Every enterprise has a declared architecture and an exercised one.

The declared enterprise appears in organization charts, policies, capability maps, governance models, and portfolio processes. The exercised enterprise becomes visible in a different set of facts: whose objection ends a discussion, which questions never reach the agenda, whose forecast is accepted as authoritative, which unit can delay a decision without bearing the cost, and which changes die without anyone formally voting against them.

Business architecture is much better at representing the first than the second. That limitation is not a minor stakeholder-management gap. It is a power problem.

The field has developed disciplined ways to describe what an enterprise does, how value moves, which information the work requires, how organizations participate, and what must change to execute strategy. These views are necessary. They are also incomplete whenever they imply that the formal enterprise is the operative enterprise.

A capability may have an assigned owner, yet its funding sits elsewhere. A governance body may hold the official decision right, yet it depends on analysis produced by a group with strong preferences of its own. An executive may approve a target state, yet local leaders control the people, exceptions, and operational knowledge required to implement it. A transformation may promise enterprise value while concentrating its costs, risks, and loss of discretion in one part of the organization.

None of those conditions will necessarily appear on a capability map. Any one of them can determine whether the architecture is adopted.

This essay opens The Unwritten Enterprise, a Field Notes series about the informal systems that determine whether the formal enterprise works: power, incentives, status, coalitions, language, institutional memory, and the practical limits of authority.

The field has two power problems

Business architects often discuss power in personal terms. We ask how the practice can gain executive sponsorship, earn a seat at the table, influence investment decisions, or avoid being reduced to diagram production. That is a real professional problem. Many architects are expected to shape consequential decisions while holding little formal authority over the people, funding, or systems involved.

There is a deeper problem, however. Business architecture also underrepresents power as an object of analysis.

The first problem concerns the architect’s influence. The second concerns the accuracy of the architecture. Solving the first without solving the second would only give the profession greater influence while leaving it with an incomplete model of the enterprise.

A field that claims to connect strategy to execution must be able to explain more than the intended structure of a decision. It must explain how a decision can actually be made, what competing interests surround it, which resources implementation depends on, and where resistance can become delay, dilution, exception, or quiet noncompliance.

If we cannot do that, we are not yet describing the enterprise as it operates. We are describing its constitutional theory.

Power is not the same as rank, and politics is not always misconduct

Power is often discussed reluctantly because the word seems to imply manipulation, empire building, or bad faith. That definition is too narrow to be useful.

Power is the practical capacity to shape what the organization considers, decides, funds, implements, measures, or leaves unchanged. Formal authority is one source of that capacity, but it is not the only one.

Classic organizational research has described power arising from position, expertise, rewards, sanctions, information, dependence, reputation, and access to people who control valued resources. Later work drew attention to a less visible form of power: the ability to keep an issue off the agenda or to define the acceptable range of alternatives before a formal decision occurs.

All of these forms appear in ordinary enterprise work.

  • A finance leader may not own a capability but can determine which benefits count in an investment case.
  • A technology team may not own a business outcome but can make one option feasible and another prohibitively expensive.
  • A respected operator may have no formal veto but can determine whether frontline employees consider a change credible.
  • A risk function may shape the decision by defining what the enterprise is permitted to accept.
  • A data owner may control which performance claims can be substantiated.
  • A sponsor may control the agenda simply by deciding which questions deserve executive attention.

These actors are not necessarily behaving improperly. Interdependent organizations create power because people and groups control things that others need. Politics is what occurs when those actors use judgment, relationships, arguments, procedures, and resources to advance an interpretation or preferred outcome.

Some political behavior is self-serving or deceptive. Some protects legitimate obligations that a simplified target state has ignored. The architect’s task is not to assume either innocence or corruption. It is to understand the arrangement well enough to tell the difference.

The org chart is evidence, not ground truth

Business architecture correctly warns against equating a capability with the organization that currently performs it. Yet the field can become strangely literal when it turns from capabilities to authority.

An organization chart shows reporting relationships. A RACI assigns categories of participation. A governance model allocates formal decision rights. None of these proves how a consequential choice will be made.

