GLP-1 wasn't developed for obesity. It was developed for type 2 diabetes — a metabolic disorder where the body's ability to process what it consumes breaks down. The weight loss was a side effect that became the headline. The mechanism that made it work got buried under before and after photos.

The mechanism is what matters. GLP-1 regulates intake by fixing the metabolic response to an environment that produces more than the body can process. It doesn't change the food environment. It restores the body's ability to function inside it.

Organizations have the same metabolic problem with data, and they've been treating it the same way — by focusing on the headline rather than the mechanism.

The data environment changed faster than organizational metabolism could adapt. Cheap storage, ubiquitous connectivity, and now AI made data available everywhere, all the time. The organizational capacity to convert data into understanding was never designed for that condition. Just as a lot of food is empty calories, a lot of information is empty zeros and ones. The emoji reaction that substitutes for thought. The dashboard view that registers as engagement and produces no decision. The AI summary that gets read and immediately forgotten because there was no question it was answering.

The corporate response has been to hire better data scientists, buy better analytics platforms, build better dashboards, and most recently inject LLMs into everything. All of it assumes the problem is insufficient intake or inadequate processing power. None of it touches the metabolism.

A faster LLM is a better fork. It doesn't help if the organization can't absorb what it's already consuming.

Data Metabolism is the rate at which an organization converts raw data into understanding and understanding into action. Every organization has one. Nobody measures it. The entire enterprise software industry has spent thirty years accelerating the intake end of that equation while the metabolism stayed roughly fixed, bounded by human attention, reflection time, and the organizational capacity to actually change behavior based on what was learned.

The flywheel spins. The transmission isn't engaged. More throughput into a broken metabolism is just a faster accumulation problem.

Knowledge Turns is how you measure whether the metabolism is working. Not how much data entered the system — how much of it completed the cycle from intake to understanding to action that wouldn't have happened otherwise. High turns means the metabolism is working. Low turns means the organization is data obese, and adding more sophisticated tooling is making it worse while looking like progress.

The GLP-1 insight wasn't that people needed to eat less. It was that the metabolic mechanism was broken and no amount of willpower or better food choices was going to fix a biological dysfunction. The data equivalent is just as uncomfortable. No amount of better tooling fixes an organization that never developed the capacity to learn from what it already has.

The question isn't whether your data is good. It's whether your organization can digest it.

How does that land?

Amazon didn’t solve its meeting problem by buying better meeting software. Netflix didn’t solve its management problem by building better approval workflows. 37Signals didn’t solve its communication problem by adopting a better messaging tool.

They solved it by redesigning how knowledge flows in their organization.

Gary Klein — a cognitive psychologist who spent decades studying how experts make decisions under pressure — gave every organization a thirty-minute practice that does the same thing.

The question most organizations avoid: does the right understanding actually exist in the right person at the right moment — or does everyone assume it does?

Most organizations never ask. They buy another tool instead.

Here is what four organizations did when they asked it — and one practice any organization can adopt tomorrow.

Amazon: Paragraphs Rather Than Bullet Points

In 2004 Jeff Bezos banned PowerPoint at Amazon’s senior leadership meetings.

This is usually described as a quirky management preference. It was something else entirely.

A presentation puts the understanding in the room during the meeting. The presenter knows the material. The audience receives it in real time, processes it imperfectly, asks questions from incomplete comprehension, and makes decisions from whatever partial understanding they managed to construct in fifty minutes.

A six-page narrative memo puts the understanding in the room before the meeting. The first twenty minutes of every Amazon leadership meeting is silent reading. The writer has to achieve clarity before the meeting because the room will notice if they haven’t. The readers have to achieve comprehension before they speak because their questions will reveal whether they did.

The meeting doesn’t create understanding. The meeting uses understanding that already exists.

Amazon added a second practice on the same principle. Before any significant product is built, a team writes a press release for it — as if the product already existed and was being announced to the world. Then a FAQ: every question a customer might ask, every objection a stakeholder might raise, every problem implementation might encounter. Only after the press release and FAQ exist does development begin.

This is called Working Backwards. It forces the understanding of what success looks like — and what the obstacles are — to exist before resources are committed.

Amazon also distinguishes between reversible and irreversible decisions and moves them at different speeds. Reversible decisions move quickly. Irreversible ones get more time, more scrutiny, more people whose understanding needs to exist before the commitment is made.

37Signals: Eliminating the Demand for Unnecessary Flow

Most organizations treat the speed of communication as a proxy for the quality of collaboration. If people are responding quickly, they must be working well together. This assumption generates a specific kind of noise — the expectation of real-time availability, the ASAP culture, the meeting that could have been an email, the email that could have been nothing.

37Signals eliminated it. No ASAP. No expectation of immediate response. Asynchronous communication as the default. People do the work and communicate when they have something worth communicating.

