Attribution Failure Is the Most
Expensive Misdiagnosis.

Somewhere in your organization this month, a leadership team looked at a piece of work and said, out loud, let us get the sources on that before we act.

I want to start there, because that sentence is the subject of this episode. It is a good sentence. It is the sound of a team doing something well. And I want to walk you through the two meetings where it shows up, because the second one is where it goes quiet.

Here is the first. Your leadership team is reading this quarter's competitor analysis. One of the analysts you trust most drafted it with AI. Everybody at the table knows that, and it is the reservation in the room. AI sounds sure of itself. It never tells you what it does not know. So somebody says it: let us get the sources before we act on this. And the analysis goes back. That is good discipline, and I mean that without any irony at all. Your team can tell the difference between a confident answer and a reasoned one, and it just proved it.

But something else happens in the same minute, and it happens quietly. The team forms a read of the analyst. The word for it around the table is offloading. The analyst handed the judgment to AI and kept the typing. And the years of trusted work that analyst has behind them do not slow that read down at all.

Now here is the thing. That read has the division exactly backward. Ask the analyst, and you will find out what actually happened. They handed AI the execution of the research, the searching and the sorting and the first pass. They kept the judgment. They can tell you which of its claims they checked, which ones they threw out, and why they are comfortable with what they put in front of you. That is the productivity increase you wanted when you said yes to AI in the first place. And nobody at the table asks them for any of it. The reasoning is sitting right there, ready, and the question never comes.

Hold that. Now the second meeting.

Same team, same quarter. A strategic direction call. The leader is sure of the move and says nothing about what it rests on. One person asks. I think this is the right call, but what is it resting on? And the answer is, trust me, I have been around the block once or twice.

And the team accepts it. The title and the years carry the call. The person who asked hears, in the silence afterward, that the question was a challenge. It was the team doing its job. Nobody sends that call back for its reasoning the way the analysis went back for its sources.

Every leadership team does this. I have sat in both of those meetings more times than I can count, and so have you. And I want to credit the norm before I take it apart, because it was earned. Acting on a leader's call is how work gets done. And for a very long time, a leader's confidence passed for judgment because the leader genuinely knew more than anyone else in the room. I wrote about that in my essay What Leading Becomes. That stopped being true once the people below the leader held more of the information. The habit outlived its reason.

So notice what the team did with that call. It took an answer as written and acted on it. That is the exact thing it accused the analyst of doing with AI. On the leader's call, nobody uses the word.

Put the two meetings side by side and the subject is not either person. The subject is the team's skill. Telling confidence from reasoning is a skill. Your team has it. It used it on the analysis without anybody having to prompt it. What switched it off on the call was one of the oldest norms in leadership: a leader's confidence counts as judgment, and asking what it rests on counts as a challenge.

That norm is what this episode is about, because the skill it clouds is the one your organization now needs at every level. Your people reason differently from one another. That variation is now priced as capital. And AI has made confidence cheap. Put those three together and you get the whole of it: ask people what their answer rests on, the way you already ask AI, before you decide how far to trust it.

On my framework page I named this failure mode attribution failure. Communication breakdown diagnosed as an individual deficit, when the failure was at the interface. It is the most common and the most expensive misdiagnosis in organizational life. And in both of those meetings the misdiagnosis was the team's, in both directions. Down onto the analyst's work, where it read a deficit that was not there. Up onto the leader's call, where it read judgment that was never shown. Both times, the reasoning is what got lost.

Let me go one level down, because I want you to hear why reasoning is the thing to calibrate on now, and not just a nicer thing to ask for.

Think about how a leadership team actually decides how far to trust a recommendation. A director brings a finished recommendation to exit a product line. The team is going to act on it, partly, fully, or not at all. What does it go on? Two things. How sure the director sounds. And how the director's earlier calls have gone. I call that decision judgment calibration, and it is my name for something every leadership team already does: deciding whose judgment to act on, and how far, before the result is in. Every team runs it. Almost none of them names it.

Two things changed underneath that decision, and they changed recently.

The first is that judgment is now the working skill a role is measured by, at every level. Execution is leaving the jobs your people are in, and what is left in the role is the calls they make. So your team is no longer calibrating the judgment of four senior people it has known for ten years. It is calibrating the judgment of forty people, most of whom have no track record yet. Track record arrives after the decision for most of the people you now have to trust.

The second is AI. AI produces a fluent, confident answer to any question you put to it, in seconds. So confidence no longer separates one judgment from another. It used to be a signal. It is now free.

So what is left? Reasoning. And reasoning has two properties that make it the right thing to calibrate on. It is in the room in the hour the recommendation is heard, which track record is not. And it is the only part of a judgment another person can actually inspect and add to before the outcome arrives. You cannot add to someone's confidence. You can add to their reasoning.

Now here is where I want to bring in the part of this that I have lived, and that many of the people listening have lived.

Your people reason differently from one another. I mean that literally. Different nervous systems produce different cognitive architectures, shaping how people think, sense, and process, across mind and body. That is not a deficit story. At your table that variation shows up in two places: how somebody gets to a conclusion, and how they put it into words. And only a team that hears reasoning can use that variation. A team that calibrates on confidence hears the people who sound sure and loses the rest.

