Why the methodology built for human execution cannot survive the shift to machine execution
I have a lot of time for the Agile Manifesto.
Not for what Agile became: the certification industry, the Jira theatre, the daily standups that run forty-five minutes and produce nothing. Not for SAFe, which took a philosophy built on small teams and fast feedback and turned it into the very bureaucracy it was designed to replace. Not for story points, which were meant to be a rough sizing tool and became a performance management weapon.
But the Manifesto itself? Four values and twelve principles, written by people who understood that the way we built software was broken? That was genuinely important work. And for two decades, it was the right answer.
The problem is that it was the right answer to a question that no longer applies.
Before going further, I want to draw a distinction that the rest of this article depends on. There is Agile with a capital A – the methodology, the frameworks, the certifications, the ceremonies, the entire industrial apparatus that grew up around the original Manifesto. And there is agile with a lowercase a – the underlying spirit of being responsive, iterative, and adaptive in the face of uncertainty.
In the Human Intent era, capital-A Agile is dying. Lowercase-a agile is more important than ever.
The Question Agile Was Built to Answer
Agile emerged in 2001 because software development had a specific, structural problem: building software was slow, expensive, and unpredictable. Requirements gathered upfront were wrong by the time the code was written. Projects took months or years. The gap between what was specified and what was needed grew with every passing week.
Waterfall tried to solve this through planning. Plan everything. Document everything. Approve everything. Then build. The industry spent decades learning the hard way that this did not work, because the world changed faster than the plans could be executed.
Agile took a different approach. Instead of trying to plan better, it embraced the uncertainty. Short iterations. Working software over documentation. Customer collaboration over contract negotiation. Responding to change over following a plan. Deliver something small, get feedback, adjust, repeat.
The underlying insight was simple and correct: when execution is slow, you cannot afford to get the requirements wrong upfront. So iterate. Learn through building. Shorten the feedback loop between idea and outcome.
That insight, the lowercase-a part, is timeless. It is how complex systems are built under uncertainty. It will outlast every methodology that ever claimed it.
But every Agile practice flows from a different problem: how do you coordinate human teams doing slow, expensive, unpredictable work?
- Sprints exist because you need time-boxed cycles when human teams build incrementally.
- Standups exist because human coordination requires frequent synchronisation.
- Retrospectives exist because human teams need structured reflection to improve.
- Story points exist because human effort is variable and hard to predict.
- Backlogs exist because there is always more work than humans can execute.
All of it, every ceremony, every artefact, every role, is a coordination mechanism for human teams doing human work at human speed.
That is the part that does not survive.
What Happens When Execution Is No Longer Human
Agentic AI systems can now generate code, write tests, run validation, fix failures, and iterate – not in sprints, but in minutes. A capability that took a team two weeks to deliver through a sprint can be generated, tested, and refined in a fraction of the time.This is not theoretical. Anthropic’s 2026 Agentic Coding Trends Report shows agentic coding tools reshaping the software development lifecycle, with cycle times collapsing from weeks to hours. Thoughtworks hosted a twenty-fifth anniversary workshop for the Agile Manifesto earlier this year in Deer Valley, Utah, explicitly asking how AI-native development changes the picture. Their finding was telling: when AI writes the code, the rigour that used to live in coding does not disappear – it moves elsewhere. Martin Fowler, one of the original Manifesto signatories who hosted the workshop, called it “way too early” for a new manifesto because “people are still experimenting with ideas.”
But none of them named where the rigour moves to.
It moves to intent.
There is a deeper change behind this, and it is worth dwelling on. For most of computing history, the gap between human intent and machine action was a translation problem. Humans had to learn the machine’s language; assembly, then C, then higher-level languages, then frameworks. Every layer of abstraction reduced the gap, but it never closed it. Specifying intent precisely enough for a machine to act on always required passing through code, or design documents, or human interpreters who could translate between domains.
Natural language has collapsed that gap. We can now type or even speak our requirements in our native language and have systems act on them directly. The translation barrier, the thing that made every previous software methodology partly a coping mechanism for that translation, is largely gone.
