From Code Writers to AI Orchestrators
The future of engineering is a discipline where humans design, supervise, and optimize AI agents that handle coding workflows, shifting the core craft from writing syntax to managing intent, constraints, and machine collaboration across the software lifecycle. In this world of AI agents software development, code is no longer the main output engineers bring to the table. Netlify CTO Dana Lawson points out that code has always been a minority of the job and now becomes the least strategic part as agents generate implementations from plain-language intent. Agentic AI forms a new abstraction layer where conversation becomes a kind of programming language. That shift opens the door to Lawson’s “builder” era, with forecasts of a billion new applications by 2029 powered by agentic AI. Engineers stay central not as typists, but as the people who frame problems, set guardrails, and own production outcomes.
Agent Experience (AX) Becomes the New Core Skill
Agent experience AX engineering treats AI agents as first-class users and collaborators that share the same platforms as humans. Lawson describes AX as the practice of designing where humans and agents collaborate seamlessly, spanning intent capture, workflows, and production behavior. It sits at the intersection of developer experience and user experience: error messages that machines can parse, logs structured for agents, and APIs that match how agents plan work rather than how humans click through dashboards. As Netlify rebuilt its platform for both citizen developers and AI coding agents, clearer error output and less hidden knowledge helped human developers too. Every assumption removed for machines removed friction for people. For engineers, the future of software engineering lies in system thinking: modeling how agents interact with CI/CD, observability, and security so that human oversight focuses on the edges, not every single line of code.
Managing AI Coding Agents: From Speed Trap to System Design
Managing AI coding agents is less about asking them to write more code and more about designing workflows that keep quality and accountability visible. Anthropic’s agentic coding forecast shows agents moving from one-shot helpers to collaborators that write tests, debug, generate documentation, and handle implementation workflows. Yet the same research notes that developers use AI in roughly 60% of their work while fully delegating only up to 20% of tasks. That gap shows why AI supervisors are becoming vital. Tools can compress implementation loops, but poor prompts or missing guardrails can create problems faster than humans can review them. Google Cloud’s DORA framing supports this: AI-assisted development is a systems problem, not a tools problem. Teams need clear escalation rules, automated tests agents cannot bypass, audit trails, and human approval gates, or else speed hides rework and security risks instead of reducing them.
A Billion Apps and the Risk of Building the Wrong Things
The promise of agentic AI is volume: forecasts suggest a billion new applications written by 2029 as more people can build with natural language. That surge will not be limited to engineering; sales, marketing, legal, and operations teams will compose their own tools using AI agents software development platforms. Outcome Engineering’s argument that AI removes human bandwidth as the main constraint changes the job of engineers in a subtle way. When almost anything can be built, judgment about what not to build grows more valuable than implementation skill. Lawson warns that many things built today will be obsolete within months. Engineers become outcome editors and ethical reviewers, deciding which experiments deserve hardening, which automations require extra security, and which agentic workflows should never reach production. The risk is not that agents cannot build enough, but that they build too much of the wrong thing.
Engineering Careers in the Era of AI Supervisors
The future of software engineering looks less like a room full of people typing and more like AI supervisors managing fleets of agents. Anthropic and others expect non-technical roles to use agentic tools, but the hardest problems will still land with engineers: architecture, observability, threat modeling, and policy. Human oversight becomes the scarce resource that determines whether AI accelerates progress or multiplies mistakes. Engineers will define agent permissions, decide which events trigger human review, and tune workflows so that agents handle routine checks while people focus on ambiguous cases. For career resilience, developers should build skills in agent experience AX engineering, systems thinking, and outcome-oriented product judgment. The teams that win will not be the ones that buy the most AI, but those that learn how to orchestrate machines and humans into reliable, auditable, and maintainable software systems.
