The Five Focusing Steps

In the AI Story

Step One — Identify the Constraint. This is the step most organizations skip or botch. The constraint is the resource whose capacity determines the system's throughput, and its misidentification is the single most costly error an organization can make. A misidentified constraint leads to improvement efforts that produce zero system-level benefit while consuming resources that could have addressed the actual bottleneck. In the AI transition, the most common misidentification is the assumption that engineering capacity remains the constraint — leading organizations to hire more engineers, purchase more compute, and optimize pipelines that are no longer the weakest link. The correct identification uses the pile test: where does work accumulate? In AI-augmented organizations, the pile sits in the builder's review queue, in unreviewed alternatives, in the cognitive inventory of unevaluated possibilities.

Step Two — Exploit the Constraint. Before spending a dollar to expand the constraint, squeeze every unit of capacity from it as it currently exists. Exploitation means removing everything that wastes the constraint's time — every activity consuming judgment capacity without producing throughput. The builder should spend her evaluative capacity exclusively on decisions that require it: what to build, whether the implementation serves the intent, when to ship. Administrative overhead, routine generation, and context-switching all represent leaks from the constraint that exploitation must seal. Deep work conditions — uninterrupted blocks, protected calendar time — are the operational expression of Step Two in knowledge work.

Step Three — Subordinate Everything Else to the Constraint. This is the step organizations resist most fiercely, because it requires the deliberate underutilization of non-constraint resources. In a factory, subordination means running machines at less than full capacity so they produce only what the constraint can absorb. In AI-augmented work, subordination means limiting generation to what judgment can evaluate — directing the AI to produce five alternatives when judgment can assess five, not twenty because the tool will happily generate twenty. The resistance is visceral: unused capacity feels like waste. The resistance is wrong. Maximizing a non-constraint produces inventory, not throughput.

Step Four — Elevate the Constraint. Only after exploiting and subordinating should investment be directed at expanding the constraint's capacity. In the judgment-constrained organization, elevation means improving the builder's judgment — which responds to different interventions than the ones that improve mechanical resources. Judgment improves through experience, mentorship, structured feedback on decisions (not just outputs), and the slow accumulation of pattern-recognition capacity. Elevation also means protecting the conditions under which judgment develops: the buffer described in Drum-Buffer-Rope is simultaneously a system-management mechanism and a judgment-development mechanism.

Step Five — If the Constraint Has Moved, Return to Step One. Every successful improvement eventually moves the constraint. The judgment constraint, elevated through training and mentorship, may cease to be the binding constraint. The new constraint might be domain knowledge, AI capability in a specific area, or market absorption rate. When the constraint moves, every management practice designed for the old constraint becomes inertia. Goldratt considered this inertia — the persistence of old rules after the constraint has moved — the most dangerous force in organizational management, more dangerous than incompetence because it disguises itself as discipline.

Origin

Goldratt articulated the Five Focusing Steps in their mature form in The Goal (1984) and systematized them in subsequent works. The steps represent his attempt to operationalize TOC in a form that managers could apply directly without requiring the full theoretical apparatus. Each step corresponds to a specific error Goldratt had observed repeatedly in his consulting practice — managers failing to identify constraints, failing to protect them once identified, optimizing non-constraints in ways that degraded the system, and most tragically, continuing to apply successful improvements after the constraint had moved.

Key Ideas

Sequence matters. The steps are not a checklist of parallel activities but a strict sequence. Elevation before exploitation wastes investment; exploitation without subordination produces inventory; subordination without identification optimizes the wrong resource.

Identification is the hardest step. Most organizations botch Step One because the frameworks they have inherited — cost accounting, local efficiency metrics — are designed to obscure constraints rather than reveal them.

Subordination requires counterintuitive discipline. Deliberately running resources below capacity violates every local-optimization instinct. It is the mathematically optimal configuration of a constrained system.

The methodology is permanent. Step Five guarantees the process never terminates. The constraint always moves, and the discipline of re-identifying it is continuous.

Inertia is the deepest danger. The organization that continues to manage for the old constraint after it has moved is strengthening non-constraint links — and its discipline becomes the source of its decline.

Debates & Critiques

Critics argue the Five Focusing Steps are too simple to handle complex multi-constraint environments, and that real systems have networks of near-constraints whose relative binding shifts rapidly. Defenders respond that the critique confuses practical complexity with theoretical adequacy: the steps apply iteratively to whichever constraint is binding at each moment, and their sequential discipline outperforms attempts to manage multiple constraints simultaneously. In the AI context, a new debate has emerged about whether Step Four — elevating judgment — is fundamentally different from elevating mechanical constraints, given that judgment resists acceleration and improves only through slow experiential accumulation.