Composite
Felix is twenty-six, eighteen months into his first job at a management consultancy, and staffed on a technology operating-model project at a European financial-services company. The board has approved a six-week diagnostic because funding is producing activity without proportional outcomes.
Felix conducts thirty-one interviews and finds ownership fragmented, strategy detached from system behaviour, and engineers unable to change the processes their software implements. His picture is accurate and disconnected from the machinery it describes, since the project contains no inspection of code, operating data, or customer outcomes.
On Friday afternoon the engagement partner reads the deck carefully. “Fragmented authority” becomes “unclear accountabilities.” The proposed process-owning unit moves to an appendix. She replaces it with “cross-functional governance model” and circles the phrase once.
The consultants cannot say that the executives who commissioned the work are part of the fragmented authority preventing funding from producing proportional outcomes. They cannot remove coordinating roles held by the project's primary stakeholders, and placing engineers inside operating units is unavailable because “the business” does not recognise engineering as part of itself. The partner has learned that advice a client accepts travels further than a correct diagnosis the client rejects.
The final deck therefore proposes better alignment, clearer accountabilities, and a governance framework connecting strategy to delivery. Its most carefully costed page offers a delivery phase. The board approves unanimously, and within a week the operating model has become a headcount negotiation. Three executives compete for the new alignment roles while a fourth adds review gates and a RACI matrix requiring three workshops to complete.
Five months later, the build programme recommends refreshing the diagnostic because the company has changed: it now has more committees.
The alignment forums proposed in the final deck are initially busy and useful. They expose dependencies people had resolved privately, give executives one place to request status, and produce decisions the old structure delayed. The difficulty appears after the first decisions: no process owner receives authority to maintain them, so each ruling becomes another precedent the forum must remember. Attendance expands because every boundary decision may affect someone absent from the previous one.
Felix moves to a logistics client in Hamburg. The industry is different, the ownership pattern is familiar, and this time he makes the adjustment before the partner asks. He has become fluent in the distance between a diagnosis and a deliverable.
The Alignment-Industrial Complex is the system that makes Felix's adjustment rational, and consultants or programme leaders can do capable work inside it. Scrum Masters and the executives who commission them can do capable work there too. Together they keep incompatible parts of the company moving through recurring meetings and the bridge roles that prepare them. Once the company depends on coordination to make fragmented rights liveable, it buys more coordination.
The Alignment-Industrial Complex, rather than the individual consultant, is the problem. The partner did not invent the client's constraints, and Felix cannot remove them from a six-week project. Their work becomes commercially useful because they acknowledge failure without requiring decision rights to move.
Alignment is easy to buy because it promises improvement without forcing the sponsor to identify which existing right must move. The first costs are workshops and roles; licence and build phases fit the current structure. A repaired boundary is harder because it gives one owner a decision that another executive, team, or committee currently expects to influence. The structure has found its incentive.
The consulting industry reflects demand for improvements that an existing client coalition can approve. A client willing to pay for diagnosis still needs a recommendation its coalition can accept. A proposal that removes recurring coordination may also remove the follow-on work used to implement it. Rational people responding to commercial and political constraints can repeatedly produce an answer whose main achievement is surviving the approval process. That is the incentive working as designed, not cynicism.
An organisation can respond to interdependence in two broad ways. It can increase its capacity to process information across the divisions, or it can reduce the need to process that information by redesigning the work. Jay Galbraith made that choice explicit in 1974.
Most alignment programmes take the first branch. They add forums, integrators, reporting, portfolio machinery and people skilled at carrying decisions between partial owners.
Adding coordination capacity is rational when the boundary cannot be moved, and cumulative when the same response becomes permanent. A complete outcome still crosses functions, each function retains part of the authority and state, and ordinary change therefore requires a coalition. The people who can assemble that coalition become unusually valuable. They are promoted because they are competent at work the organisation requires, not because they have a psychological preference for meetings. Roles, budgets, status and commercial demand form around the same requirement.
WORKED EXAMPLE: The timetable hides the limit
A railway can look complete when its timetable is built around the limits of the line. Signals, crossings and recovery time fit the ordinary service, while dispatchers resolve the remaining conflicts. A weak section of track receives a speed restriction, and the timetable absorbs it. Passengers experience a slower journey rather than a design failure. Skilled people keep the service punctual, so the structural limit appears as the schedule.
Put a much faster train on the same line and the hidden constraints become visible. It catches the service ahead, waits for a free path and reaches each restriction sooner. Its unused speed measures conflicts that the old timetable had absorbed.
Galbraith's organisational choice has a railway equivalent. The operator can improve traffic management and process more conflicts, or reduce the conflicts by separating the flows and rebuilding the route. The first answer is rational when the line cannot change. It reaches a physical ceiling when fast and slow services continue to share the same path.
The corporate timetable is the meeting calendar. Review buffers absorb late decisions, and a forum created after one incident becomes part of the permanent schedule. Alignment keeps fragmented authority moving, often with real skill. Faster execution reveals how many ordinary decisions still have to pass through the crossings.
Documented case
The dedicated line. New infrastructure did not remove the service's institutional seams.
The Netherlands opened HSL-Zuid to commercial passenger traffic in September 2009. The route was purpose-built for high-speed service and used ERTMS signalling and a 25 kV power supply. Infrastructure had been renewed rather than patched.
The international Fyra service using V250 trains began in December 2012. The trains left service on 17 January 2013 after only a few weeks of operation. The rolling stock had serious defects, so the case cannot show that divided authority caused the failure. It does show that new infrastructure did not create one complete owner of the service.