It is a redesign of what knowledge flow is required in the first place. When you remove the expectation of real-time response you remove the noise that was masquerading as communication. What remains is the signal — the thinking that actually needed to travel from one person to another.

The insight is subtractive. Organizations systematically undervalue subtraction — we add tools, processes, communication channels, meetings, and rarely ask what should be removed. 37Signals asked the question and found that most of what organizations call communication is actually interference with the knowledge flow that matters.

Better design. No new tools.

Netflix: Context Instead of Control

Reed Hastings at Netflix described their management philosophy in a phrase that sounds simple and is radical: context, not control.

Most organizations manage information flow through approval systems. Decisions travel up the hierarchy for authorization. Information travels down as directive. The knowledge that would enable good decisions at the front line is filtered through layers that each add latency and remove context.

Netflix inverted this. Instead of building better approval workflows, they invested in making sure the people making decisions had enough understanding of the organization’s goals, constraints, and priorities to decide well independently. The management job was not to control decisions but to ensure understanding existed in the right place.

Instead of routing decisions upward for approval — which treats knowledge as something that lives at the top — Netflix routes understanding downward continuously so that knowledge lives where decisions are made.

No approval software required.

Bridgewater: Making Reasoning Visible

Ray Dalio at Bridgewater Associates built radical transparency — every meeting recorded, every decision documented with its reasoning, every mistake examined openly.

In most organizations the reasoning behind decisions is invisible. The decision gets made, the outcome gets measured, and if the outcome is bad the organization cannot examine whether the reasoning was flawed because the reasoning was never recorded. The system has the decision. It doesn’t have the thinking.

Bridgewater records the thinking. Which means the thinking can be examined, challenged, and improved. Knowledge turns happen not just on what was decided but on how the decision was made. The organization gets smarter not just about outcomes but about the quality of its own reasoning.

This is expensive and uncomfortable. It is also the only way an organization can improve the quality of its judgment rather than just the efficiency of its processes.

Gary Klein: The Pre-Mortem

Before a project begins, Klein asks teams to imagine it has already failed. Not to worry that it might fail. To assume it did — and ask why.

The knowledge that a project might fail in specific ways exists before the project starts. It lives in the intuitions of experienced people, in memories of similar projects that didn’t work, in quiet concerns nobody is voicing because the momentum is toward optimism and raising concerns reads as not being a team player.

The pre-mortem creates a moment where that knowledge has to flow. The concern that was staying private gets permission to become visible. The understanding that something might go wrong surfaces before it’s too late to act on it.

One structured conversation. Thirty minutes. A room full of knowledge that was present but not flowing, made to flow.

No tool required. A single question asked before the project starts.

What They Have in Common

None of these organizations solved an information technology problem. None of these organizations bought their way out of it.

In each case the organization asked a question most organizations never ask: when does the right understanding need to exist in the right person, and what is currently preventing that?

The information management industry never asked this question because it cannot be answered with a product. You cannot sell the Amazon meeting structure as software. You cannot license the pre-mortem as a platform. The insight is organizational and behavioral — a design problem, not a technology problem.

Better design. Sometimes different tools followed. The tools were never the answer.

The Question Worth Asking

The question is not: how do we manage our information better?

It is: when does the right understanding need to exist in the right person for our decisions to go well, and what is currently in the way?

Amazon asked it in 2004 and banned PowerPoint. Netflix asked it and dismantled their approval hierarchy. Klein asked it and invented a thirty-minute conversation.

The tools exist to support whatever answer you find. They always did.

Most organizations are still waiting for someone to sell them the answer as software.

“Tell me how you measure me, and I will tell you how I will behave. If you measure me in an illogical way, don’t complain about illogical behaviour.”

— Eliyahu Goldratt, The Haystack Syndrome

The knowledge flow problem is largely invisible. That’s part of why it stays unsolved.

K-TIP 01: Amazon’s Six-Pager — Paragraphs Rather Than Bullet Points

Knowledge Turns in Practice — one organization, one practice, one question.

The Practice

Every significant meeting at Amazon begins with twenty minutes of silent reading. No presentation. No slides. A six-page narrative document — written in complete sentences, no bullet points — is distributed at the start of the meeting. Attendees read it before any discussion begins.

The writer must achieve clarity before the meeting because the room will notice if they haven’t. Readers must achieve comprehension before they speak because their questions will reveal whether they did.

Jeff Bezos introduced this in 2004. His description was simple: a presentation is easy for the presenter and hard for the audience. A narrative memo is hard for the writer and generative for the room.

What It Actually Does

The six-pager is not a document format. It is a knowledge turn built into the meeting structure.

In a standard meeting, understanding is supposed to emerge during the discussion. In practice it rarely does — people arrive with different levels of context, discussions are dominated by whoever speaks most confidently, and decisions get made from whatever partial comprehension the room assembled in real time.

