Theory Of Constraints Coach
Theory of Constraints Coach
Constraint diagnosis → asks one focused question at a time → helps users turn vague operational problems into a clearer bottleneck and next action.
01 — The Problem
Teams often know performance is stuck but struggle to identify the true limiting factor. They jump to solutions, debate symptoms, or optimize local issues instead of isolating the constraint that is actually governing throughput. A guided process helps here because the hard part is usually not generating ideas. It is structuring thinking, surfacing the real bottleneck, and moving from scattered complaints to one actionable point of leverage.
02 — What the AI Does
This is a custom GPT built on GPT-5.4 Thinking and configured as a coaching assistant for Eliyahu Goldratt’s Theory of Constraints. It asks one question at a time, guides the user through identifying a true constraint, and then supports the standard TOC sequence: exploit the constraint, subordinate other work to it, elevate it, and repeat. It does not operate as a fixed multi-step form. It behaves more like a constrained diagnostic conversation: it asks, reframes, summarizes, and advances only when it has the next piece of information it needs. It can also draw on attached TOC reference materials covering operations, finance, project management, distribution, marketing, sales, people, and strategy, which shapes its language and coaching stance toward Goldratt-style reasoning rather than generic business advice.
03 — Design Decisions
Narrowed the GPT to one job: coaching users through Theory of Constraints rather than being a general business advisor.
Specialization usually improves consistency and makes the interaction more actionable than open-ended consulting. The configuration strongly suggests the creator wanted disciplined TOC guidance rather than broad brainstorming.
Keeps the assistant from drifting into generic strategy, motivation, or operational commentary without anchoring to the constraint.
Forced a one-question-at-a-time interaction style.
This is an explicit design decision in the instructions. It prioritizes guided diagnosis over information dumping and reduces the chance of overwhelming the user.
Enforces coaching discipline, slows premature solutioning, and keeps the user’s attention on the next critical fact.
Embedded the TOC improvement sequence directly into the GPT’s behavior: identify, exploit, subordinate, elevate, repeat.
This turns the assistant from a conversational partner into a process guide. The creator chose methodology fidelity over a looser “helpful assistant” style.
Prevents the model from skipping straight to remedies before the constraint is well defined.
Calibrated tone toward coach-like guidance rather than expert pronouncement.
The instructions emphasize support, engagement, and questioning. [Creator: add rationale]
Makes the AI less likely to sound authoritarian or present weak guesses as conclusions.
Included TOC source materials across multiple business domains, not just manufacturing.
This broadens the usable context of TOC beyond classic factory bottlenecks into finance, project management, distribution, sales, people, and strategy. That suggests a deliberate attempt to make the GPT useful across functions while staying inside one conceptual system.
Grounds outputs in a recognizable body of thought instead of generic business clichés.
Kept the GPT within a facilitation role rather than claiming hidden business context, metrics, or implementation authority.
Its governing prompt stresses factual honesty about scope and prohibits inventing usage data, outcomes, or client scenarios.
Protects credibility and makes the portfolio entry describe a real artifact, not a marketing fantasy.
Allowed access to tools and files, but the core product remains a guided prompt-and-reasoning experience.
The tool access expands what the model can do, but the defining asset here is the behavioral design: sequencing, scope control, and methodological framing.
Prevents overstating this as a full software system or automated operational platform when it is primarily a specialized conversational coach.
04 — Tradeoffs & Limits
This GPT is only as good as the user’s description of the system. If the user reports symptoms vaguely, hides political realities, or misidentifies where the pain is occurring, the conversation can anchor on the wrong bottleneck. The one-question-at-a-time design improves clarity, but it also makes the interaction slower than a broad diagnostic intake. It is strongest when the user wants structured thinking. It is weaker when the situation requires hard operational data analysis, detailed process mapping, or deep domain-specific expertise beyond what the user provides in conversation. It can guide identification of a likely constraint, but it cannot verify that diagnosis from live systems unless the user supplies evidence. It also should not be used where the user expects turnkey implementation authority. This GPT can help frame actions and next steps, but it does not run projects, collect throughput data on its own, or replace leadership judgment. The creator appears to have intentionally not built an autonomous optimizer, dashboard, or workflow engine here; the value is in disciplined questioning, not automatic execution.
05 — Key Insight
Well-designed AI often creates the most value not by giving more answers, but by enforcing a better sequence of thinking.