This changes the meaning of iteration itself. Agile’s short cycles were partly a response to genuine uncertainty about what users needed, but they were also a response to the fact that you could not specify what you wanted accurately enough upfront. Iteration was how you closed the gap between “what I meant” and “what the system does.” When natural language closes that gap directly, the iteration moves upstream, from refining the implementation to refining the intent itself.
When a machine can execute in minutes, the bottleneck is no longer execution speed. The bottleneck is the quality of the instruction. How clearly have you defined what you want? How precisely have you specified success? How well have you articulated the constraints, the boundaries, the trade-offs?
If the intent is clear, execution is fast and reliable. If the intent is vague, execution is fast and wrong. And because execution is fast, the consequences of poor intent arrive fast too.
This is the same shift that underpins Intent-Driven Development: the constraint moves from implementation to specification. From building to defining. From execution to intent.
The Spirit Survives. The Methodology Does Not.
Here is where I differ from both the “Agile is dead” camp and the “Agile is fine” camp.
The conversation in the industry right now is polarised. Some are declaring the Manifesto dead and proposing replacements — agentic manifestos, new frameworks, new acronyms. Others insist the Manifesto’s values are timeless and that only the ceremonial bloat needs trimming. AWS has suggested in its Prescriptive Guidance that sprint planning should become “Intent Design.” Kent Beck has been exploring what he calls “augmented coding.” Martin Fowler says it is too early for a new manifesto.
InfoQ has documented the full debate, capturing positions from Capgemini’s Steve Jones (who argues the Manifesto is dead) to Forrester’s 2025 finding that 95% of professionals still consider Agile relevant.
I think all of them are circling the right answer without quite landing on it.
The Agile Manifesto’s values are not wrong. Iteration still matters. Fast feedback still matters. Responding to change still matters. Collaboration still matters. These are not artefacts of a particular era, they are fundamental principles of how complex systems are built under uncertainty. They are agile, lowercase a. And they will continue, perhaps in stronger form than ever, into the next era.
But the practices, sprints, standups, retrospectives, velocity tracking, backlog grooming, story points, these are human coordination mechanisms. They exist because human teams need rhythm, synchronisation, and structured reflection. They exist because building is slow and effort is unpredictable. This is Agile, capital A. And it is what is ending.
When execution is delegated to machines, those coordination mechanisms lose their purpose. Not because the values behind them are wrong, but because the constraint they were designed to manage, slow, expensive, human execution, is disappearing.
You do not need a two-week sprint when execution takes two minutes. You do not need a daily standup when there is no human team to synchronise. You do not need story points when effort is no longer variable in the way human effort is. You do not need backlog grooming when the backlog can be executed as fast as it can be defined.
The capital-A apparatus collapses. The lowercase-a spirit lives on, free at last from the bureaucracy that was suffocating it.
Intent Shaping Replaces Sprint Planning
In an AI-enabled world, the iterative cycle does not disappear. It transforms. In fact, it becomes more agile, in the truest sense of the word, than capital-A Agile ever managed to be.
Instead of iterating on implementation across sprints, you iterate on intent. You define what you want. A system executes against that definition. You measure the result. You refine the intent. The system executes again. The loop is the same: define, execute, measure, adjust – but it operates in near real-time rather than in two-week cycles.
This is intent shaping.
In Intent-Driven Development, intent shaping is the process through which human goals are transformed into executable, measurable, and governable specifications. It is not a one-time activity. It is continuous. Delivery reveals missing constraints. Measurement reveals where intent and outcome diverge. Governance reviews reveal risks. All of this feeds back into shaping.
The parallel with Agile’s original vision is clear, and worth dwelling on. Agile said: learn through building. Intent shaping says: learn through measuring the gap between what you intended and what was produced. The principle is identical. The lowercase-a spirit is the same. The mechanism is different, because the speed of execution has changed by orders of magnitude.
