Sheila Taormina had one of the finest aerobic engines her coaches had ever measured. She trained relentlessly — logging the yardage, following the program, doing everything the sport asked. And for years, the part of her she trained hardest was the part that was already world-class: her endurance. She kept polishing the strength she already had, while the thing actually limiting her went untouched.
Then she encountered an idea that had come from the factory floor, and it reorganized how she thought about her entire body as a system. The idea was Eliyahu Goldratt’s Theory of Constraints, laid out in his 1984 business novel The Goal. And the way Taormina applied it turned a stalled swimmer into the first woman in history to reach the Olympic Games in three different sports. Her story is the clearest possible illustration of the single most common, most expensive mistake organizations are making with AI right now.
Goldratt’s insight
Eliyahu Goldratt was a physicist who turned his attention to why factories underperform. His central insight, dramatized in The Goal through the story of a struggling plant, is deceptively simple: every system has exactly one slowest step — one bottleneck — and the output of the entire system is limited by that single step, no matter how fast every other part moves.
The consequences are counterintuitive and profound. Improving any step that is not the bottleneck produces no improvement in the output of the whole system. You can make the fast parts faster forever, and the system’s total throughput will not move by a single unit, because it is still throttled by the one slow step you did not touch. Effort spent anywhere but the constraint is, in terms of system output, wasted — however hard you work, however much it improves that individual step.
Goldratt called this the Theory of Constraints, and its practical discipline is a reversal of intuition. Do not improve everything. Find the one constraint that limits the whole system, and pour your effort there — because that is the only place effort translates into system-level improvement. Everywhere else, effort feels productive and changes nothing.
The swimmer who trained the wrong thing
Taormina, the story goes, took Goldratt’s factory-floor idea and pointed it at herself. She asked the question the Theory of Constraints demands: what is the one thing actually limiting my performance?
It was not her aerobic endurance. That was already elite — the part of the system that was working best. For years she had been improving exactly that, the fast step, the strength she already had, precisely because it responded well to training and felt productive to work on. Meanwhile, the real constraint sat untouched. At around 5’3″, small for an elite swimmer, she had never seriously developed her power and strength. That was the bottleneck — the one slow step throttling the output of the whole system — and it was the one part she had never trained.
So she changed everything about where she spent her effort. She found a coach who could help her attack the constraint, and she trained the weakness instead of the strength. In 1996, in Atlanta, Sheila Taormina stood on the Olympic podium with a gold medal — part of the U.S. 4×200-meter freestyle relay that set an Olympic record. She had made the team not by getting even better at what she was already best at, but by fixing the one thing that was holding the entire system back.
The proof it was the method, not the talent
What happened next is the part that proves the point. If Taormina’s success had been about raw talent, it would have stayed in the pool. Instead, she kept applying the same method — find the constraint, train the constraint — to entirely new sports.
After the 1996 gold, she took up triathlon, competing at the 2000 Sydney Games (finishing sixth) and the 2004 Athens Games, and winning the 2004 ITU Triathlon World Championship along the way. Then she took up modern pentathlon — a sport combining swimming, running, fencing, pistol shooting, and equestrian show jumping — and qualified for the 2008 Beijing Games. Four Olympics, three sports, across four different summer Games. She remains the only woman in history to have competed in the Olympics in three different sports.
You do not do that on a single physical gift. Each new sport presented a completely new system with completely new constraints — a new slowest step to find and fix. Taormina’s transferable superpower was not her body. It was the method: the discipline of finding the one thing that actually limits the system and training that, rather than polishing what already works. The Theory of Constraints was the real athlete.
Why organizations make Taormina’s original mistake with AI
Here is why this matters urgently in 2026. The single most common pattern in enterprise AI adoption is Taormina’s original mistake, at organizational scale: pouring AI into the part of the system that is already working, because it responds well and feels productive, while the actual constraint goes untouched.
Watch how AI gets deployed. It goes to the function that is already efficient and data-rich and eager — the part of the organization that is easiest to improve. The marketing team that was already fast produces content faster. The engineering team that was already productive ships code faster. These are real improvements to individual steps. And exactly like Taormina’s years of endurance training, they frequently produce no improvement in the output of the whole system — because the thing actually limiting the organization was somewhere else entirely, and AI was never pointed at it.
If your organization’s constraint is a slow approval process, a decision bottleneck, a single overwhelmed team, or a broken handoff between functions, then making your already-fast marketing team faster with AI does nothing for total throughput. The work still piles up at the constraint. You have made the fast step faster and left the slow step untouched — the precise error Goldratt spent a career warning against, now committed at speed and expense with AI.
