"A man should never be ashamed to own he has been in the wrong, which is but saying, in other words, that he is wiser today than he was yesterday."
Alexander Pope
In 2023 I published my first Blank Collar book. At its center was an earlier version of the framework, built the way you build anything you are willing to put your name on: slowly, out of things that had actually happened to me. It brought together integrated data, vision and mission, communication, process, knowledge, technology, user interface, user experience, augmentation, automation, and AI. I believed every part of it when I wrote it.
There is a date this chapter is not allowed to skip. The manuscript was finished before conversational AI became publicly prominent in late 2022, and the book was published in 2023, after that shift had begun. I let it ship while the first cracks were already widening, because a finished manuscript and a publication date carry a momentum that doubt is not invited to interrupt. The questioning had started in my head before the book reached anyone's hands. So the uncomfortable sentence has to come first: I published a framework whose audit I had already begun. If you bought that book, nobody told you this at the time. I am telling you now, which is late, and late is the accurate word for it.
Earlier in this movement I said that AI grades whatever it touches. This chapter is where I stop grading your world and hand you my own report card, because the framework this book gives you was not drawn fresh on a whiteboard. It is what was left of the earlier one after reality had finished with it, and you deserve to watch the demolition before you are asked to trust what stayed standing. A framework you never revise is a belief, not an instrument.
Where the framework came from
The components came from real problems. Across different kinds of work, in large organizations and small ones, I kept running into the same failures: technology bought with no direction behind it, knowledge that lived only in people's heads, processes nobody could describe, communication that reached everyone and changed nothing. Each encounter left a mark, and I did what a systematic person does with marks. I turned them into components. Data earned its place because I had watched decisions made on numbers nobody could trace. Communication earned its place because I had watched good plans die of silence. Knowledge, technology, interface, experience: each one was a real wound with a label attached.
The mistake was not in the observing. It was in what I did next, which was to stop there. I translated experience into components and never put the components through the test an instrument has to pass. And I can name the first design error precisely now, because I have watched other people commit it since. Every scar received a label. Accumulation felt like rigor. Each addition made the framework feel more complete to me. To everyone else it only made the thing heavier to use, and I read the weight as seriousness. A framework that grows with every experience is recording its author. It is not yet measuring anything.
The failure that should have stopped me was reproducibility, and I saw it and talked myself out of it. Two people could read the same organization through the framework and arrive at different results. Which component was weakest, what the reading meant, what to do first: on all of it, the framework offered no rule that could settle a disagreement between two competent users. I told myself this was depth, that the tool was a thinking aid and the divergence was nuance. The plainer reading was available the whole time. An instrument that cannot bring two readers to the same answer cannot change a decision consistently, and a framework that cannot change a decision consistently is an illustration. Good for explaining, useless for deciding, and not yet entitled to the word diagnosis.
There was a human failure folded inside the design failure, and it took me longer to see. The people who would have to carry any change the framework implied showed up in it as terms: knowledge to be captured, experience to be improved, communication to be increased. Reduced to labels, they could not object, adopt, refuse, or improve anything, and the framework had no way to tell the difference between a change people carried and a change people waited out. I had built a model of an organization with the organization's own people abstracted out of it. I had done it while believing that respect for those people was the whole point of the work. What I believed did not survive contact with what I had actually built.
Then the world ran my exam on me
Look. When conversational AI became public in late 2022, I did to my own framework what this book has spent its first chapters doing to your company. I pointed the amplifier at it and watched what it brought out. What follows is the audit, taken by error type rather than by anecdote, dated where the dating matters, because some of these are judgments from a fast-moving stretch and could age poorly themselves.
The first error was the one already confessed: the framework was not consistently runnable. Two readers, one organization, different results, no rule to arbitrate. Every other error sits downstream of this one, because an instrument that cannot bring readers to the same answer cannot be tested, and a structure that cannot be tested can only be believed.
The second was a category error. The framework mixed properties of products with conditions of organizations. Interface and user experience are attributes of a thing you ship; the framework claimed to diagnose companies. My dated observation from after 2022 is that chat-style interaction spread quickly and interfaces changed faster than any revision cycle could track. But the deeper fault needs no observation to stand. A product property has no business inside an organizational diagnostic, whatever interfaces do next. Those components were removed for the category error. The market only made the error easier to see.
The third was how the framework treated knowledge. I had counted accumulated expertise as a compounding asset, more of it meaning more strength. The framework never asked whether the knowledge was current, whether anyone could inspect it, whether it could be transferred, or whether the people who held it were free to challenge it. My reading of the years since 2022, offered as a reading, is that machine systems began to absorb precisely the kind of expertise I had been treating as ballast, which is uncomfortable for the old framework. The design fault stands without that. An instrument should have asked those four questions. Mine could not ask a single one.
