Systemicity

Configuration dependence means an element’s effect is not fixed. Its contribution changes with the other elements around it and with how they are arranged. A capability that adds value in one configuration can add little or destroy value in another.

This is why many AI implementations are a disappointment. The technology has potential, but the systemic configuration doesn’t align to realize the potential.

Complementarity is a strong form of configuration dependence. Two or more changes produce greater value together than their separate returns would suggest. Evaluating each change independently misses the return created by their combination. Substitution runs in the other direction: adding one capability reduces the contribution of another. Interaction effects arise when conditions jointly determine an outcome, even though none of them has that effect alone.

These dynamics make component-level economics unreliable. The value of an investment depends on the configuration it enters, while the configuration that would make it valuable may not yet exist. The enterprise must judge not only whether to acquire a capability, but what else must change for that capability to produce the intended result.

Coupling describes how a change in one part alters conditions elsewhere. Propagation is the movement of that effect through dependencies. The effect can weaken, accumulate, or amplify as it travels.

This separates local performance from system performance. A local improvement can transmit costs to other activities; a local disruption can constrain activities that performed correctly. The size of the originating change does not determine the size of its eventual consequence. Strong coupling makes small changes consequential. Weak coupling contains them.

In a nonlinear system, outcomes do not change in proportion to inputs. A small change can have little effect until a limit is reached, then produce a sharp shift. Thresholds may arise from capacity, queueing, compatibility, or another condition that changes how the system responds.

Constraints govern how much the system can produce or absorb. Changing one constraint can shift pressure to another, so the original intervention changes which limitation governs the outcome. A forecast based on the previous constraint then misstates the new system’s behavior.

State dependence means an action’s effect depends on the system’s current condition. The same action can produce different results as capacity, accumulated wear, demand, knowledge, or other conditions change.

Path dependence means prior choices shape the current state and the options now available. They can make some future moves cheaper, harder, or impossible. Sequence dependence is related but distinct: the order of actions changes their costs and effects. In general, doing A and then B does not produce the same result as doing B and then A.

These dynamics make past performance conditional evidence. Results from one state, path, or sequence do not automatically predict results after the enterprise changes them. A decision can also alter the options against which later decisions will be judged.

A delay is the interval between a cause and its observable effect. During that interval, the enterprise may judge the action, repeat it, or extend it. The eventual consequence may arrive in another part of the system or appear in a different measure from the one that captured the original benefit.

Different parts of a system also change at different speeds. Commitments can harden while technologies, markets, capabilities, and behavior continue to shift. Analysis may be accurate when completed and obsolete by the time the enterprise acts. Delay and asynchronous change make timing part of the causal structure, not a scheduling detail.

Feedback occurs when an outcome changes the conditions that produced it. Reinforcing feedback amplifies a direction of change; balancing feedback limits it or pushes the system back. Either can make effects build over time rather than remain proportional to the initial action.

Adaptation adds purposeful response. People, organizations, and other actors change their behavior as conditions or policies change. Their responses alter the outcome and the evidence used to evaluate the intervention. The enterprise is not acting on a passive system: action changes the system that will produce the next result.

Systemic dependencies do not stop at the boundaries used to organize analysis. Relevant causes and effects can cross functions, organizations, markets, or other established categories. The boundary that matters is the one containing the relationships that govern the outcome.

Emergent properties arise from interactions among elements. They cannot be inferred by adding up the elements’ individual characteristics. The system’s performance depends on how those elements interact and on the conditions in which they do so.

Systemic change can also alter the possibility set itself. A new configuration can create value that no existing component or option possessed independently. The enterprise is then not simply choosing among fully formed opportunities. It is creating a configuration whose value depends on how its elements work together.