I keep hearing it lately: “Agile is dead because of AI.” At conferences, in customer meetings, on LinkedIn. The reasoning is usually the same: If an AI agent builds in two hours what a team used to build in two weeks, why do we still need sprints, standups, and retrospectives?

I think something is being confused here. The idea and its packaging.

The twelve principles, read again

The Agile Manifesto is 25 years old. It consists of four values and twelve principles. I recently read the principles again, this time with one question in mind: Which of them does AI make obsolete?

My answer: none. On the contrary, several of them only become truly important now.

Working software is the primary measure of progress. With AI agents, output has become cheap. An agent produces more code in an hour than a developer does in a day. That is exactly why the amount of code no longer measures anything. The only question that counts: Does it run, and does it deliver value? The principle was always right. Today it is the only defense against the mountain of generated code.

Simplicity, the art of maximizing the amount of work not done. AI likes to produce too much. Too many abstractions, too many options, too many features nobody asked for. Leaving things out becomes a core skill. If you do not tell the AI what not to build, you get everything. This principle fits my motto: Keep IT simple.

Continuous attention to technical excellence and good design. The question of whether generated code is still maintainable in two years does not get smaller. It gets bigger. When nobody writes every line themselves anymore, somebody has to look even more closely at how everything fits together.

Business people and developers must work together daily. For me, this is the most important point. Typing used to be the bottleneck. Today it is the ambiguity in the requirement. An AI agent implements exactly what the specification says. If something is missing there, it is missing in the software. The collaboration between business and development does not become less necessary. It becomes the actual job.

At regular intervals, the team reflects and adjusts. The tools change every month. What was the best approach last spring is outdated today. A team that does not regularly pause and adjust its way of working will be overrun by the pace of change.

I could go on. Early and continuous delivery? Gets easier. Welcoming changing requirements, even late? Gets cheaper. Sustainable pace? Gets more important, because the temptation to overrun the team gets bigger.

What is actually under pressure

It is not the manifesto that is dying. It is the industry that grew around it.

Certificates you can earn in two days. Scaling frameworks with dozens of roles and meetings. Story points, velocity charts, and burndown diagrams as reporting tools for management. Sprint rituals performed because the framework says so, not because they help the team.

Much of this was never “agile” in the sense of the manifesto. It was overhead meant to manage the slowness of manual development. A two-week sprint makes sense when two weeks is the smallest useful unit in which a team can deliver something that works. When an iteration no longer takes two weeks but two hours, this whole administrative apparatus collapses.

This is what the “agile is dead” voices are observing. And they are right. They just draw the wrong conclusion.

Back to the idea

When I read the four values of the manifesto, they sound more current in 2026 than in 2001:

  • Individuals and interactions over processes and tools. Yes, including AI tools.
  • Working software over comprehensive documentation. With one nuance I want to add, more on that in a moment.
  • Customer collaboration over contract negotiation.
  • Responding to change over following a plan.

The nuance about documentation: The manifesto does not say “no documentation”. It says working software is more important. With AI agents, the weight shifts. The specification is no longer a document written after the code that then goes stale. It is what the human contributes. The agent delivers the implementation. The specification is the input, not the output.

This is exactly where I start with the AI Unified Process. Requirements and use cases as the human contribution, developed together with the business, as principle 4 demands. The architecture emerges from the team, as principle 11 says. This is not a departure from the Agile Manifesto. It is the logical consequence when implementation becomes almost free.

Conclusion

Agile was never a framework. It was an attitude: Deliver early, work closely with the customer, keep it simple, stay technically clean, and adapt continuously.

AI does not make this attitude obsolete. AI makes it a prerequisite. What disappears is the apparatus the consulting industry built around it. And honestly: I will not shed a tear for it.