The Island of Agility: Why a Fast Team Cannot Transform a Slow Organisation
The improvement is real and the transformation is illusory, and the gap between them is exactly the width of the boundary.
Executive Summary
Somewhere in most large organisations by now there is a team that has genuinely become agile. It plans in short cycles, it ships working software every fortnight, it inspects and adapts, and the people on it will tell you — accurately — that they have never delivered so well. And somewhere just past the edge of that team, the organisation that surrounds it carries on exactly as before: funding once a year, deciding by committee, gating by quarter, reporting by milestone. The team is fast. The system it sits inside is not. Between them lies a boundary at which most of the promised transformation quietly disappears.
This essay is about that boundary, and about why it has proved so durable. The dominant story we tell ourselves is that agility spreads outward from the team — that a successful pilot is a seed, and that with enough seeds the enterprise itself becomes agile. Three or four years into the enterprise adoption of these methods, that story is looking thin. The teams took. The organisations did not. The pattern is too consistent, across too many sectors, to be explained by weak sponsorship or the wrong coach. It is structural.
My argument is that the un-agile organisation is not a backdrop against which agile teams struggle; it is the actual subject. The team’s agility is real but local, and locality is the whole problem: a fast cell inside a slow body inherits the body’s tempo at every point where it must touch it — money, decisions, releases, people, reporting. I will try to show the mechanism at each of those points, and to take seriously the strongest case for the organisation’s caution, because that case is better than agile’s advocates usually allow. But I will also argue that most organisations have done something stranger than resist agility. They have adopted its ceremonies while declining its substance — buying the stand-up and the sprint and the retrospective, all of which can be installed without moving any real authority, while keeping untouched the distribution of decision rights that would have made those ceremonies matter. The result is the worst of both worlds, and it tells us something uncomfortable about the distance between what we say we want from transformation and what we are actually willing to change to get it.
The Review and the Gate
Consider two rooms, a fortnight apart, in the same organisation.
In the first room it is the end of a sprint. A team of nine is demonstrating what it built over the last two weeks. The work is real: a running increment, tested, integrated, shown live rather than on slides. The product owner accepts most of it and defers one item with a shrug that costs nothing, because there will be another increment in two weeks and the decision can wait. The retrospective that follows is candid; two small improvements to how the team works are agreed and will be in place by Monday. Anyone watching would conclude that this is what good looks like. It is.
In the second room, ten days later, it is the quarterly programme board. The same team’s work is now one line in a portfolio pack that runs to fifty pages. The increment they demonstrated cannot actually be released to customers, because releases are bundled into a quarterly window that is eleven weeks away, and the change-approval process that governs that window requires documentation the team has not been asked to produce and a sign-off from a group that has not seen the software. A dependency on another team — one that is not agile, and is working to a plan set last autumn — has slipped, and there is no mechanism to reconcile the two tempos except to wait. The board notes the status. It does not decide anything, because deciding is what next quarter’s board is for. The meeting closes on time.
Both rooms are working exactly as designed. That is the point. The first room is agile and the second is not, and the software the first room built will sit, finished, for the better part of a quarter, waiting for the second room’s machinery to turn. The team has compressed the part of the process it controls from months to a fortnight, and in doing so has revealed, with painful clarity, how much of the total lead time was never inside the team at all.
The Island
The image that practitioners reach for, when they describe this, is an island. The agile team is an island of one operating model inside a sea of another. On the island, work flows; at the shoreline, it queues. The team has changed how it works but not what it is embedded in, and everything that crosses the boundary — a funding request, a release, a decision, a person’s time, a status report — has to be translated out of the island’s tempo and into the mainland’s, where it slows to the mainland’s pace.
An agile team inside an un-agile organisation does not make the organisation faster. It makes the organisation’s slowness visible, by removing its own contribution to the delay and leaving only the part it does not control.
What is striking is how stable these islands are. We might expect that a team demonstrably delivering better would exert a pull on its surroundings — that the organisation, seeing the improvement, would reshape itself to remove the friction at the boundary. It rarely happens. The island persists, year after year, neither spreading nor being reabsorbed, and the friction at its shoreline is treated as a fact of geography rather than a problem to be solved. To understand why, we have to look at what the boundary is actually made of, because it is not made of resistance to agile ideas. It is made of the ordinary machinery of the organisation, each piece of which is there for a reason and none of which moves because a team downstream has started working in fortnights.
Why the Island Does Not Spread
The boundary is not a wall. It is a set of couplings, each running on the organisation’s native tempo, and the agile team is coupled to every one of them. Name them individually and the durability of the island stops being mysterious.
- The money arrives once a year. The team may plan in fortnights, but it is funded in an annual round that allocates a fixed envelope against a business case written before the work began. When the team learns something in sprint three that ought to change what it does in sprint eight — exactly the learning the method exists to produce — there is no mechanism to move the money, because the money was committed whole, last autumn, to the plan as it was then understood. Agility generates the ability to change course, and the budget removes the ability to fund the change. The two cancel.