The Dutch parliamentary inquiry, De reiziger in de kou, reported in October 2015. It found that the parties involved repeatedly placed their own interests above the passenger's interest. The record covers the state, the infrastructure manager, operators, procurement decisions and the manufacturer. The case therefore supplies a narrower lesson: a new technical route can leave the institutional seams of the service intact.
Structural repair then threatens visible positions. Moving a right removes an executive's expected influence, a programme's coordinating purpose, a committee's agenda or a consultancy's next phase. The operating benefit is dispersed across future decisions; the loss is immediate and concentrated. An acceptable proposal therefore tends to add coordination without moving a right. Under a commission bounded that tightly, a capable consultant can become an expensive pencil: able to render the coalition's answer more clearly, but not to redesign the authority that makes the answer necessary.
The loop needs no villain because it rewards people for making authority nobody designed as a system survivable.
A six-step loop begins with authority and state divided by function. A complete outcome crosses several owners, ordinary change requires alignment, coordination becomes leadership and institutional value, moving rights creates concentrated losses, and the approvable answer adds coordination, making the original design survivable. AI enters at the alignment step by making local production cheaper, so more work reaches the same political boundary. A dashed escape moves ordinary rights, state and capability around a complete process, reducing crossings while retaining explicit genuine seams.
Alignment is not the cure for fragmented authority. It is the work generated by fragmented authority.
A standing alignment forum around ordinary change is a coupling detector. Coupling means that one part cannot change without forcing change on others; propagation is the toll each change pays at every boundary it crosses. When ordinary work must convene the holders of process, budget, data, control, architecture and release, the company is coupled in that precise sense, whatever the organisation chart says. Engineers call a system whose parts claim independence but must move together a distributed monolith. Chapter 4 shows why the identity is structural and where genuine interdependence limits it.
AI increases the pressure without changing the loop. It can reduce the cost of analysis, code and documentation, but the rights to choose, fund, integrate, control and expose the result remain divided. More proposals therefore reach the same six holders of authority over the complete result. AI operates at machine speed. Alignment operates at the speed of human politics. The comparison is not between intelligent machines and foolish people. Politics is the protocol by which legitimate and positional holders of authority understand a dispute, protect their interests and form an acceptable decision.
The loop can break. An incident, regulatory finding, acquisition, leadership change or sustained economic pressure may concentrate the cost of preserving it. A sponsor can also choose deliberately to test a better boundary before a crisis does the concentrating. The rest of this book is the design and evidence required for that smaller, harder move.
A recurring alignment meeting assembles partial owners because nobody can decide for the complete process. Minutes preserve the conversation but leave the process definition and software untouched; the same is true of data, contracts and controls. One decision then absorbs the delay until the meeting, private preparation, bridge work during it, approval afterwards, and another meeting when an absent stakeholder objects.
Temporary joint work remains necessary during an incident or while establishing a new seam. It may also support a decision that crosses legitimate decision-makers. Before you end it, record the ruling in the governing artefact, name who owns the result, and state when the joint work ends. To test a standing forum, you can route one class of ordinary decision around it while preserving legal and incident escalation. Stalled work reveals a missing decision right, conflicting actions reveal a missing contract, and silence suggests the forum was maintaining a ritual.
Composite
A payments director removes a weekly release forum after finding that most of its agenda is status. Two teams begin approving routine changes for themselves. Three weeks later, each makes a locally sound change on the same day and the combined effect breaches the trigger for a group liquidity control. The forum had been clumsy, but it was also the only place where anyone joined those exposures.
The payments director does not restore the whole meeting. She gives Treasury one declared stop right, a response time, and a route back to ordinary release. The other agenda items stay gone because meeting reduction was never the control.
Agility is the speed and fidelity with which evidence becomes changed operating behaviour. The loop begins with a customer or market signal, or with a control or operating signal. It ends only after the responsible unit changes the process and exposes that change safely, then observes the result and corrects course. Shipping code is one step inside it.
Many companies changed ceremonies; the right to change the process stayed elsewhere. The same holders retained commercial context, operating data, control, and authority over release. Scrum can coordinate local work and OKRs can express intent, but neither supplies the missing decision right. A team may complete every sprint yet wait weeks for the decisions that determine whether its output matters.
The loop can absorb its own diagnosis. A new operating model becomes a representation with maturity measures, communities and a reporting cycle. The company improves its score.
Ordinary decisions continue to cross the same owners. Repeated reorganisations change the route through which work waits without changing the reason it has to wait.
Process definitions, contracts, and evidence reports have stricter tests than a maturity model. A process definition fails when the live system follows rules it omits, a contract fails when either party cannot test its promise, and an evidence report fails when its claims do not lead to the behaviour compared. Consultants can still turn the vocabulary into a target operating model, and a central office can certify every unit yet retain its decisions. If decision rights have not moved and runtime cannot defeat the represented company, the framework industry has absorbed the criticism and returned it as merchandise.
Operating tests safeguard against framework capture; local vocabulary remains disposable. Can one owner decide an ordinary process change, can the affected systems and data change with it, can another unit rely on a published boundary, and can evidence reproduce the customer result? A design that answers those questions needs less explanation over time because its decisions remain in the mechanism instead of the meeting record.
If you work in the middle, your immediate task is smaller than reorganising the company. Trace one consequential decision from promise to operating result, show where rights and evidence separate, and ask the executive who owns the boundary to grant or refuse a bounded fix. A refusal identifies the veto that alignment usually hides; a grant creates something the rest of this book can make real.