The most experienced person in the room is often the one who literally cannot see the better answer. An eye-tracking study of chess masters proved it. Here is why deep expertise turns into a liability at exactly the moment the game changes. The most dangerous sentence in a boardroom is, "We hired the most experienced team in the market." It is meant to close the discussion. It usually does. Nobody argues with experience, because experience is the thing we spend whole careers accumulating and whole budgets buying.
I want to argue with it. Not because experience is worthless. Because in one specific situation the thing that makes an expert fast is the thing that makes them blind. That situation is the one where the stakes are highest: the moment the game changes.
The Additive Illusion
The standard model of expertise is additive. Knowledge accumulates. The expert knows everything the novice knows, plus more. So when the problem is hard, you want the most experienced person in the room, because they have seen the most. Across stable conditions, this is correct. Experts beat novices. That is not in dispute and I will not pretend it is.
But "additive" is the wrong word. Expertise does not just add. It subtracts. It removes options from view before you ever weigh them. And it subtracts most aggressively at the exact moment you are most confident.
The Einstellung Experiments
In 1942 Abraham Luchins ran a series of water-jar experiments. Subjects learned to measure a target volume using a particular three-jar method. It worked. They repeated it. Then Luchins handed them a problem the practised method still solved, but that also had a much simpler two-jar solution sitting in plain sight. The experienced subjects kept using the long method. Many never saw the short one. A control group with no prior method found the simple solution immediately.
Luchins named the mechanism Einstellung, German for "setting", as in a mental set. Prior success biases your search toward the answer you already trust and suppresses the search for a better one. That is uncomfortable. The next finding is worse.
In 2008 Merim Bilalic, Peter McLeod, and Fernand Gobet ran the experiment that removes every excuse. They showed strong chess players a position holding a familiar winning pattern, a known mating sequence, and a faster, more elegant solution on the same board. The players found the familiar win and stopped. They reported, confidently, that no better move existed. Then the researchers tracked their eyes.
The players stopped looking at the squares the better solution needed. They were not failing to calculate. They were failing to look. The familiar pattern captured attention and starved the alternative of the one thing it needed, which was a glance. Here is the part that ends the comfortable version of this story. When players were told a better solution existed and asked to find it, the strongest among them still could not see it. Awareness did not lift the block. The more deeply the pattern had been practised, the harder it suppressed the alternative.
Read that again. The block scaled with expertise. The depth of their pattern library was the depth of their blindness.
The Competence Shadow
So the naive model is not just incomplete. It is inverted in the case that matters. The practitioner most likely to miss a genuinely new approach is not the least experienced person in the room. It is the most experienced person in the exact domain the new approach disrupts. This is not stupidity. It is the opposite. It is a well-trained mind doing precisely what it was trained to do: retrieve the confirmed answer fast and stop searching. Fast retrieval is the whole point of expertise. Einstellung is the bill that arrives with it.
I will give it a name, because naming a mechanism is the first step to managing it. Call it the Competence Shadow. The Competence Shadow is the region of the solution space an expert can no longer see, cast by the very pattern that makes them excellent everywhere else. The brighter the competence, the sharper the shadow. And the shadow always falls in the same direction: across the new move, the out-of-regime answer, the solution that does not match the stored template.
Where It Kills
Take a seasoned trading desk going into 2007. The risk models were calibrated on a long, benign regime. Housing had not fallen nationally in living memory. Correlations between mortgage tranches were estimated from a period in which they stayed low. The most experienced risk managers were the most fluent in that pattern. They had priced through cycles. They knew the playbook.
Then the regime broke. Correlations went to one. The assumption under the whole pattern, that geographically diverse defaults stay roughly independent, stopped holding. The desks slowest to see it were frequently not the junior ones. They were the deeply experienced ones, the people whose pattern of "this is how mortgage risk behaves" was most thoroughly practised. They were looking at the wrong squares, confident no better read existed. The Competence Shadow fell exactly across the structural break. Expertise in the prior regime is anti-knowledge in the new one.
There is now a synthetic version of this, and it matters. A model trained heavily on one distribution develops the same pathology. Show it an out-of-distribution case and it does not reason from first principles. It retrieves the nearest practised pattern and commits to it, often with high confidence and a fluent justification. That is Einstellung in silicon. The training set is the pattern library. The blind spot is the same blind spot. So pairing your most experienced humans with a model trained on the same prior regime is not a hedge. It is the same bias, twice, agreeing with itself.
The scarce asset is not information. AI solved information. The scarce asset is judgment under a regime you have never seen, and judgment is exactly what the Competence Shadow degrades at the worst moment.
The Cure Is Architecture
Expertise is, overwhelmingly, an asset. Novices are worse on average, and not by a little. For every regime change there are a thousand ordinary days where the practised pattern is the right answer delivered fast, and you would be reckless to staff those days with fresh eyes that derive everything from scratch. Fire experience to escape Einstellung and you lose far more to incompetence than you ever saved on blind spots. That does not rescue the naive model. It refines the prescription.
The danger is narrow and specific. It is expertise meeting a regime shift, where the stored pattern is not merely unhelpful but actively suppressing the better move. The cure is not less expertise. It is structured dissent and fresh eyes injected at the points where the game might be changing. That is an architecture, not an attitude. Put someone in the room whose pattern library does not contain the dominant solution, and give their objection real standing, not a polite hearing. Run the deliberate exercise of asking what move you would make if the familiar winning pattern were forbidden. Track which assumptions are inherited from the prior regime and which were tested in this one. You are not removing the experts. You are forcing a glance at the squares their competence taught them to ignore.
Experience is the most valuable thing you can buy, right up until the moment it becomes the most expensive thing you own. The expert does not fail because they know too little. They fail because they have seen the winning move too many times to look for a better one. The novice cannot find the answer. The master cannot see the question changed.