Why the constraint is the one thing AI rarely gets pointed at
There is a cruel logic to why this happens, and naming it is the key to escaping it. The constraint is, almost by definition, the hardest and least pleasant part of the system to work on. That is often why it became the constraint — everyone has been avoiding it. AI, meanwhile, gets deployed where deployment is easy: the data-rich, well-understood, cooperative parts of the organization. The path of least resistance leads AI directly away from the constraint and toward the steps that were already working.
Taormina’s endurance was easy and rewarding to train, so she trained it. Her strength was hard and unrewarding, so she avoided it — until the Theory of Constraints forced her to see that the avoided thing was the only thing that mattered. Organizations deploy AI exactly where Taormina spent her early training years: on the responsive, rewarding, already-strong part of the system. And they get exactly her early result: enormous effort, real local improvement, and no change in the outcome that matters.
Mapped to the Mutation Readiness framework
The Theory of Constraints maps onto three dimensions of the Mutation Readiness diagnostic — the operational instrument of the Mutation transformation practice we run for enterprise leaders.
Signal Sensitivity — finding the constraint is a signal-reading discipline of the highest order. The constraint is often quiet, unglamorous, and easy to overlook precisely because attention flows to the parts of the system that are performing visibly well. Signal Sensitivity is what let Taormina read past her elite endurance to the silent weakness beneath it. An organization with high Signal Sensitivity finds the true bottleneck; one without it mistakes its strongest, loudest function for the place that needs the most help.
Ambidextrous Capital — the Theory of Constraints is a capital-allocation discipline. It says: do not spread investment evenly across every function; concentrate it at the single constraint, because that is the only place it produces system-level return. Applied to AI, Ambidextrous Capital means resisting the temptation to fund AI everywhere it is easy and instead funding it at the constraint, where it is hard — the disciplined allocation that actually moves the whole system.
Structural Flexibility — Taormina’s genius was that once she found a new constraint, she reshaped her entire training around it, and did so again for each new sport. That is Structural Flexibility: the capacity to reorganize the system around its true constraint rather than around its existing strengths. Organizations that can point their people, budget, and AI at the constraint — and repoint them as the constraint moves — have it. Those locked into optimizing their historical strengths do not.
The signals your organization is missing right now
The master signal is the gap between where your organization is deploying AI and where its actual constraint sits. If those two places are different — and they usually are — then your AI investment is Taormina’s endurance training: real effort, real local improvement, no system-level result.
Look for the specific tells. Where is work actually piling up in your organization — which step is the queue forming behind? Is that where your AI investment is going, or is AI going to the fast, easy, data-rich functions instead? Which of your AI initiatives have improved an individual team’s output without moving any outcome the whole organization cares about? Each is a sign you are training the endurance and ignoring the strength.
Three practical questions
One: what is your organization’s actual constraint — the one slowest step limiting the whole system? Find where the work piles up, where the queue forms, where everything waits. That is the bottleneck. If you cannot name it, that is the first problem, because you cannot be deploying AI at a constraint you have not identified.
Two: is your AI investment going to the constraint, or to the part that already works? Map where your AI spend is landing against where your constraint sits. If AI is improving your already-strong functions while the constraint goes untouched, you are making Taormina’s original mistake — and getting her original result.
Three: are you avoiding the constraint because it is hard? The constraint is usually the unpleasant, difficult, long-avoided part of the system — which is exactly why it became the constraint. Be honest about whether you are deploying AI where it is easy rather than where it matters. The whole return is at the hard place you are avoiding.
The closing thought
Sheila Taormina nearly walked away from sport having done everything right except the one thing that mattered. She had trained hard, followed the program, and improved relentlessly — all on the part of herself that was already elite. The Theory of Constraints gave her the one insight that changed everything: effort only matters where the constraint is, and everywhere else it is a beautifully executed waste. She found her constraint, trained it, and won gold — then proved it was the method and not the gift by doing it again in two more sports across four Olympics.
Most organizations are, right now, at the stage of Taormina’s career before the insight: working incredibly hard on the wrong part of the system, deploying AI where it is easy and rewarding, watching individual steps improve, and quietly bewildered that the outcome that matters will not move. The bewilderment has a cause, and Goldratt named it forty years ago. The system is limited by its constraint, and AI is not being pointed at the constraint.
You are probably not failing to work hard enough. You are, like the young Taormina, doing the work in the wrong place. Find your bottleneck. The constraint shows you where to focus — and that, and almost only that, is how you change everything.
The world has changed. The leaders who notice will be the ones the next decade is built around.