The six-pager forces the knowledge turn before the meeting. The floor of shared understanding when discussion begins is higher than any presentation produces.

Amazon reports getting roughly twenty times as much information processed in the same meeting time — not because the document is longer, but because the understanding exists before the conversation starts.

The Question It Asks

In your organization, when does the understanding needed for this decision actually have to exist — and does it?

Most meetings assume understanding will emerge during discussion. The six-pager assumes it has to exist before. That assumption changes how you prepare and what the meeting is actually for.

They know how information is supposed to move. The org chart. The reporting structure. The meeting cadence. The documentation system. The approval workflow. These describe the designed path — the linear sequence from input to decision that somebody drew on a whiteboard and called a process.

What actually happens is different. The insight that changes the decision comes from an unexpected conversation in a hallway. The understanding that prevents the mistake lives in someone who wasn’t invited to the meeting. The context that would have reframed the entire problem existed in three different people who never spoke to each other. The decision got made anyway, on incomplete understanding, and the organization filed the outcome and moved on.

This is not a failure of execution. It is a failure of design. And it has been designed in from the beginning.

The Linear Approximation

When organizations began building information systems, they faced a genuine constraint. Knowledge — the real thing, the understanding that exists in people’s minds, the pattern recognition that develops through experience, the contextual judgment that resists articulation — couldn’t be put in a system. It was too distributed, too tacit, too dependent on the specific person holding it and the specific moment in which it was needed.

So they built systems for the next best thing. Documents. Records. Processes. Artifacts that captured the explicit residue of knowledge — the part that could be written down, stored, and retrieved. And they designed those systems around the assumption that knowledge moved linearly. From source to recipient. From input to output. From the person who knew to the person who needed to know, along a path that could be mapped and managed.

The approximation was reasonable. It was the best available option given the tools. A document is better than nothing. A process is better than chaos. A linear path is better than no path.

But an approximation treated as truth produces a specific kind of blindness. The organization stopped asking whether the approximation was accurate. The linear model became the design constraint for every system built on top of it. Which meant every system was designed for a version of knowledge flow that was always a simplification of the real thing.

What Knowledge Actually Does

Knowledge in organizations is not linear. It is emergent.

Understanding doesn’t travel from A to B along a designed path. It crystallizes at the intersection of partial understandings held by different people at different times in different functions. The person who understands the customer and the person who understands the technical constraint and the person who remembers what happened last time — when those three people are in the same conversation at the right moment, something becomes possible that none of them could produce individually.

That emergence can’t be stored. It can’t be retrieved. It can’t be documented after the fact in a way that recreates it. The filing cabinet can hold the output of the conversation — the decision, the record, the artifact — but not the understanding that produced it. The understanding was an event. Events don’t file.

This is what quantum means in an organizational context. Not the physics — the principle. The state of knowledge in an organization cannot be fully described at any given moment. It exists in superposition across the people who hold pieces of it. It collapses into a decision when the right collision happens. Trying to manage that by building better linear paths is like trying to catch light by building a better box.

The organization has been building better boxes for fifty years.

The Cost of the Approximation

The cost is visible once you know what to look for.

The decision that keeps getting made wrong. Not because people are incompetent but because the understanding that would change it exists somewhere in the organization and never reaches the person making the decision. The path was never designed. The collision never happened.

The project that fails in implementation. The strategy that was coherent at the executive level and incoherent at the point of execution. The understanding that the strategy required never crossed the functional boundary between the people who made it and the people who had to carry it out.

The institutional knowledge that walks out the door when someone leaves. Not because it wasn’t documented — sometimes it was documented thoroughly — but because the documentation captured the explicit residue of the understanding, not the understanding itself. The tacit knowledge, the judgment, the pattern recognition that made the person valuable — that was in their head and it left with them and the filing cabinet has a perfectly organized record of everything except what actually mattered.

Every organization has all three of these. Most have all three operating simultaneously, continuously, at significant cost. None of it appears on the dashboard because the dashboard measures the linear approximation, not the underlying reality.

What Designing for Emergence Looks Like

The question nobody has designed for is: how do you create the conditions for the right collision at the right time?

Not how do you store the information so it can be retrieved. How do you hold the context — the distributed partial understandings across functions, across time, across people — in a way that surfaces the intersection when a decision needs it.

This is what AI makes newly possible. Not because it’s smarter than the people in the organization. Because it can hold context at a scale and across a scope that no human and no previous system could manage simultaneously. It can notice that the understanding the sales team needs exists in a conversation that happened in customer success six months ago and a pattern that the finance team has been seeing for two quarters and a constraint that the engineering team documented last year. It can surface that collision rather than waiting for the hallway conversation that may never happen.

