Software factories return to life, this time powered by AI agents

Software factories are making a comeback, but not as a management buzzword; they are being deployed as an engineering infrastructure that the companies leading the agent economy are currently implementing. According to a ZDNET report, Anthropic, Cognition, Cursor, Factory, Google, GitHub, OpenAI and Ramp all arrived at the same architecture exactly, independently and within a year and a half.
The term first circulated in 2008 when Microsoft talked about component-based, repeatable software production, but at that time the code itself was the bottleneck. “You had to hire more engineers, and that was very hard and expensive,” explained Moritz Plassnig, CEO of CloudBees. Automation and DevOps did not solve the problem; they only moved the jam to the next stage.
Foundational models and agentic coding have changed the equation. Plassnig said the writing side is no longer the limiting factor; engineers are moving from writing code to exercising judgment about what to build, what to release and what meets standards. The craftsmanship does not disappear, it moves up the value chain.
Jaymin West, front-deployment engineer and technology evangelist, described the phenomenon in a recent YouTube video: each of the eight built exactly the same machine. Not “similar”, the same shape. It is not coordination, it is an engineering convergence that marks a first-mature pattern for how software engineering looks in the agent era.
The operational model is simple: a support worker with a customer-feedback-driven idea sends it to the factory. The factory produces a first iteration; product, engineering and IT test it; then the factory takes it to mass production. The automobile-factory analogy is intentional: a prototype experiment followed by a repeatable, automated production line for wide supply.
Disclosure: Ziff Davis, the parent company of ZDNET, filed in April 2025 a lawsuit against OpenAI alleging copyright infringement in training and operating its systems.