AI Will Create More Engineering Work, Not Less
In How GenAI Is Transforming Software Engineering, But Not Replacing It, Sr. Director of Global Business Innovation at Syntax, Matthias Steiner offers a comforting perspective for software engineers (which could have saved me a few sleepless nights had I read it in December). While many, including myself, have misinterpreted the AI boom as an existential threat to software engineering, Steiner argues the opposite.
The reality is that AI makes the mechanical work of software construction light work, but not the craft of software engineering. Engineers are still responsible for architecture, pattern adoption and reliability, domain modeling, security, scalability, reliability, systems integration, and the alignment of technical decisions with business outcomes."
The good news according to Steiner is that the need for more software will only increase the need for more engineers to manage it. This is "Jevons' paradox: increasing efficiency expands demand." In other words, AI multiplies the amount of software and therefore the amount of software engineering.
The point is that "writing code is only one step of the software development lifecycle (SDLC)," which typically includes seven phases: planning, analysis, design, implementation, testing, deployment, and maintenance. Steiner points out that, every phase of the software development lifecycle benefits from AI: "Early-stage requirements can be transformed into structured specifications; Architectural diagrams can be drafted or critiqued; Test suites can be auto‑generated; CI/CD workflows can be orchestrated; Logs and incidents can be summarized and diagnosed; Legacy code can be refactored into modern patterns"
One approach that I've been experimenting with lately is spec-driven development, "where a structured specification becomes the system’s single source of truth." It has added a level of control that I didn't have when purely vibe-coding. No doubt, AI can still be unwieldy.
For many, the fun part of software development has always been writing the code. AI takes that part and makes it more efficient. That's pretty unfortunate for the engineers who love writing code, but the good news is that there is still plenty to do.
Via The New Stack