That is not a filing cabinet. That is not storage and retrieval. That is a system designed for the actual nature of organizational knowledge rather than for the linear approximation that was the best available proxy for fifty years.

Designing for emergence means starting with different questions. Not where does this information live. What understanding needs to exist in this person at this moment for this decision to go well. Not how do we document what we know. How do we create the conditions for knowledge to crystallize when it’s needed rather than after it’s too late.

Not what is the path. What is the collision.

The Design Constraint That Changes Everything

Every system built on the linear approximation asked: how do we move information from where it is to where it needs to be?

The right question is: how do we create the conditions under which understanding emerges in the people who need it, at the moment they need it, from the distributed context the organization already holds?

Those are different questions. They produce different systems. They require different metrics. They cross functional boundaries that the linear model never crossed because linearity respects the path and emergence doesn’t respect anything except the collision.

The organizations that start asking the right question will build something the current model cannot produce. Not incrementally better knowledge management. A different category of organizational capability — one where the understanding needed for a decision is available at the moment the decision is made, not stored somewhere that nobody navigates to in time.

The approximation served its purpose. It was the best available option for fifty years.

It is no longer the best available option.

The question is whether enough people notice before the next generation of systems gets built on the same wrong constraint, faster and more expensively, with better metrics for the thing that was never the point.

The industry defined the problems it could solve and called them the problems worth solving. Knowledge management meant storage and retrieval because storage and retrieval could be built, sold, and measured. Collaboration meant shared workspaces. Information flow meant faster access. The actual question — whether anyone understood anything better as a result — never made it into the specification.

For thirty years, an entire industry solved the wrong problem with considerable competence and complete confidence. The analysts validated the framing. The conferences amplified it. The buyers followed the experts. The experts followed each other. The loop closed

.

In the 1990s I worked alongside people who were trying to do something genuinely difficult — measure knowledge work the way manufacturing had been measured. The lean revolution had given us inventory turns, cycle time, throughput, work in progress. These were precise, actionable, honest. They told you exactly how well your factory was working and exactly where it was failing.

Knowledge work resisted every attempt at the same precision. The metrics that emerged — headcount, hours logged, tickets closed, documents filed — measured the wrong thing. They measured storage, not flow. Accumulation, not application. They told you how much knowledge you had. They said nothing about whether any of it was working.

We never built the right metric. We still haven’t. But the concept existed then and it exists now, and it’s more urgent than it’s ever been.

I want to propose one. Call it the Knowledge Turn.

————————

What Inventory Turns Actually Measure

In lean manufacturing, inventory turns tell you how fast stock moves through the system. High turns mean lean operations — material flows quickly from intake to production to customer. Low turns mean waste — stock sits, ties up capital, drifts toward obsolescence.

The insight behind the metric is simple: inventory that isn’t moving isn’t working. It’s just occupying space and costing money while pretending to be an asset.

Knowledge works the same way. Knowledge that isn’t moving — that sits in a database, a documentation system, a ticketing archive — isn’t working. It’s occupying server space and costing attention while pretending to be organizational intelligence.

The difference is that unlike inventory, knowledge doesn’t get consumed when it’s used. A part that gets installed is gone. A piece of knowledge that gets applied can be applied again — refined, updated, passed forward. Knowledge that turns well doesn’t deplete. It compounds.

Which makes low knowledge turns not just wasteful but compounding in the wrong direction. Every piece of knowledge that sits unused drifts toward entropy. It carries assumptions that decay. It reflects a moment in time that has passed. It hardens into something that looks like truth but is actually an artifact of a previous reality.

————————

The Physics of Organizational Entropy

Entropy is the default state. Without active energy input, systems drift toward disorder. This is true of physical systems and it is true of knowledge systems.

Every organization is fighting entropy whether it knows it or not. The question is whether it’s fighting it effectively or just generating the appearance of fighting it.

Most knowledge management systems generate the appearance. They capture. They file. They organize. They produce dashboards showing how much has been captured, filed, and organized. What they don’t measure — what they have no mechanism to measure — is whether any of it is actually working against entropy or merely encoding it at higher resolution.

A knowledge system with zero turns is a particularly expensive form of entropy. It gives the organization the confidence of having captured something without the benefit of having used it. It produces the illusion of organizational memory while the actual memory decays underneath.

Richard Feynman understood this. Reality must take precedence over public relations, for nature cannot be fooled. Entropy doesn’t care how well documented it is.

————————

The Knowledge Turn Defined

A Knowledge Turn is the complete cycle of:

Capturing a piece of context — not just information, but the conditions under which it applies.

Refining it into a signal — removing noise, surfacing what actually matters for a specific purpose.

Applying it to a problem — the moment the knowledge does actual work.

Archiving and updating it for the next person — with the context of how it performed and what changed.