- The decisions are made by committee, on the committee’s calendar. Inside the team, decisions are cheap and continuous. At the boundary, they revert to being expensive and periodic, because the authority to make them was never delegated to the team — only the work was. A choice the team could settle in an afternoon waits weeks for a board that will, in the end, mostly note it. The team has been given responsibility without the decision rights that would let it discharge that responsibility, and the gap between the two is filled by waiting.
- The release is gated by a process built for risk, not for flow. Potentially shippable is not the same as shipped. Between the team’s definition of done and an actual customer sits a release process — change advisory, environment booking, a deployment window, a separate operations group — that was designed in and for a world of infrequent, high-ceremony releases, and that treats a request to deploy every fortnight as an anomaly to be managed rather than a capability to be enabled. The divide between those who build and those who run is not a technical fact; it is an organisational one, and it sits squarely on the boundary.
- The people belong to functions, not to the team. In the matrix, the tester, the analyst, the specialist are lent to the team by line managers who own their appraisals, their careers, and ultimately their priorities. When the function needs them back, or spreads them across three teams at once, the team’s stable, cross-functional wholeness — the thing that made it work — dissolves, and no amount of ceremony inside the team can hold together people the organisation’s structure is pulling apart.
- The reporting demands a language the team does not speak. The organisation governs by milestone, percentage-complete, and RAG status against a plan. The team produces working software and an empirically observed rate of progress. To be legible to the mainland, the team’s reality must be translated into the mainland’s artefacts — a translation that is lossy, time-consuming, and quietly corrupting, because once a team is asked to report a velocity as a target rather than a forecast, the number starts to be managed, and the honest signal the metric was meant to carry decays into a figure produced to keep the board content.
Each of these couplings is individually reasonable. The annual budget keeps faith with the capital plan; the committee exists to coordinate; the release gate protects a shared production environment; the functional line develops deep expertise; the reporting gives the board its oversight. None was built to obstruct agility, and none yields to it, because each answers to a logic that a fast team downstream does nothing to change. This is why pilots do not spread. The island is not failing to convert the mainland; it was never coupled to the mainland in a way that could.
In Fairness to the Organisation
It is easy, from the island, to treat the mainland as simply benighted — a mass of bureaucracy that would dissolve if only the people on it were braver or better informed. That view is comfortable and it is wrong, and the argument is not worth making unless it can survive the strongest version of the organisation’s defence.
Begin with the fact that a team’s local success does not, by itself, prove anything about the enterprise. A single team, well-staffed, working on a well-bounded problem it happens to control end to end, is close to the ideal conditions for these methods. Most of what a large organisation does is not like that. It is many teams that must move together, on shared platforms, against dependencies that no single team owns, toward outcomes that require dozens of contributions to arrive in the right order. Coordination at that scale is a real problem, and the annual plan, the stage gate, and the programme board are real if imperfect answers to it. Remove them in the name of agility and you do not automatically get a faster enterprise; you may get a set of locally optimal teams pulling in directions that no longer add up.
Then there is accountability. We are governing in the years after the corporate scandals that reshaped what boards understand themselves to be for, and in the aftermath of a financial crisis that has left every institution more anxious, not less, about being able to demonstrate control. A board that funds change is answerable for that money to people who will not accept “the teams are self-organising” as an account of how it was spent. The demand for predictability, for an auditable trail of decisions, for a plan against which performance can be judged, is not a failure of nerve. It is the price of using other people’s capital, and it does not lift merely because a delivery method has improved.
“The organisation’s caution is not the opposite of good judgement. It is a form of it — applied to risks the island cannot see from where it stands.”
And there is the honest observation that not everything should move at the team’s tempo. Some decisions ought to be slow. Committing capital, changing the direction of a large programme, altering something on which many other things depend — these are choices where deliberation is a feature, and where the team’s instinct to decide quickly and adapt later would be a liability rather than a virtue. An organisation that fixed nothing, that held no commitment long enough to deliver it, would not be agile; it would be incapable of finishing anything.
All of this is true, and none of it is a defence of what most organisations have actually done. Because the honest case for the mainland is a case for coordination, accountability, and deliberate commitment — and one can accept every word of it and still observe that the annual freeze, the deciding-by-drift committee, the release gate that adds risk by batching, and the reporting that corrupts its own metric are not those virtues. They are the crude, inherited forms those virtues have congealed into, defended as though they were the virtues themselves. The strongest argument for the organisation is an argument for keeping the substance of control while changing its form. What most organisations have done is the reverse.
The Boundary, Measured
Let me make the cost concrete, because the argument turns on a number that is almost never looked at.
Take a composite team — recognisable, I think, to anyone who has worked near one of these programmes. Inside the team, the cycle is genuinely two weeks: an idea accepted into a sprint is built, tested, and integrated into a shippable increment within the fortnight. By any measure the team controls, it is fast, and its reported velocity is stable. Everyone involved, asked how delivery is going, says it is going well, and they are not wrong about the part they can see.