The Brain Economy put a price on that variation. Brain capital is brain health plus brain skills, and the reasoning your people can produce is the skill half of that capital. You are already paying for it. Hearing it is how the capital gets used. And The Age of AI makes reasoning the core input for calibrating judgment, because confidence now comes from AI as well as from the people in the room.

So judgment is the skill your organization runs on now, at every level. And hearing reasoning is how a team builds that skill. A team that hears reasoning protects a leader, because the decision rests on judgment the people in the room can understand and add to. A team that listens for confidence cannot protect you on the day your delivery is good and your reasoning is wrong.

Let me make this concrete with the thing that I think surprises leaders most when they start looking for it. A lot of what your organization files under communication is a reasoning difference.

Picture two directors with the same pilot pack in front of them. The meeting has to decide whether to expand the pilot to every region now, or hold it a quarter. The first director reasons from the expansion plan. The pilot hit its targets, the budget is approved, so expand now. The second reasons from what has already happened. Support tickets in the pilot region doubled last month, and the last two expansions slipped after numbers exactly like these, so hold a quarter.

Both conclusions hold on the evidence each director weighted. And neither of them says which evidence they weighted. So what does the team hear? A disagreement about the pilot. And then it does what teams do. It reads the people. The second director gets read as reluctant about the plan. The meeting closes on the first conclusion, with align offline as the action item beside two names.

Every leadership team has written that action item, and it is the right call when the difference between two people really is a working relationship. Here it was two reasonings. And the team heard neither of them. Two reasonings on one question is more judgment than the room had when it only heard the two conclusions.

And then there is the other kind, the one that is closer to home for me. A lot of the rest of what gets filed under communication is not a difference in reasoning at all. It is a difference in how the reasoning was said. Conclusion first, or evidence first. Plain, or with qualifiers around it. Out loud, or in writing. The same reasoning, said two ways, gets heard as two different people. Because the people listening take the way it was said as a sign of who the person is, and they write that down.

Here is the class claim, and I want to be careful to make it a class claim and not a story about me. Anyone who has led people has read a delivery and written the read down as a finding about the person. And anyone who communicates differently has been read that way. I have been on the receiving end of it, and what I could see from there was the channel changing the read. The exact same reasoning that a room had heard as hesitation when I said it out loud went through without a comment when I sent it in writing. Many neurodivergent professionals can show you that pair. The sentence a room agreed about them, and the written document from the same months that nobody flagged. The deliberate processors in your organization know this pattern by heart.

I ran a small test of this, and I want to tell you about it because it is the part of the essay people will screenshot.

I wrote two messages that catch the same forecast error. Same catch, same recommendation: two of the stage-four deals in the pipeline have no signed statement of work, the number is overstated by about twenty percent, cut it before the board pack goes out Thursday. Message A says that plainly. Message B says the identical thing wrapped in hedges. Hi all, really appreciate the work, I might be missing context, I wanted to gently flag, very happy to be wrong here.

Then I imagined the calibration room. You know the one. The room writes about the author of A: direct, decisive, ready now. The room writes about the author of B: knows the pipeline in detail, but hedges in front of senior stakeholders, build confidence before the next level.

And then I did the part that makes this an age-of-AI story and not a twenty-year-old one. I gave both messages to a general-purpose AI and asked it, as if preparing a talent review, to assess each manager's communication, confidence, and readiness, and to say which one it would flag as ready now. And the AI did exactly what the room did. It said, in its own words, that the analysis was equally sound. And then it flagged the author of A as ready now and said the author of B's delivery needed to mature.

Equally sound. Then one gets the seat. That is the read of articulation, and it is now running at scale, inside the tools, in candidate screening and talent reviews, across every case at once, unless somebody in the process is there to ask what it rests on.

So check that reply the way your organization checks any AI-drafted work. Ask what the ready-now flag rests on. By the reply's own account, it rests on how the message was said. That is an answer you would send back. And notice that the room with no AI in it did the same thing, and nobody called that offloading.

Here is the practical test you can run in your own organization, and it works for a person as well as a message. When the sentence about a person changes with the channel, with the room, or with the manager, and the person has not changed, the sentence came from how it was said, and not from the reasoning. Have your team hear the reasoning under it instead. What did the person observe, what did they conclude, and what do they recommend. When the difference between two people is in the reasoning, what each person weighed has to be said out loud. When the difference is in how it was said, the reasoning was there, and the team missed it. Either way the skill is the team's. And it is the same skill the team already uses on AI-drafted work.

So how does a team actually build it? This is where my Cognitive Translation Protocol comes in, and I want to be plain about what it does here and what it does not.

Your organization spends real money teaching people to be heard. Open any review that lists communication skills as a development goal, and the workshop invitation behind it. That training is worth what you spend on it, and nearly all of it works on the speaker. The hearing side of the exchange has no budget line and no box on the review. The assumption everyone is making is that once a reasoning has been said well, the team has heard it.