An organization with high knowledge turns moves knowledge quickly from capture to application to update. Each cycle makes the knowledge more accurate, more contextual, more useful. The system gets smarter over time.

An organization with zero knowledge turns has knowledge storage. It piles up documentation like a digital hoarder — everything captured, nothing refined, nothing applied, nothing updated. The technicians eventually stop consulting it and start guessing. Which is rational. The knowledge system has made guessing more efficient than retrieval.

————————

The Information Utility Index as the Velocity Engine

This is where the Information Utility Index becomes useful as something more than a diagnostic framework.

The IUI asks two questions about any piece of knowledge: is it accurate, and is it useful? The resulting matrix has four quadrants. Intelligence — accurate and useful. The Data Lake — accurate and useless. The Working Map — inaccurate but useful. Noise — inaccurate and useless.

Most organizational knowledge management optimizes for the wrong quadrant. It pursues accuracy obsessively — more data, better data, cleaner data — without asking whether the accurate data is doing any work. The result is an organization full of Data Lakes. Accurate. Useless. Zero turns.

The IUI reframes the goal. The objective isn’t accuracy for its own sake. It’s utility — knowledge that actually moves through the system, gets applied, gets refined. Directionally correct beats precisely wrong. A working map that gets used beats a perfect map that sits in the archive.

High IUI means high knowledge turns. The knowledge is moving. It’s being applied. It’s being updated. It’s fighting entropy.

Low IUI means zero knowledge turns. The knowledge is sitting. It’s accurate and inert. It’s encoding entropy at higher resolution.

The IUI is the velocity metric that the knowledge turn requires. You can’t measure turns without measuring whether the knowledge is actually doing work at each stage of the cycle. The IUI tells you whether the knowledge moving through your system is intelligence or just expensive noise.

————————

The Refinery, Not the Warehouse

The metaphor that holds this together is the refinery.

A warehouse stores things. Its job is to keep what goes in intact until something takes it out. Success is measured by capacity and preservation. The warehouse is optimized for accumulation.

A refinery transforms things. Crude input enters, refined output leaves. The value isn’t in the storage — it’s in the transformation. A refinery that stops refining and starts accumulating crude is no longer a refinery. It’s a very expensive warehouse with a fire hazard.

Most organizational knowledge systems are warehouses pretending to be refineries. They take in crude context and store it at crude resolution. The AI layer being bolted on top of most of them is a faster, more confident warehouse. More intake. Better indexing. Same absence of refining.

A genuine knowledge refinery takes in context, removes noise, surfaces signal, applies it where it’s needed, and updates it based on what it learned in application. Each turn through the cycle makes the output more refined. The system doesn’t just store knowledge — it improves it.

That is what knowledge management was always supposed to be. We just never built the metric that would tell us whether we were achieving it.

The Knowledge Turn is that metric. The IUI is the engine that drives it. And entropy is what happens when neither is working — which is, at the moment, the condition of most organizations on earth.

————————

In seventh grade I went on a field trip to the General Motors assembly plant in Framingham, Massachusetts. Our social studies class had been studying the industrial revolution — the productivity gains and the displacements both. (Hat tip to Dr. Martin Sleeper for teaching us to think critically rather than just admire.)

What nobody mentioned was that the plant was already dying. A few years later Michael Dukakis would hold it up as proof of the Massachusetts Miracle — the economic revival that became the centerpiece of his 1988 presidential campaign. That particular plant had become a symbol of American industrial resurgence at the precise moment it was past saving.

Years later I read The Machine That Changed the World. Jim Womack and his colleagues had spent years studying that exact plant. It was their central example of everything wrong with the industrial model. Overstaffed. Inflexible. Producing defects at a rate that lean manufacturers in Japan had made unthinkable.

The plant closed in 1989.

I’ve been thinking about that gap ever since. Between the symbol and the reality. Between the confident presentation and what was actually happening underneath it. Between the system working perfectly and the system working.

When ChatGPT arrived in November 2022 the reaction was nearly universal. Transformative. Revolutionary. The moment everything changed.

My reaction was different. I thought it was impressive. And then I thought — all I can picture is two AIs talking to each other.

It wasn’t a breakthrough in human understanding. It was more generation, more output — more of the thing we already had too much of, produced faster and formatted better.

I had seen this movie before.

I wrote about this pattern last year using an example from my own professional past. In the 1980s CAD/CAM arrived with the same energy AI carries today. Engineers would think differently. Products would be conceived differently. The creative act of design would be fundamentally transformed.

What actually happened was more instructive. CAD/CAM accelerated the wrong layer. Drawing got faster. Decisions got slower. Dependencies exploded. And as one engineer from that era put it — once everyone used the same CAD system, designs started looking the same.

The tool got better. The thinking stayed the same. The biology didn’t move.

