents shows how agentic AI adoption can force teams to clarify architecture, signals, and responsibilities. According to Dana Lawson, “the judgment to build systems that agents can work with, that is the engineering skill of the future.” In this view, engineers stop thinking only in API calls and start thinking in capabilities like create_a_site or deploy_repository, so agents can act based on intent rather than stitching together low-level endpoints.
The Machine Management Race and the Risk of AI Chaos
As agentic AI spreads, companies are shifting from a hiring race for more developers to a machine management race for better-controlled AI systems. A billion new applications are expected by 2029, enabled by AI that lets almost anyone become a builder. That scale makes AI management skills critical: engineers must decide what not to build, because many new apps will be obsolete within months. Lawson warns that without strong engineering practices and clear intent, teams will drown in experiments and fail to see real agentic AI adoption. Success depends on rethinking internal stacks, processes, and ownership boundaries so agents can be trusted participants in production. Engineers become shepherds of production, overseeing what goes in and out, and designing guardrails that keep autonomous activity aligned with business goals, security constraints, and environmental impact, instead of chasing speed at the expense of understanding.
Designing Systems for Human-Agent Loops, Not Hand-offs
In the AX era, software is built around continuous human-agent loops rather than hand-offs between teams. Systems evolve from APIs that expose low-level methods to capabilities that encode intent, from request-response workflows to event-driven streams that agents subscribe to, and from merely machine-readable formats to full architectural blueprints that are agent-legible. Agents write code, generate tests, detect faults, propose fixes, and open pull requests, while humans supply judgment, taste, and business context. To keep this safe, Netlify grounds its AX practice in strict boundaries: each agent runs in a sandbox, human-in-the-loop is the default for meaningful changes, and every action is logged for instant rollback. Lawson frames the core trust question bluntly: “If you can’t explain what the agent did, why would you trust it in production?” The engineer’s value lies in ensuring those explanations are always possible.
