From Writing Code to Orchestrating AI-Driven Development
AI agents in software development are specialized autonomous tools that convert intent into working software by owning discrete stages of the delivery lifecycle, from requirements and coding to testing and release, shifting engineers’ focus from writing code toward orchestrating and validating automated workflows at scale. This shift sits at the heart of an AI-driven development workflow, where natural language instructions trigger coordinated agent activity rather than manual coding. Instead of starting every feature from scratch, engineers define business outcomes, align them with technical constraints, and configure agent pipelines to achieve them. As Netlify’s Dana Lawson argues, writing code has long been a minority slice of engineering work, and now the least strategic slice is the first to be automated. What remains is high-impact decision-making: understanding systems, risks, and trade-offs while AI agents execute the repetitive implementation details.
Endava’s Agent Network and the Automated Software Delivery Pipeline
Endava’s approach to automated software delivery shows how far AI agents can reach into the full lifecycle. Instead of a single coding assistant, Endava is building a modular network of specialized agents, each responsible for a specific phase: one turns raw business requirements into user stories and functional specs, another generates boilerplate logic and executes unit tests, a third quietly reviews pull requests for vulnerabilities and formatting problems before humans ever see them. Teams string these agents into custom workflows, so a web project might chain frontend, API testing, and accessibility agents, while data engineers assemble agents for pipeline creation and schema validation. In this model of AI agents software development, engineers kick off work and then supervise a sequence of agents that carry it to completion. Their responsibilities move toward choosing the right workflow, tuning prompts, and confirming that outcomes meet quality, security, and compliance standards.
Agent Experience (AX): The New Frontier for Engineering
As agentic platforms mature, agent experience engineering is emerging as a discipline on par with developer experience and user experience. Netlify has rebuilt its platform to speak not only to traditional developers but also to AI agents and non-technical “builders” who may not know what git is. According to Netlify CTO Dana Lawson, agent experience is “the practice of designing where humans and agents collaborate seamlessly — and it isn’t just about making API calls agent-friendly.” Clear agent error messages, machine-structured build output, and fewer hidden assumptions improve both AX and UX. Engineers now design interfaces, events, and feedback loops that agents can consume reliably, while making those same flows understandable to humans. This forces teams to expose tribal knowledge, implicit workflows, and unclear internal APIs, replacing them with explicit contracts and well-documented pathways that both software agents and people can follow.
A Billion New Apps and the Builder–Orchestrator Divide
Agentic AI is lowering the barrier to creation so far that Lawson predicts “there will be a billion new applications written by 2029 because AI enables what [she] calls the builder.” In this world, many more people can describe what they want and have AI agents assemble it. That explosion of output reshapes the developer role transformation: engineers become orchestrators, outcome designers, and quality gatekeepers. Their work shifts toward deciding what not to build, setting constraints, and protecting production from runaway complexity or insecure patterns. While citizen developers and agents handle composition, engineers safeguard business goals, security posture, and environmental impact. They must decide which intentions deserve durable systems versus quick experiments that may be obsolete in months. Success depends less on typing speed and more on judgment, architecture, and an ability to manage risk in a landscape where human bandwidth is no longer the main limit.
Engineering for AX: New Skills, Stacks, and Responsibilities
For AI-driven development workflow patterns to be reliable, engineering teams must rebuild systems around agentic intent. Enterprise stacks were designed for humans in the loop; now, agents need consistent signals, schemas, and events across services that speak different API dialects. Engineers design those connective tissues and guardrails. They define event models that trigger agent workflows, standardize interfaces so agents can call them without brittle prompt engineering, and expose routes to production in ways agents can follow safely. Lawson describes engineers as the “shepherd of production,” responsible for what goes in and out. That includes thinking about resource use, compression, and compaction so the internet remains open and environmentally responsible even as a billion new applications appear. The job centers on systems thinking, collaboration design, and continuous verification, transforming engineering from handcrafting code into owning the quality, ethics, and sustainability of automated software delivery.