AI is repeating this at a scale CAD/CAM never approached. Not just in one discipline but across every form of knowledge work simultaneously. More generation. More output. More of the thing we already had too much of, produced faster and formatted better. And underneath it — the same unexamined assumptions about what knowledge is, how it moves, and what it’s for.

The legal disclosure is the perfect metaphor for what we have built.

Lawyers write it for other lawyers. It gets updated when a policy changes or someone finds a loophole — not when users fail to understand it. The person clicking “I agree” was never really the intended audience. And it isn’t genuinely negotiable — you can decline to agree, which means declining to use Chrome, or Gmail, or any of the infrastructure modern life runs on. Even regulations designed to protect users get captured by the same logic. GDPR was meant to restore meaningful consent. It produced the cookie banner — a popup that interrupts everyone thousands of times a year to obtain permission that nobody meaningfully gives. The compliance theater replaced the protection it was supposed to provide.

The disclosure performs informed consent without producing it. It feels responsible. It protects the institution. It communicates nothing to the person it was nominally written for.

AI meeting summaries work the same way. Accurate. Formatted. A confident record that the meeting happened and that things were said. The tension in the room — the idea that almost got said, the moment when someone’s face changed, the thing everyone understood but nobody wrote down — none of that survived the summarization. The left hemisphere produced a perfect record of what the right hemisphere actually understood.

We did not get more productive. We got better at generating the appearance of productivity.

Here is what I think is actually happening.

We are still running a model of knowledge work that emerged from 1913 and was never questioned.

In that year Henry Ford’s factory in Highland Park demonstrated something the world had never seen — work decomposed into repeatable steps, optimized, measured, and scaled. Frederick Taylor had provided the intellectual framework. Ford made it undeniable. What followed was not just a manufacturing revolution. It was an invisible template for how every organization on earth would come to think about knowledge, decisions, and coordination.

There is a detail about 1913 that matters enormously. The physical systems — the assembly line, the division of labor, the measurement apparatus — were deliberately and precisely engineered. The knowledge architecture underneath them was never designed at all. Nobody decided how understanding would travel through the organization. How meaning would move from the factory floor to the boardroom. How assumptions would get surfaced and questioned. That part accumulated. It was assumed. It became organizational common sense.

There was a reason nobody noticed the gap. The hierarchy filled it. Information traveled up one chain of command and down another — and for most of the industrial era that was sufficient. The environment was stable enough, decisions were slow enough, and information was scarce enough that the org chart functioned as a knowledge system. It wasn’t designed as one. It just worked as one, well enough, for long enough that nobody thought to ask whether something better was possible.

That condition no longer holds. It hasn’t held for decades.

Slack, Teams, and email created informal channels that work around the hierarchy for day to day coordination. That’s genuinely useful. But they didn’t replace the hierarchy as a knowledge system. They created a parallel layer alongside it. Now you have the formal hierarchy still making decisions based on information that traveled up and down the official chain — distorted at every handoff as before. And underneath it a vast unstructured layer of threads and messages and email chains where the actual thinking happens, the real decisions get pre-made, and the context lives that never makes it into the formal system.

Two patchwork emergent knowledge systems. Rarely talking honestly across the divide.

The hierarchy has the authority. The informal layer has the understanding. The gap between them is where organizational disasters quietly incubate. The night before the Challenger launch, the engineers who understood the O-ring risk and the managers who approved the launch were looking at the same information. They interpreted it differently — because different incentive structures were bearing on them. The knowledge system had no mechanism for surfacing unwelcome truths to people with strong reasons not to hear them. I’ve written about this in more depth here.

We inherited this whole. Not as a blueprint that could be revised. As the invisible water we swim in.

The industrial model then migrated further. Into job descriptions. Org charts. Performance metrics. Knowledge bases. Ticketing systems. And into the education system that produces the people who run all of those organizations.

Not because anyone decided to teach compliance. The reward structure did it quietly. Education is built around demonstrating that you’ve absorbed the existing explanation — finding the right answer, the one the teacher already knows, the one that gets graded. Generating better explanations is mostly irrelevant to whether you succeed. Nobody tells you not to ask questions. The system just makes asking questions mostly beside the point.

David Deutsch would call this teaching people to memorize rather than to conjecture and refute. The factory needed people who could do the former. The age we’re entering desperately needs people who can do the latter.

AI doesn’t fix this. It deepens it.

Every assumption baked into the industrial model gets more embedded, more confident, and harder to question. Bad explanations get generated at volume. Wrong questions get thoroughly answered. The organization mistakes fluency for understanding and output for insight. The factory doesn’t run faster. It runs deeper into the wrong direction.

And this is where the title of this piece means two things simultaneously.

AI changes everything in your organization’s environment. Your inbox. Your meetings. Your competitive landscape. Your industry’s economics. The transformation is real.