The executive named accountable may defer to a trusted subordinate with deeper knowledge. A committee described as consultative may hold a practical veto because implementation requires the assent of its members. A central function may issue standards that regional operations routinely reinterpret. A program sponsor may formally control scope while a vendor, regulator, customer, or labor constraint determines what can actually change.

The gap between formal and effective authority is not an exception to be cleaned up after discovery. It is part of the current-state architecture.

This is especially important when we use the language of ownership. Saying that someone “owns” a capability can conceal several distinct powers:

  • defining the capability and its boundaries;
  • setting policy and performance expectations;
  • controlling its funding and workforce;
  • choosing the systems that enable it;
  • granting exceptions;
  • accepting operational or regulatory risk;
  • and being held accountable for the outcome.

Those powers frequently sit in different places. Assigning one owner may be a useful governance choice, but treating that label as a description of existing reality can obscure the dependencies that make the capability work.

The question is not merely who owns the box. It is who can cause the ability to perform, who can prevent it, and who bears the consequences when it fails.

A stakeholder grid is not a power analysis

Many architecture and change methods respond to organizational complexity with a stakeholder matrix. People are placed on a grid according to interest and influence, then assigned an engagement strategy.

This is useful as a first inventory. It is inadequate as a model of power.

The grid makes influence look like a stable personal attribute. Power is usually relational and issue-specific. A senior executive may have broad authority but little control over the technical evidence needed for a particular decision. A midlevel manager may appear less influential while controlling a scarce operational resource, an indispensable customer relationship, or the local conditions required for adoption. A subject-matter expert may be peripheral in the hierarchy but able to discredit the assumptions on which the business case depends.

The same stakeholder can also support one part of a target state and resist another. A local leader may welcome better information while opposing centralized exception authority. A technology group may support platform consolidation while rejecting the operating responsibility that consolidation would transfer to it. A risk function may endorse the strategic outcome while disputing the proposed sequence.

Labeling any of these people “resistant” compresses the analysis precisely where it should become more specific.

A serious assessment asks what each actor controls, needs, fears, owes, and stands to gain or lose. It also asks what evidence supports that interpretation. The purpose is not to psychoanalyze stakeholders or produce a secret ranking of allies and enemies. It is to understand the dependencies around a decision well enough to design a viable path through them.

Every target state redistributes something

Business architecture often presents the target state as a more coherent arrangement of capabilities, value streams, information, technology, and organization. If the logic is sound and the benefits are persuasive, stakeholder resistance can appear to be an adoption problem that begins after the architecture is complete.

In reality, the target state is already a political proposal.

It may redistribute budget, headcount, discretion, visibility, workload, expertise, status, risk, or the right to define performance. It may expose previously incomparable units to a common measure. It may turn a locally controlled exception into an enterprise rule. It may remove a manager’s ability to solve problems through relationships and place that authority in a shared service, a platform, or an algorithm. It may ask one function to surrender resources now so benefits can appear later in another function’s results.

Consider a composite example. An enterprise wants to centralize work intake and scheduling across several operating units. The target design promises better capacity use, more consistent prioritization, and a clearer customer commitment. On the architecture, the case looks strong.

The local units see additional consequences. They will lose discretion to reorder work for important customers. Their performance will become more comparable. Some experienced coordinators may move into a shared team. Local managers will still receive the first complaint when the centralized decision is wrong, but they may no longer possess the authority to correct it. The shared function will gain control but also inherit an operational obligation it has never carried.

No amount of communication turns those consequences into misunderstandings. Some resistance may reflect habit or parochialism, but some is a rational response to an operating model that has separated accountability from authority or concentrated sacrifice without designing a credible exchange.

Before describing resistance as a change-management problem, the architect should identify what the target state asks each actor to surrender and what institutional protection replaces it.

Many target architectures fail not because stakeholders misunderstand them, but because stakeholders understand their local consequences better than the architects do.

The most consequential use of power may be a nondecision

Architecture work naturally follows visible decisions: approved strategies, funded initiatives, declared priorities, governance forums, and active transformation programs. Power also operates through what never becomes visible enough to enter those channels.

