Agile has become one of those words that means almost nothing because it has been used to describe almost everything. Daily stand-ups, sprint ceremonies, ticket rituals, planning poker, Scrum Masters, release trains, velocity charts and entire organisational hierarchies have all been presented as Agile. In many organisations, adopting Agile has meant replacing one rigid management system with another, only with more meetings and a new vocabulary.
That is a long way from the original idea.
The principles behind the Agile Manifesto are remarkably direct. Deliver valuable software early and continuously. Welcome changing requirements. Put businesspeople and developers together. Trust motivated people. Measure progress through working software. Maintain technical excellence, pursue simplicity and regularly improve how the team works.
Read those principles now and they do not feel outdated. They feel unusually well suited to software development in the age of AI.
The central idea of Agile was not Scrum. It was reducing the distance between an organisation having an idea and discovering whether that idea works.
Traditional software projects tried to manage uncertainty through prediction. Requirements were gathered, plans were produced, teams were assembled and delivery was divided into phases. By the time working software reached a real user, months or sometimes years had passed. Inevitably, the organisation had learned something new in the meantime, the market had moved, or the original requirements had been wrong.
Agile accepted that software development is an act of discovery. You cannot fully understand the product before you begin building it because building it changes your understanding of the problem. The answer was shorter feedback loops, closer collaboration and frequent delivery of working software.
That reasoning remains completely sound. AI makes it more important, not less.
When the cost and time required to produce software fall, the limiting factor shifts. Writing code is no longer necessarily the slowest part of the system. Understanding the problem, making good decisions, validating assumptions and judging the quality of the result become more significant. The organisation that learns fastest gains the advantage.
This is the thinking behind our Software Factory model.
A Software Factory is not a larger development team with an AI coding tool added to it. It is an operating model designed to turn priorities into working software continuously. A small, senior team works directly with the organisation, supported by an AI harness that can generate, test, inspect and improve software at a scale that would previously have required many more people.
The result is much closer to the original Agile ideal than many of the frameworks created in its name.
The customer is not separated from delivery by layers of account management, project management, analysis and handover. Business and technical people work together throughout. Priorities can change because the factory is a persistent capability, rather than a project plan that must be protected. Progress is demonstrated through software that runs, not through reports showing how much of the plan has been completed.
Technical excellence also becomes central. AI can produce code quickly, but speed without architectural judgement simply creates problems faster. The Software Factory places experienced architects and engineers around the harness so that design, security, maintainability and operational quality are continuously assessed. Good architecture is not a stage completed before development begins. It evolves alongside the product.
The principle of simplicity may be the most relevant of all. The Agile Manifesto describes it as maximising the amount of work not done. AI allows us to generate enormous quantities of software, but that does not mean we should. A mature Software Factory uses AI to remove unnecessary effort, reduce ceremony, automate repetitive work and focus human attention on decisions that matter.
Software Factory is not a rejection of Agile. It is what Agile looks like when many of the old constraints around software production begin to disappear.
We no longer need to organise work around the assumption that every feature requires a large team and months of coordinated human effort. We do not need process for its own sake, nor do we need increasingly elaborate frameworks designed to make large, slow organisations appear adaptive.
We need small groups of trusted, capable people. We need direct access to customers and users. We need working software delivered frequently. We need the freedom to respond when we learn something new. We need technical excellence, sustainable delivery and constant reflection on how the system can improve.
That was Agile in 2001. It is still Agile now.
Software Factory simply gives us the opportunity to practise it properly.