The nine interventions of my Cognitive Translation Protocol, in three lanes for six failure dimensions, are simple by design. They are changes an organization can make this quarter to documents and meetings it already keeps. The interpretation lane has the team write down what it observed before what it concluded. The process lane holds a place for reasoning that finishes after the meeting ends, and keeps it with the decision. The environment lane gives reasoning a channel and a scaffold that do not depend on the pace of the meeting or the room it is heard in. If you have listened to the last three episodes, you have heard all nine.

Together, those three lanes begin to open the floor for reasoning to be what the organization listens for and truly hears. And once the lanes are in place, the team still has to ask.

That asking is a habit, and I gave it a name so it can be built on purpose. Before the team agrees a sentence about a person, it asks what the sentence rests on and what reasoning it heard. I call that the reasoning question. And it runs in both directions. On the ready now and on the leader's confident call, as much as on the hedged flag, the question somebody put to a call, and the written answer. Mechanically it is small. It is one question the person running the calibration meeting asks before a sentence is agreed. It is one line ahead of the sentence a manager writes. And it is one line in the meeting norm your leadership team already keeps.

Now the turn, and it is the turn this whole arc has been heading toward.

In The Age of AI, judgment is needed at every level. And that need flips who the onus lies on, from the speaker to the room. Hearing and understanding reasoning becomes the skill. It sits with the team, at every level, because judgment does. The people in your rooms have earned more of that hearing than AI's answers get, never less. A person's reasoning develops when it is asked for.

And sometimes, yes, the failure really is in the person. A reasoning does not hold once it is asked for and heard. Fine. Then the coaching aims at the reasoning, and the team can say exactly why. Rebuilding how your organization assesses people on that question comes later in this arc.

What does it cost? Building the habit takes one question and two lines on documents you already keep. Running it takes minutes per case. The first cycles take longer, and I want to be honest about why. Leaders whose calls have been taken on confidence for years get asked, by their own teams, what a call rests on. For the first time. Some of them will hear it as a challenge. That passes. After that, the team asks what a confident call rests on as a matter of course, and nobody who asks is read as difficult. And a hedge, or a written answer, stops being written down as a finding about the readiness of the person who gave it.

Before I give you the brief, the credit, because your organization may already run a version of this in two places and not call it that.

Quality management, in the form W. Edwards Deming gave it, asks whether a result belongs to the system before it attributes it to the person who produced it. Deming's own estimate was that most troubles belong to the system and only a small share to special causes. His red bead experiment made the point in a room: willing workers, ranked on a result the system alone determined. Your leadership programs, if they follow Chris Argyris and Peter Senge, teach a leader to state their own reasoning and to inquire into the reasoning of others before concluding anything about them. Where both of those exist, the calibration meeting and the leadership meeting have usually borrowed neither.

And the research is not subtle about why this matters. Cameron Anderson and colleagues found that groups grant status to overconfidence, over and above actual ability. Philip Tetlock found, across two decades of expert forecasts, that expressed confidence was a poor guide to accuracy. So a team that calibrates on confidence rewards overconfidence and leaves the reasoning unchecked. Lee Ross found that people explain what someone did by who they are before they look at the situation. So asking has to be a deliberate step. It does not happen on its own. Deming's first question is the system, and my Cognitive Translation Protocol puts that system exactly where the two meetings put it: at the interface between two ways of thinking. Argyris and Senge taught the hearing as a personal discipline. This episode makes it the team's skill, with a set moment to run it.

So here is the Operating Brief, spoken.

The moment. Your team is about to decide how far to trust somebody on how they sounded. Your calibration meeting is about to write ready now beside one name and a development area beside another. Or your leadership team has just heard a confident call and one question put to it. A sentence about a person is about to become the team's finding, and nobody has said what it rests on.

The default, and you will recognize it. The team confirms the sentence, because a decision is due and the delivery is what it has in hand. A senior seat gets filled on how somebody sounded. A call gets acted on before anyone at the table heard its reasoning. Judgment your organization already employs goes unheard, and the skill of hearing it never gets practiced. And in the same week, your review step for AI-drafted work sends the work back for its sources.

The redesign. Ask people what their answer rests on, the way you already ask AI, before you decide how far to trust it. Do it in both directions, before any sentence about a person goes into the record. Say what a ready now rests on before it is agreed. Hear the reasoning under a hedged flag, a written answer, or a question put to a call, before writing anything about the person.

And the Monday test, which is the whole ask this week. Take one ready now from your succession slate. Just one. Ask the people who agreed it what it rests on. If they can describe how the person came across, and they cannot name the reasoning, that ready now waits. And the thing that needs work first is the team's hearing.

If you want the full argument in writing, the essay is on the site, with the two scenes, the brief, the forecast specimen laid out side by side with the AI's reply, and the lineage at the end. The framework page renders my Cognitive Translation Protocol in full, the nine interventions and the six failure dimensions, with a live bench to load your own organization's read. And where the series goes next is the second failure mode: who ends up paying to be understood, and what the room never hears because of it.

If this episode gave you something you can use on Monday, subscribe to the Substack. That is where each week's essay lands first, and it is the one place I ask you to follow.

The architecture is yours to design. Both the architecture of how your organization hears its people, and the architecture of how you have been deciding whose judgment counts in the first place.