The fourth was placement. Augmentation, automation, and AI sat in the framework as a suffix, a final modifier tacked onto a structure that was mostly about other things. By my own dated reading, AI was already changing the behavior of the other components when the book shipped: what data was worth, which processes could run without people, which expertise still commanded a premium. Pricing the most consequential force in the system as an afterthought is the kind of proportion error that invalidates the whole drawing. In the current framework this became a promotion, and the reason travels with it: a force that changes the behavior of the base cannot be a footnote to the base. The competing explanation, that AI's centrality is itself a feature of this moment and will fade, stays open, which is exactly why the new framework has to carry dated tests instead of my conviction.
The fifth was weight. Human experience, in the operating sense this book uses, whether people adopt, trust, can use, and can challenge what the organization runs, carried almost no operational weight in the earlier framework. The design argument holds on its own: a framework whose components can all score well while the people carrying the work refuse it, or route around it, is measuring the wrong things.
Underneath the five sat the load-bearing error, the one that still embarrasses me, because it was the one doing the protecting. Complexity stood in for rigor. The size and the interlocking parts made the framework feel serious, to me and to the people I showed it to, and the size did a second job I could not see at the time. A structure with that many interacting parts can explain any outcome after the fact, and a structure that can explain any outcome can name no outcome that would prove it wrong. The complexity was not incidental. It was the framework's defense against ever having to say what result would falsify it. The last chapter described organizations digesting threats to their machinery. I had drawn the same defense into a diagram and called it a model. What the audit produced is a shorter candidate instrument, and candidate is the word that carries the weight. Removing parts does not certify the parts that remain. That question gets its own section.
What survived the demolition
Five parts, and before I list them, the limit that governs the list. This audit was a lens read backward, run by the man who owned the lens. Retrospection with that much freedom can locate possible errors with some confidence. It cannot certify its own survivors, because the same author who chose the components I removed also chose which ones stayed, and choosing is not testing. The evidence, if it comes at all, comes from use I do not grade: the reading you will run on your own organization later in this book, and the failure conditions chapter twenty-two states without pretending they have already been tested.
So here is the arrival report, with the provenance of each part stated plainly, because the parts did not all arrive the same way.
Vision, Data, and Process were retained, and retained is not the same as vindicated. They stayed because they kept turning out to be where the trouble was when I went back through the failures I had seen, and each had to be made more operational to keep its place. Vision survives as a decision rule that returns yes or no, and mostly no. Data survives as the record of what happened, tested by whether it returns the same answer to the same question twice. Process survives as how work actually moves, tested by whether the flow survives the absence of any one person. The vague versions, direction as inspiration, data as an asset, process as documentation, did not survive, and the distance between those and the operational versions is the distance between what I published and what this book teaches.
AI was promoted, from a suffix to the force that changes the behavior of the base. That is a claim about where it belongs in the structure, and it carries the same humility as everything else here. If the technology's trajectory flattens, the promotion gets retested with the rest.
Human Experience was the late recognition, and I want to state its promotion carefully, because this is exactly where a framework's author is most tempted to reach for drama. It gained operational weight because adoption, trust, incentives, usability, and the freedom to challenge the system decide how much of any designed change is actually carried into practice. That is the claim, whole. Where those conditions fail, the other four can all score well and still describe a company that only works on paper.
Let me be plain about what I am not claiming for the five. I am not claiming every organizational failure fits them; a scheme flexible enough to sort anything after the fact earns no credit for the sorting. I am not claiming the number is sacred. Five is what survived this audit, and a sharper audit might cut further.
What I will claim is the standard the new framework has to meet, because the old one is the record of what happens without it. It has to point attention at an operating constraint specific enough to inspect. It has to change a decision: something gets done differently because of the reading, or the reading was decoration. It has to expose itself to a disconfirming result stated in advance, so that failure is recognizable when it shows up. And it has to be revisable by someone other than me, because a framework only its author can correct dies with its author's blind spots, and mine has already shown what those cost.
That standard turns into an audit you can run on any framework anyone hands you, this one included, in four questions. What decision does it change? What result would make you revise it? What did you last remove from it, and why? When will it be tested again? An owner with fluent answers is maintaining an instrument. An owner who hears the questions as an attack is defending a belief. The deletion question is the sharpest of the four, because a real removal comes with a reason attached and a consequence that followed, and a belief keeps no such records.
This chapter cost something to write, and I want to be exact about the cost, since exaggerating it would betray the whole point. The earlier framework still carries my name on a printed spine. Retiring it in public means every reader of that book can now ask why the next one deserves more trust, and the only answer I have is the one this chapter just performed: the new framework arrives with its own demolition attached, and its tests are set where I cannot reach them.
Movement I ends here. Its chapters argued that the pressure is real, that the calm is a property of averages, that the bubble cannot answer an operating question, and that the machinery, yours and mine alike, defends what it has accumulated. What comes next is the surviving framework itself, set out properly, after a short pause for one story told at the right distance. Do not extend the new framework the confidence I have just taken back from the old one. It is shorter. Shorter is not the same as right. Easier to test is the only claim I will make for it until the tests have run.