Now measure the thing no one on the island measures: the time from the moment a feature is accepted into a sprint to the moment a customer can actually use it. The two-week build sits, finished, waiting for a quarterly release window — on average six weeks away. The release itself requires a change-approval cycle and a hardening phase bolted on outside the team, adding three or four weeks. A dependency on a plan-driven team, working to a schedule set the previous autumn, adds more. Count it honestly and the lead time to real value is not two weeks. It is four or five months. The team compressed the small fraction of the process it owned and left the rest untouched, because the rest was never inside the team to compress.
| Measure | Inside the team | Across the boundary |
|---|---|---|
| Unit of planning | The two-week sprint | The annual budget round |
| Cadence of decisions | Continuous, in the team | Quarterly, in committee |
| Definition of done | Integrated, tested increment | Deployed, in front of a customer |
| Cycle time | About two weeks | About four to five months |
| What the metric reports | A forecast, freely given | A target, carefully managed |
The number that matters — the four to five months — is invisible in every report the organisation produces, because every report stops at the boundary of what the team controls. The team’s velocity looks excellent; the programme’s actual responsiveness is barely changed from the plan-driven world it thinks it has left. And this is the quiet tragedy of the island: it can be a complete, sincere success on its own terms and make almost no difference to the outcome the organisation was trying to buy when it funded the change. The improvement is real and the transformation is illusory, and the gap between them is exactly the width of the boundary.
What the Method Concealed
Why, then, do organisations adopt agility so enthusiastically at the level of the team and so rarely follow it across the boundary? The usual answers — inertia, fear, insufficient training — are true enough but shallow. The deeper answer is that the parts of agility an organisation adopts and the parts it declines are not chosen at random. There is a pattern to what gets taken up, and the pattern is revealing.
What gets adopted is the ceremony. The daily stand-up, the sprint, the demonstration, the retrospective, the board covered in cards, the named role of the ScrumMaster — all of these can be installed inside a team without asking anyone outside it to give up a thing. They are visible, teachable, certifiable, and above all contained. A manager can send a team on a course, watch the ceremonies appear, and report that the organisation is becoming agile, and none of it disturbs the annual budget, the committee’s authority, the release gate, or the reporting line. The ceremony is adoptable precisely because it is powerless at the boundary.
What gets declined is the redistribution of authority that the ceremony was supposed to serve. The stand-up is meant to be the visible surface of a team that decides its own work; the sprint is meant to be a real boundary inside which the team is trusted and outside which the organisation does not reach; the retrospective is meant to produce changes that the team has the standing to make. Strip out the delegated authority and what remains is the form without the function — a stand-up that reports upward instead of coordinating sideways, a sprint the organisation reaches into whenever it likes, a retrospective whose improvements stop at the team’s edge. The organisation has bought the appearance of a new operating model and kept the old one intact underneath, and it has done so not by accident but because the appearance was affordable and the reality was not.
Agility was adopted as a method, which asks a team to change how it works, rather than as a redistribution of authority, which would ask an organisation to change how it decides. The first is a training exercise. The second is a transfer of power, and power is not transferred by attending a course.
This is what the method concealed, from its adopters as much as from its observers. By packaging a change in the distribution of decision rights as a change in team practice, agile made itself adoptable — and in becoming adoptable, it made itself possible to adopt in name only. The ceremonies became a way of not transforming while appearing to: a demonstration of intent that cost nothing, precisely because it changed nothing about who decides.
The Boundary Always Wins
If there is a single lesson in three or four years of watching agile teams live inside un-agile organisations, it is that the boundary always reasserts the organisation’s true operating model. You cannot install responsiveness as a local optimisation and expect the surrounding system to yield to it. The system was there first, it is larger, and every coupling that crosses the boundary runs on its terms. A fast team does not, by existing, make a slow organisation fast. It makes a slow organisation legible — it exposes, by subtraction, exactly where the time and the decisions actually go — and then it waits at the shoreline like everyone else.
This is not an argument that the organisation should surrender its judgement to the team, nor that coordination and accountability and deliberate commitment are dispensable. It is an argument that these things can be honoured in forms that move at something nearer the pace of the work, and that the annual freeze and the quarterly gate and the milestone report are not those forms but their fossils — kept in place, and defended with the language of control, long after they stopped delivering control and started merely delaying delivery. The teams have shown that a faster tempo is possible. The question the boundary poses is whether the organisation is willing to change how it funds, decides, and governs in order to run at that tempo — and that is a question about the redistribution of authority, not about method.
Which returns us to the gap between transformation intent and transformation reality, because the island is that gap made visible. We say we want to be responsive, adaptive, fast to learn and fast to change. We are willing to send our teams on courses, to hang the boards and hold the ceremonies, to adopt every part of agility that can be adopted without anyone ceding anything. What we have been unwilling to do is move the authority — to fund in tranches rather than in years, to let the standing governance decide rather than note, to release when the work is ready rather than when the calendar permits, to trust the team’s account of its own progress rather than translate it into a number we can manage. The agile team inside the un-agile organisation is not, in the end, a story about method at all. It is a mirror, and what it reflects is the precise distance between the change we say we want and the power we are prepared to give up to have it.