The AI Software Factory: Why Building Software Will Stop Being the Bottleneck

Software development is becoming a much smaller part of building software. I expect that over the next few years, a product owner may still start with a Jira ticket, but instead of handing it to a conventional delivery team, a set of AI agents can analyse the requirement, produce a design, write the code, review it, test it and prepare it for deployment. Yes, people will remain responsible for deciding what should be built, resolving ambiguity, setting constraints and judging whether the result is right. But far fewer human hours may be required to turn that decision into working software. In manufacturing or retail, there is a word for such a process, where everything runs with high automation. It’s a dark factory. That is what is coming for software, and I put it at two or three years because we have already run several of them. For a solar energy storage company in Texas, we built a lab management portal with six agents covering business analysis, design, development, review and security, QA, and project management. Our engineers stepped in when the agents needed clarification or intervention. The application went live in seven days, compared with roughly three months for a conventional delivery model.

This is a short preview supplied by the publisher. The charts above are separate ApplyDjinn observations. The complete story, including images and updates, remains with the publisher.