The critics of Agile are often right.
Not about everything. But about enough that dismissing them is a mistake. The thing most people mean when they say “Agile” — the two-week sprints, the daily standups, the SAFe diagrams that span entire walls, the certification programs, the consultants who turn a mindset into a compliance exercise — much of that deserves the skepticism it gets. The “Agile Bug,” as practitioners have taken to calling the corruption of Agile into its opposite, is real — and the Agile Bug deserves its own examination. It caused real harm. Teams burned out running ceremonies that produced motion without progress. Organizations paid for Agile transformations that delivered heavier bureaucracy with a sanitized vocabulary that missed deadlines and business opportunities.
We’re not here to defend any of that. In fact, we join you in strong solidarity.
So when we say “Agile,” what do we actually mean?
Not frameworks. Not SAFe, not LeSS, not anything in between. Not micromanagement dressed in sprint clothing. Not theater.
We mean what was there before the commercialization of what was a brilliant realization: the values and principles that the Manifesto authors captured in 2001, which themselves were distilling approaches that predated the Manifesto by decades — from Toyota’s production philosophy to NASA’s incremental delivery to early iterative development work. The core of it, stripped of everything layered on top: Individuals and interactions over processes and tools. Working outcomes over exhaustive documentation. Collaboration over contract negotiation. Responding to change over following a plan.
Not as slogans. As actual operating principles — the kind that change how you behave and, more importantly, how you think when a situation is ambiguous, complex, and fast-moving.
That’s what we mean by Proper Agile.
The thesis
Proper Agile is the mindset that belongs to complex work. AI is among the most complex work we’ve ever done. The fit isn’t accidental — it’s inevitable.
Here is a framing that nobody else in the Agile community promotes, but that I hold as central to everything this site is built on: Agile is a cultural antidote to the systemic decline of human cognitive performance. Not a productivity framework. Not a delivery methodology. An antidote — to something that is actively happening in organizations, at scale, right now.
That definition elevates the human. Proper Agile requires and cultivates higher cognitive performance: the capacity to engage, question, adapt, and reason under uncertainty rather than execute prescribed steps. This is precisely what AI requires from the humans working alongside it. AI produces better outputs when paired with cognitively engaged humans who interrogate the work, catch the drift, redirect when needed.
And when AI is used well, it amplifies human thinking rather than replacing it — it becomes a force multiplier for exactly the cognitive capacity that Proper Agile develops. The best case for Agile in an AI era isn’t that it helps you ship faster. It’s that it creates humans capable of being genuine AI partners.
The reasoning traces back to first principles. Agile thinking starts with a clear-eyed observation: in complex work, you cannot plan the full path in advance. You work empirically — start with what you know, learn as you go, inspect and adapt continuously. You prioritize real outcomes over exhaustive upfront specifications. You keep things as simple as possible. You collaborate, because no single perspective has the full picture.
Now describe working with AI. The outputs are non-deterministic. Quality depends on how well you engage, not just what you ask. Early feedback catches problems that late correction cannot fix. The relationship between human judgment and model output is iterative by nature. No single interaction is the whole picture.
The same first principles apply — not because someone designed it that way, but because both Agile and effective AI collaboration are responses to the same underlying reality: complex work, done well, requires a particular mindset that performs at a higher cognitive level than the norm. That mindset is Agile.
The strongest objection — we’ll name it for you
Agile was built for human teams. It assumes fatigue, forgetting, imprecise estimation, social dynamics — properties specific to people working together. An AI doesn’t tire, doesn’t forget within its context, and carries none of the social fears that human teams needed a whole philosophy to address. So: the premises don’t apply, and “Agile for AI” is a category error.
It’s a real argument. We engage it seriously rather than dismiss it.
And we think it misidentifies what the enduring core of Agile actually is. Some of what Agile developed as workarounds for specifically human limitations — the retrospective ritual designed to surface what a tired team won’t say directly, the standup built to briefly synchronize distracted humans — those don’t transfer unchanged. But the underlying principles weren’t workarounds. They were observations about the nature of complex work itself. And complexity doesn’t change because your collaborator is a model. In fact, because your collaborator is an AI you now have to deal with new levels and types of complexity that nobody has ever seen before — squarely why Agile emerged in the first place. The fit is unavoidable.
What this looks like with AI in the room
Before you engage — a few concrete examples, because abstract principles need grounding.
Are we proposing to retrospect with AI? Yes — and it matters more, not less. Working with a model, you build patterns of interaction. Those patterns can sharpen over time, or degrade to failure. A regular pause to examine what’s working, what’s drifting, and what assumptions need resetting keeps you aligned in effective collaboration. The form a retrospective takes changes. The discipline of stepping back and inspecting doesn’t.
Are we proposing to run check-ins with AI? Yes — and not just once a day. Brief alignment checks — at natural inflection points throughout the work, not just at the start of a session — keep the collaboration grounded. A model can accumulate context in ways that drift from your original intent without signaling that it’s happening. Frequent, lightweight synchronization is how you catch that before it costs you.
The same mindset. Adapted form. The underlying Agile principle travels intact.
Now: what’s your case?
That’s our position. Stated clearly enough that you can pointedly push back on it.
If you think the premise is wrong — that Proper Agile doesn’t translate, that the category error objection is decisive, that we’re forcing a mindset where none fits — make your strongest case. Not the strawman version. The real one.
The rule here is simple: discuss the idea, respect the person. We’ll engage every genuine challenge in good faith.
Bring it.
Comments 0