But it changes nothing about how human beings actually think. The biology hasn’t moved. The same brain that struggled to make meaning in 1913 is struggling to make meaning now — the same cognitive limitations, the same capacity for motivated reasoning, the same tendency to mistake fluency for understanding. Iain McGilchrist spent two volumes demonstrating that the western mind has been systematically suppressing its own capacity for contextual, relational, meaning-making intelligence. That suppression didn’t get a software update in November 2022.

The gap between a transformed environment and an unchanged biology is where the real problem lives.

Most organizations are running on explanations that accumulated over the last century and were never examined.

The industrial model wasn’t just a management technique. It was an explanation of what knowledge is, how it moves, and what it is for. Capture it. Codify it. Control it. That explanation made sense when information was scarce and coordination was expensive. It makes no sense now. And AI doesn’t update the explanation. It just runs it harder.

Every explanation is provisional. The best available understanding at a specific moment, always replaceable by a better one. The industrial model was a genuinely good explanation for its time. The problem isn’t that we adopted it. The problem is that we stopped looking for what comes next.

That is what Best Practices really means. We found an answer and stopped asking the question.

We are at a precipice

A small number of organizations will recognize the sorting event for what it is and ask the question the industrial model was never designed to answer: what would knowledge work look like if we were designing it today, from scratch, for the world we actually live in?

The rest will let Microsoft decide what knowledge is worth capturing. They will let Google determine what connections are worth surfacing. They will let OpenAI define what a good answer looks like. Those are not neutral technical decisions. They are the most consequential strategic choices of the next decade, being made by people whose interests are not yours, at a moment when most organizations don’t realize the choice is being made at all.

We remember the Amazons. We forget the Kmarts.

The foundational claim of the work I’m developing is simple.

Organizations treat knowledge as permanent. It never is.

Every piece of organizational knowledge is a best available explanation at a specific moment in time. It carries assumptions that decay without anyone noticing. It fractures as it travels. It hardens into policy and process and eventually into something nobody questions because nobody remembers it was ever a choice.

The Framingham plant closed because the explanation it was built on stopped being true. The people running it had every incentive to believe otherwise. The data they were looking at told them things were fine.

New organizations have always had one advantage — they get to start with modern technology rather than inherit their predecessors’ decisions. That same opportunity now exists for knowledge architecture. Most organizations are running information systems built on industrial era assumptions. The question is whether anyone is willing to build something optimized for how knowledge actually behaves.

That is the argument. The book that develops it .is coming. If you’ve been watching the gap between what the system says is happening and what you know is actually happening — you’re in the right place

Move fast and break things. We knew things would break. What we didn't account for was losing the ability to know what broke, when it broke, or whether it was breaking at all.

That's the world developing at the Speed of AI has delivered. And it arrived before most organizations noticed they were in it.

In the late 1990s, "developing at Internet Speed" felt like the most dangerous idea in business. It wasn't. It was still humans moving fast. Humans were cutting corners, shipping earlier, iterating in public — but humans were still in every step. The judgment, the context, the understanding of what was being built and why — that was still in the room, even if the room was on fire.

What's happening now is different in kind, not just degree. It's not humans moving faster. It's execution happening at a scale and speed that removes humans from the loop entirely — not by choice, but by arithmetic. You cannot verify what you cannot observe. You cannot observe what moves faster than human attention can follow.

A new economics paper from MIT, Washington University, and UCLA — "Some Simple Economics of AGI" by Christian Catalini, Xiang Hui, and Jane Wu — has been making rounds in the right circles. It deserves a wider one.

The paper's central argument: as AI decouples cognition from biology, the binding constraint on economic growth is no longer intelligence. It is verification bandwidth — the scarce human capacity to validate outcomes, audit behavior, and understand what the systems we've built are actually doing.

Catalini and his colleagues arrived at this conclusion through economic modeling. The work here has been arriving at the same place through organizational practice — specifically, through the question of what happens to knowledge integrity when the systems organizations depend on were never designed to preserve it in the first place. [The Risk That Doesn't Have a Date explores this in full.]

When two lines of inquiry converge independently, it is usually a signal that something real is underneath. This is that signal.

Everything Is Happening At Once

The dangerous thing is not that any one of these is happening. It's that all of them are happening simultaneously, at speed, and the window for doing something about it is closing faster than most organizations have noticed.

Consider what is converging right now, in the same moment, inside the same organizations:

Agents are coordinating agents. Until recently, AI felt like a tool you handed tasks to one piece at a time. Something shifted in late 2025. Agents stopped feeling like tools and started feeling like coworkers — ones you could brief, send away, and check back in with. The next step, already underway, is agents that don't just execute long-running tasks but coordinate with other agents to do it. The human is no longer in the loop between steps. They are only at the top and bottom. Everything in the middle is unmeasured, moving fast, and largely invisible.