Sprint planning asked: what can the team build in two weeks? Intent shaping asks: what do we actually want, and how will we know we have it?
That is a harder question. And it is the right one.
What Stays, What Goes, What Changes
The Agile Manifesto valued individuals and interactions over processes and tools. In an AI-enabled world, this becomes even more important – but the nature of the interaction changes. The interaction is no longer between developers coordinating implementation. It is between humans defining intent and systems executing against it. The value lives. The interaction transforms.
The Manifesto valued working software over comprehensive documentation. In an AI-enabled world, this inverts. When machines can generate working software in minutes from natural language specifications, the software itself is almost disposable. What matters is the specification, the intent, that produced it. Documentation of intent becomes the primary artefact, not the code. The old tension between documentation and working software dissolves entirely: when your specification is also your instruction to the machine, they are no longer separate things. The value lives. The artefact transforms.
The Manifesto valued responding to change over following a plan. In an AI-enabled world, the speed of response collapses. Change can be responded to in real time. The question is no longer whether you can respond quickly enough. The question is whether you understand the change clearly enough to reshape your intent. The value lives. The response transforms.
The Manifesto valued customer collaboration over contract negotiation. In an AI-enabled world, this remains essential, but the collaboration shifts from “what should we build next sprint” to “what does success actually look like.” The conversation moves upstream, from implementation to purpose. The value lives. The collaboration transforms.
In each case, the lowercase-a value survives. The capital-A practice transforms beyond recognition. And the direction of transformation is always the same: upstream, toward intent.
This Is Not Just About Software
One more point, and it is an important one.
The “is Agile dead?” debate is happening almost entirely within software engineering. But in the first article in this series, we explored how the shift from execution to intent applies across every function: operations, finance, HR, marketing, product. Every department is discovering that work previously done through human process and coordination can increasingly be delegated to systems.
Agile never claimed to apply beyond software development, but many organisations adopted Agile principles for broader work management. Kanban boards, sprints, standups – these spread far beyond engineering teams. If those capital-A practices are losing relevance even within software, where AI is most advanced, then the implications for every other function are obvious. The same shift is coming. Just slightly later.
And when it arrives, the same question will apply: if execution is no longer the constraint, then what is? The answer is always the same.
Clarity of intent.
Long Live agile
Agile was the right answer to the question of its time: how do we coordinate human effort under uncertainty? It served the industry well for two decades. And its core insight – iterate, learn, adapt – does not just survive the shift to AI. It becomes the operating principle of an entirely new Human Intent era.
But the question has changed. The constraint is no longer slow execution. The constraint is clear intent. And that requires a different discipline, not a faster version of capital-A Agile, but a fundamentally different approach to what humans do when machines handle the building.
So Agile, the methodology, is dying. The certifications, the ceremonies, the bureaucracy that grew up around the Manifesto, that era is ending. It will not be missed.
But agile, the spirit – iteration, fast feedback, responsiveness to change, learning through doing – that has never been more alive. It moves upstream, into the work of shaping intent. It moves into the human-system loop where intent is defined, executed, measured, and refined in real time. It becomes faster, tighter, and more responsive than capital-A Agile ever managed to be.
The Manifesto’s authors wrote that they valued responding to change over following a plan. In an AI-enabled world, that value is no longer aspirational. It is structural. We can finally live it, fully, because the execution layer can keep up with the speed of human thought.
Agile is dead. Long live agile.
Govern Intent. Delegate Execution.
This is the second article in the Human Intent series. In the next article, we will explore why even the most thoughtful models of organisational change, including Kotter’s Dual Operating Model, were not designed for a world where execution is delegated to machines, and what that means for how organisations must now be structured.
A Note About the Banner Image
The circle in the image is an ensō, a symbol from Zen Buddhism, traditionally painted in a single, unbroken brushstroke. It represents clarity, presence, and the disciplined expression of intent in a single moment. It is not about perfection, but about completeness of intention. In this context, it reflects the shift explored in this article: as execution becomes automated and continuous, the defining human contribution is no longer the work itself, but the quality of the intent behind it.