An issue can remain permanently outside scope. A proposal can be referred for more analysis until its window closes. A metric can be defined in a way that makes one form of harm disappear. A pilot can succeed but never receive a scaling decision. A known dependency can be labeled an implementation detail. A governance body can repeatedly discuss a problem without naming a decision owner.

No one has to reject the change. Delay and ambiguity do the work.

This is why minutes from the formal decision meeting are often less revealing than the path by which the agenda was assembled. Who decided which alternatives were credible? Which costs were treated as real? Whose evidence was considered objective? Which stakeholder never entered the room? What would have happened if the organization had done nothing, and who could tolerate that outcome?

A portfolio can therefore be politically shaped before prioritization scores are applied. The scoring model may be mathematically consistent while the candidate set, benefit definitions, risk categories, and funding boundaries already reflect an allocation of power.

The architect does not need to allege conspiracy. Most agenda control is embedded in routines, language, and inherited governance. The obligation is simpler: do not mistake a procedurally tidy decision for a complete one.

Informal power can protect the enterprise

There is a danger in discussing the informal organization only as an obstacle to rational design. The same networks and tacit authority that frustrate a transformation may also preserve capabilities the formal model does not recognize.

Experienced employees know which policy exceptions prevent customer harm, which data cannot yet be trusted, which supplier relationship depends on personal credibility, and which apparent redundancy provides resilience during disruption. Informal leaders can translate enterprise intent into language that a local workforce trusts. Long-serving managers may remember why a control was created after the failure that justified it has disappeared from institutional memory.

An architect who sees only obstruction may remove the mechanism that has been compensating for a defective formal design.

This does not mean every workaround deserves preservation or every veteran deserves a veto. Institutional memory can defend obsolete arrangements as easily as valuable ones. The point is that informal power carries information. It tells us where the formal architecture depends on trust, scarcity, tacit knowledge, or unresolved ambiguity.

The proper response is investigation. What problem does the informal arrangement solve? Is that problem still real? Can the target state preserve the benefit while removing the fragility, inequity, or dependence on particular people? What evidence would show that the inherited control is no longer needed?

Power-aware architecture is therefore not cynical. It is less willing to confuse administrative neatness with institutional capacity.

Add a power overlay to the architecture

Power should not become another permanent map maintained for its own sake. It is better used as a decision-specific overlay on the existing architecture. For a consequential change, examine seven lenses.

Lens What to ask Evidence to seek
Authority Who can formally decide, fund, approve, accept risk, or grant an exception? Charters, policies, approval thresholds, delegated authorities, prior decisions
Dependencies Who controls people, money, data, systems, access, vendors, or relationships that the outcome requires? Budgets, service agreements, data rights, staffing models, contracts, operational handoffs
Interpretation Who can define the problem, the metric, the acceptable evidence, or the meaning of success? Business-case rules, metric definitions, reporting ownership, analytical methods
Agenda Who determines what is discussed, when it is discussed, and which alternatives are considered legitimate? Forum membership, intake rules, scope decisions, prioritization criteria, meeting history
Coalitions Which relationships can create support, credibility, or coordinated opposition? Repeated decision patterns, trusted intermediaries, cross-functional alliances, informal consultation
Execution Who can accelerate, delay, reinterpret, or quietly withhold adoption? Local procedures, exception rates, capacity constraints, implementation sequencing, adoption evidence
Exposure Who receives the benefit, bears the cost, carries the risk, or becomes more visible under the target state? Incentives, scorecards, chargebacks, regulatory accountability, customer commitments, role changes

The overlay should remain evidence-based. A formal authority can be documented. A practical veto should be supported by observed behavior, a resource dependency, or a repeated decision pattern. An inference about motives should be labeled as an inference and tested directly when possible.

Because this analysis can be sensitive, it also requires judgment about audience and permanence. The enterprise needs an honest account of implementation conditions. It does not need an ungoverned dossier of speculation about individuals. Focus on roles, dependencies, incentives, and observable behavior. Record only what is necessary to support the decision.

