Applying the lens in a large company
Most transformation programs fail not from lack of budget, but from treating a living system like a machine to be re-engineered.
Where large organizations rust
Roughly 70 percent of digital transformations fall short of their objectives — a global finding, not a German specialty, and the German-language data offers no comfort: in a survey of about a thousand organizations, most of them German and Austrian, a clear majority had not reached their agile-transformation goals. A failure rate that high and that stable suggests the cause is structural rather than a run of bad luck or bad managers. It was not the money: German firms entered the decade with historically high equity and liquidity reserves. What kept failing was the model of change itself — transformation treated as a machine upgrade, announced from the top, executed by decree. The three traps below are that model meeting an organization that is not a machine.
The board announces "we are agile now"; middle managers run the ceremonies and keep the control. Stand-ups happen, decisions don't move. The org chart changed costume while the incentive structure — who gets promoted, who gets blamed — stayed untouched, so behavior stayed untouched too.
SOURCES · BearingPoint 2023 · Gonçalves 2016Each department optimizes its own KPI, and the sum is a company nobody chose. Cross-functional value streams die in the gaps between org-chart boxes, because every agent is playing the game it was given.
SOURCES · Gonçalves 2016Decades of patched core systems make every change risky, so the IT department becomes a gatekeeper whose rational answer is "not until we replace system X" — a multi-year project that keeps not finishing. The people running the old systems become invested in their continuation.
SOURCES · Couchbase 2019What the lens says
To see why the decree fails, slow the film down to a single decision. The board has announced the agile transformation; the ceremonies are running. A mid-level engineer now chooses whether to say, in the new stand-up, that the flagship project is in trouble. The methodology asks for transparency. The promotion record says the last person who escalated bad news early got moved sideways, and the manager running the meeting still writes the performance reviews. The engineer can read incentives as well as anyone, and reports green. Repeat that one small calculation across forty thousand employees, daily, and you have the corporation's actual behavior — orderly, self-reinforcing, and unmoved by the announcement.
This is what it means to call a corporation a complex adaptive system: its behavior is produced by thousands of people separately adapting to the incentives in front of them, not by the intentions at the top. "Culture", in this reading, is the stable pattern of what gets rewarded and what gets punished, as experienced daily — which is why culture change cannot be announced, only re-priced. Now revisit the engineer under a re-priced regime: the flagship project demos working results every thirty days, so its true state will surface within a month whether anyone reports it or not, and the last team that brought an honest negative was thanked in front of the division. The same self-interested calculation that produced the green report now favors saying so early. Nobody's character changed; the prices did. Change the vocabulary and leave the prices alone, and you get stand-ups run by the old incentives. The transformation problem is loop-work: find the feedback loops that are slow, severed, or pointing the wrong way, and repair them. Where does a hierarchy's loop-work actually start?
2016 Incentives, networks, emergence → 2026 Play the coordination games →
What to do instead
Replace the annual program review with a monthly demo of working results — not slides, the thing itself. Your slowest loop is currently a year long; make it thirty days and watch what the organization suddenly notices.
M4 Make failure cheap.Fence off one unit, one quarter, one budget line — and decide before it starts which result ends it. Then praise the team that reports the honest negative: whatever happens to the first failed pilot is what everyone learns to do with failure.
M5 Skin in the game.Let the team that builds it run it — and feel it when it breaks. Handovers to 'operations' sever the loop that makes builders careful.
M7 Decentralize decisions.Give the people closest to the customer real decision rights with a budget line, not a suggestion box — then count how many decisions still queue for the board each month, and treat that number as your speedometer.
Retire these
Announcing culture change in a town hall— culture is what is rewarded, not what is said.Five-year detailed transformation roadmaps— plan the direction, not the path; the path will not survive contact.Buying an innovation lab instead of changing incentives— the lab becomes a zoo the organization visits.Waiting for the big system replacement before improving anything— strangle the legacy system with working seams, piece by piece.