Frequently Asked Questions
Is Agile dead?
Capital-A Agile — the methodology, the certifications, the ceremonies, the bureaucratic apparatus that grew up around the original Manifesto — is dying. Lowercase-a agile — the underlying spirit of iteration, fast feedback, and responsiveness to change — has never been more alive. The distinction matters because the practices were designed to coordinate human teams doing slow, expensive work. When agentic AI systems can execute in minutes rather than weeks, those coordination mechanisms lose their purpose. But the values that drove Agile in the first place become more important than ever, because they describe how complex systems are built under uncertainty.
What is the difference between capital-A Agile and lowercase-a agile?
Capital-A Agile refers to the methodology — Scrum, SAFe, sprints, standups, story points, backlog grooming, the certification industry, and the entire framework that grew up around the Agile Manifesto. Lowercase-a agile refers to the spirit behind it — being responsive, iterative, and adaptive in the face of uncertainty. The distinction is critical because the methodology and the spirit have very different fates in the AI era. The methodology was built for a world where human teams coordinated slow human execution. The spirit is timeless and applies to any system being built under uncertainty.
What is intent shaping?
Intent shaping is the continuous process of refining human intent in near real-time as AI systems execute against it. Instead of iterating on implementation across two-week sprints, you iterate on intent itself: define what you want, let a system execute, measure the result, refine the intent, and repeat. The loop is the same as Agile’s original cycle of define, build, learn, adjust — but it operates in minutes rather than weeks. Intent shaping is the discipline that replaces sprint planning when execution is delegated to machines. It is core to Intent-Driven Development and to the broader Human Intent shift across organisations.
How does agentic AI change the Agile Manifesto?
The four values of the Agile Manifesto survive the shift to agentic AI, but their practical expression transforms. Individuals and interactions still matter, but the interaction is now between humans defining intent and systems executing against it. Working software still matters, but the specification — the intent — becomes the primary artefact. Responding to change still matters, but the response can be near real-time rather than confined to sprint cycles. Customer collaboration still matters, but the conversation moves upstream from “what should we build” to “what does success look like.” The values live. The practices transform. And the direction of transformation is always the same: upstream, toward intent.
What replaces sprints, standups, and story points in the AI era?
The coordination mechanisms designed for human teams lose their purpose when AI systems handle execution. Sprints exist because human teams need time-boxed cycles; when execution takes minutes, sprints are unnecessary. Standups exist because human coordination needs synchronisation; when there is no human team building the code, there is nothing to synchronise. Story points exist because human effort is variable; when machines execute, effort is not variable in the same way. What replaces them is intent shaping — a continuous loop of defining, executing, measuring, and refining intent in near real-time. The discipline shifts from coordinating execution to governing intent.
Why is natural language so important to this shift?
For most of computing history, the gap between human intent and machine action was a translation problem. Humans had to learn the machine’s language — assembly, then C, then higher-level languages, then frameworks. Every layer of abstraction reduced the gap, but it never closed it. Natural language has collapsed that gap. We can now type or speak our requirements in our native language and have systems act on them directly. This changes the meaning of iteration itself. Agile’s short cycles were partly a response to genuine uncertainty about what users needed, but they were also a response to the fact that you could not specify what you wanted accurately enough upfront. When natural language closes that gap directly, the iteration moves upstream — from refining the implementation to refining the intent itself.
Human Intent – Organisational Change in the AI Era
For most of the last century, organisations have been built to coordinate human execution. But as AI systems take on more of the work, across every function, not just technology, the constraint shifts. It is no longer how fast we can build. It is whether we actually know what we want.
Human Intent – The End of Taylor’s Century
For over a century, organisations have been built on a single premise: that they exist to optimise human work. From Taylor’s scientific management to Kotter’s Dual Operating System, every major theory has been a variation within that frame. The third article in the Human Intent series explores why agentic AI is the first technology to genuinely overturn that assumption – and what comes after.
0 Comments