A power-aware impact assessment

The analysis becomes useful when it changes the design. A practical power-aware impact assessment can be completed in seven steps.

1. Define the decision, not merely the initiative

State the choice the organization must make and the outcome it is intended to produce. “Implement a new platform” is not a decision. “Adopt one enterprise intake and prioritization model, with common data and defined local exception authority” is closer.

Power attaches to specific choices. Vague transformation language allows stakeholders to endorse the aspiration while preserving incompatible interpretations of what will change.

2. Trace the formal path to authorization

Identify who recommends, who decides, who funds, who accepts risk, and who is accountable for benefit realization. If those roles do not align, make the gaps explicit before assuming governance will resolve them.

3. Identify the practical dependencies

For each affected capability and value-stream stage, identify the resources, information, expertise, relationships, and local actions required. Then ask who controls them and whether that control includes the practical ability to delay or alter implementation.

4. Make the distributional effects visible

Document where the target state changes discretion, workload, cost, status, accountability, risk, or access to information. Separate enterprise benefit from local effect. A net-positive business case can still contain concentrated losses that must be governed rather than explained away.

5. Test the stated objections

An objection may reveal a genuine design defect, a legitimate obligation, a negotiable interest, or an attempt to preserve advantage. Do not decide which one it is from the stakeholder’s title or tone. Ask what evidence would confirm or disconfirm the claim.

6. Design the political operating model

If the target state redistributes power, specify the reciprocal arrangement. New centralized authority may require transparent service levels, local escalation rights, shared measures, review points, or a retained exception mechanism. Standardization may require a legitimate process for variation. Greater visibility may require fair attribution of factors outside a unit’s control.

This is not a side agreement added to make people feel comfortable. It is part of the operating model.

7. Establish review and stop conditions

State what evidence will show that the new allocation of authority is working and what result would trigger redesign. A power transfer without a review mechanism easily becomes permanent even when its original justification fails.

The ethical line is visibility, legitimacy, and accountability

A power-aware architect can become more useful or merely more manipulative. The difference lies in purpose and method.

The purpose is not to help a sponsor outmaneuver every opposing stakeholder. It is to make the real conditions of enterprise action visible enough for legitimate governance to work. That includes surfacing valid dissent, concentrated costs, misaligned accountability, and informal controls that the preferred solution would rather ignore.

The method should be equally disciplined. Distinguish facts from interpretations. Do not attribute motives when incentives and dependencies explain the behavior. Do not disclose sensitive observations more broadly than the decision requires. Do not confuse access to leaders with authority to decide. Do not use architecture language to make a political preference appear technically inevitable.

The architect should be candid about one further fact: architecture itself can confer legitimacy. Naming a capability, drawing a boundary, assigning an owner, selecting a measure, or placing an option in the target state can make one interpretation of the enterprise appear official. The artifact does not merely record a debate. It can alter the debate.

That influence creates an obligation to show assumptions, alternatives, tradeoffs, and decision rights clearly. Neutrality is impossible, but procedural integrity is not.

Strategy execution is organized power

Business architecture often describes itself as a bridge between strategy and execution. Power is part of the bridge.

A strategy becomes executable when people with the necessary authority, resources, information, and credibility can act in concert; when conflicts can be resolved through legitimate governance; when those who bear new costs or risks understand the reciprocal arrangement; and when implementation cannot be defeated through ambiguity that no one is accountable to resolve.

Without those conditions, the target state may still be logically elegant. It is simply not yet an operating model.

The profession does not need to become cynical or Machiavellian. It should continue to value evidence, stewardship, transparency, and enterprise outcomes over factional advantage. It does, however, need to abandon the comforting fiction that politics begins only after the analytical work is complete.

Power shapes which problems become architecture work, which evidence enters the model, which options appear feasible, who receives ownership, and whether the adopted design survives contact with the organization.

A field charged with describing enterprise reality should not avert its eyes from the force that organizes it.

This essay is part of the Business Architecture Cookbook at StephenKlahr.com, a free, no-registration collection of field notes, tools, templates, study resources, and decision-centered recipes for real enterprise work.

No fluff. Just what helps.

Sources and Further Reading