Read Your Own
Organization.

If you read about AI and organizational change, the future of work, how organizations make decisions, you know the gap this note is about. The writing has gotten good. The diagnosis is convincing, everyone agrees with it, and on Monday nothing in your organization has changed, because an argument about organizations in general was never a reading of your organization.

This week I shipped the thing that closes that gap. The Workbench is a free instrument for running the reading on your own organization: which of your stated values your systems quietly contradict, which of your decisions have collapsed into one seat, and which judgment calls have moved to AI without anyone deciding they should.

The reading you could not get.

Take the last offsite where everyone agreed on the diagnosis. Everyone could restate the problem. And nothing changed afterward, because agreeing that a problem exists is different work from locating it, and a repair starts from a location: which decision, which stated value, which handoff to AI, named specifically enough to change. Leaders have had two ways to get that location. You could hire consultants, who produce a reading of your organization at six or seven figures, months later, built from what people were willing to tell an outsider. Or you could read books, which give you the model and leave the work of applying it to you. Neither one equips the person who actually knows the organization from within: the one who was in the room when the decision went wrong.

Organizations already run diagnostics on every other capability they depend on. Finance has the audit, safety has the floor inspection, and security has penetration testing. Each one produces a reading of your own operation, with a verdict at the end and a repair order. Cognition, how the organization actually thinks, has never had a diagnostic, because thinking was never treated as something an organization runs. It was treated as something people bring with them, already finished, so there was nothing to inspect.

What went live this week.

That diagnostic now exists, in two parts. The first part has been on my site for a while: four framework pages at theautisticleader.ai, each carrying a live diagnostic with example cases already loaded. Their job is recognition: you see a finished diagnosis of a familiar failure before you run anything on your own. You load an example case, a culture to test, a decision to map, or a whole organization, and its verdict is already on the screen before you touch anything. The second part is the Workbench, where the same instruments run on your organization instead: a guided read, one question at a time, that builds your verdict as you answer. On the pages you see a diagnosis of an example case. In the Workbench you run the diagnosis on your own organization.

What each instrument reads.

They run in a fixed order: how information moves between people, how decisions get made from that information, how the thinking divides between people and AI, and whether the whole system holds.

Culture is the first read. Edgar Schein mapped organizational culture in three levels: the artifacts you can observe, the values you espouse, and the tacit assumptions that actually drive behavior. The most expensive cultural failure is the distance between the values you espouse and the assumptions that actually drive behavior, and that distance persists because of the one level you can directly design: how people are required to communicate and work, how meetings run, what gets recorded, whose input gets heard. That is where the Cognitive Translation Protocol operates. Its culture test asks only what you can observe, whether each value you espouse is written down, said out loud, and visible in how people behave, and what enforces it, and it decodes the assumption your systems actually run on, then names the stated value your systems contradict most and the one change that closes the gap. The protocol’s six failure dimensions and nine interventions came from my own life at the interface between people who think differently. Most people run social processing automatically. I run it as deliberate, manual work, so I have spent a career watching, from the inside, exactly where communication fails between two people who think differently.

Decisions are the second read. You already run a role matrix for execution: RACI names who is responsible, accountable, consulted, and informed on every task, and DACI, Intuit’s variant, brings that clarity to decisions inside a project. Both do their job, and neither names the judgment across the entire system. In every consequential decision there are three distinct cognitive roles to hold: who reads the situation, who makes the call, and who owns the outcome. Most organizations collapse all three into one seat, which was tolerable while execution filled most jobs and judgment could concentrate at the top. As execution moves to the machine, judgment becomes the work that stays, and it needs the role breakdown execution always had. DecisionOS is that breakdown, and it works alongside the role matrices you already run. Its map shows which of your decisions have collapsed, why the highest-paid opinion keeps winning the call, and the one separation that makes the decision hold under pressure.

The third read is the thinking itself. Your organization routes cognitive work between people and AI hundreds of times a week, and if nobody designed that routing, habit decides which work goes to the machine, and the judgment your people should own is the first thing habit hands over. The AI Cognitive Strategy Matrix is the allocation policy, and the move it makes before any routing is the one most AI conversations skip: you route the parts of a decision, not the decision itself. Take one restructuring call. Pulling the data together is one kind of thinking, modeling the scenarios is another, and making the call is a third, the step only a human can own and the one most often delegated by default. The matrix routes each step to one of three strategies: preserve it in human hands, enhance it with AI, or offload it to the machine entirely, with guardrails on what AI may and may not do at each, and its verdict says whether the judgment inside the decision is intact, at risk, or exposed.

The Organizational Judgment Diagnostic combines the three readings into one verdict on whether your organization’s judgment holds, and one repair order that starts at the most upstream break rather than the most visible symptom, because a break in communication cascades into every decision and every routing beneath it. That order is the part no single framework can give you.

Simple, on purpose.

None of that is worth anything if the instrument costs too much time or money to use, and most organizational diagnostics fail one of two ways. Some can only be run by an enterprise with a transformation budget, so the reading arrives a year late and belongs to a steering committee. The rest produce abstractions, so the output never turns into something a leader can do next week: the deck gets presented, and the organization goes back to work unchanged. So the Workbench is built to one standard: the reading has to translate into action. It is built for leaders of organizations of every size, a founder with twelve people or an executive inside a company of forty thousand. A reading runs in one sitting, one question per screen, and the verdict is on screen the moment you give the last answer. There is no login and no account, your answers never leave your browser, and the session ends when you close the tab. Every reading ends the same way: the exposure named in plain language, the one move to make first, and, when you run all three instruments, the order for the rest.

Where to start.

Start in the order the instruments were designed to be used. Load an example case on the framework pages first; the verdict is on screen before you touch anything, and the cases are written so you can pick the one that most resembles your own situation. Then open the evidence beside each instrument, the research and the reasoning behind each instrument, and argue with what you read; the instruments are claims, and you should test them against your own experience before you trust them with your organization. When you are ready, click the link marked run your own. It carries the reading you are looking at into the Workbench, and the guided read starts from there.

Tell me where it breaks.

I built these instruments from one perspective, and there are failures one perspective cannot see. The questions were shaped by what one cognitive architecture can see in an organization. The range has to come from organizations I have never seen, and that means your organization. So run a reading and send me what did not fit: the question that did not apply, the verdict that read wrong, and any move that would not have worked in your actual week. Reply to this note or find me on LinkedIn. From here, the essays take up the instruments one at a time, a series for each, the Cognitive Translation Protocol first, and what you send goes straight into how those essays get written and how the instruments get revised.

The personal claim behind this work was mine to make, and the argument was mine to write. The foundational layer of how organizations redesign the way they think, in the Age of AI, at the intersection of the Neurodiversity Movement and the Brain Economy, is the part nobody builds alone. It gets built the way the rest of this work has been built: in public, corrected by the people who use it. That correction is what the instruments are now open for.

You do not have to trust any of this before testing it, or even start with your own organization. Load one example case this week and read the diagnosis of someone else’s organization; that is the whole first step. The Workbench is open. Run a reading, tell me where it breaks, and the next version gets built with you. The architecture is yours to design.