The Codifier's Curse is running at full speed. Catalini names a dynamic that most organizations are living inside without knowing it. The people with the deepest tacit knowledge — the senior practitioners, the domain experts, the people who know why the decision was made and not just what it was — are precisely the people being asked to create training data, write evals, and document processes that will eventually automate their domain. They are doing the work of their own replacement, rationally and willingly, because it's valuable work that gets rewarded in the short term. The cruel mechanic: the better you are at your job, the more valuable your codification work is, and therefore the faster you accelerate your own obsolescence. Expertise is being weaponized against itself.

The Missing Junior Loop is already broken. Human expertise is a stock that gets built through the friction of doing — through entry-level work, through mistakes made under supervision, through the apprenticeship pipeline that has always been how organizations grew their next generation of senior practitioners. That pipeline is being cut. When AI handles tier-1 work, the junior role that used to be the first rung of the expertise ladder disappears. Organizations are eliminating the very process by which they would have built the verifiers they will desperately need. They won't notice for five years. By then it will be too late to rebuild quickly.

Shadow AI is already proliferating. This is the Trojan Horse at the individual level. Employees are running agents to do work that used to take days in hours, delivering output that looks identical to what they would have produced, and the organization has no mechanism to detect it. The deliverable passes review. The metrics look fine. The work gets done. But the organizational understanding that should have formed through the work — the tacit knowledge that develops through doing, the judgment that builds through struggle — never forms. The employee isn't learning. The organization isn't learning. The output exists. The understanding behind it is hollow.

And all of this is landing on infrastructure that was already broken. The CHEF environment — the Complex, Hidden-dependency, Expensive, Fragile technology stack that most organizations have accumulated over thirty years of applying technology to the explicit layer while the tacit layer evaporated [explored in full here]— was already degraded before any agent touched it. Copilot doesn't know why the decision was made. It only knows what's in SharePoint. And what's in SharePoint is a graveyard of documents that nobody updated, built on assumptions that nobody preserved, reflecting a version of the organization that may no longer exist.

Each one of these alone would be manageable. All of them at once, moving at the Speed of AI, on a broken foundation — that is the actual danger.

The Measurability Gap Is an Organizational Problem, Not Just an Economic One

Catalini's paper formalizes something called the Measurability Gap — the widening structural distance between what agents can execute and what humans can afford to verify. The paper models this as an economic phenomenon with policy implications for governments, investors, and firms.

That framing is correct and important. But the Measurability Gap is also, right now, today, an organizational problem that practitioners are living inside without a name for it.

Every organization deploying AI agents is creating unmeasured activity. Not because they chose to — because they had no choice. The execution scales. The verification doesn't. The gap opens automatically, structurally, the moment you deploy an agent capable of acting faster than a human can observe.

What accumulates in that gap is not just risk in the abstract economic sense. It is the specific organizational knowledge that was never captured — the why behind the what, the reasoning behind the decision, the context that would allow someone to understand not just what the agent did but whether it did the right thing for the right reasons in a way that can be trusted going forward.

This is knowledge integrity failing at a new speed and a new scale. The failure mode is the same one that killed the Chief Knowledge Officer thirty years ago — invisible, undated, unattributable. The difference is the rate of accumulation. What used to take years now takes quarters. What used to take quarters now takes weeks.

The Window

Catalini's paper identifies three countervailing forces that can prevent the drift toward what he calls a Hollow Economy — high nominal output, collapsing actual value, organizations generating activity that looks like understanding and functions as sediment.

Those forces are observability, accelerated mastery, and graceful degradation. They operate at the infrastructure and policy level. They are real and they matter.

But there is a more immediate intervention available, at the organizational level, before any infrastructure exists and before any policy gets written.

It is the question of who, exactly, is responsible for what your organization actually knows.

Right now, in most organizations, the honest answer is nobody. [The organizational response to that question — the Knowledge Integrity Officer and the Knowledge Steward — is explored here.]

The security team is trying to stop employees from pasting internal data into Claude and ChatGPT. That concern is legitimate. But the reason employees route around internal systems is that the internal systems don't give them anything useful. You can lock down the perimeter all you want. If the knowledge infrastructure inside the perimeter is so degraded that people have to go outside to get work done, you haven't solved the problem. You've made the workaround more expensive and more furtive.

The KIO isn't the security team's adversary. It's the function that makes the security team's job actually solvable — because employees who have good internal knowledge infrastructure don't need to take the risk in the first place

What the Speed Changes

Internet Speed changed the pace of development, but humans could notice what was breaking and course correct. The knowledge infrastructure that degraded during that period degraded slowly enough that organizations could pretend it wasn't happening and mostly get away with it.

The Speed of AI doesn't offer that grace period.

Agents coordinating agents means the unmeasured activity is no longer linear — it's compounding. The Codifier's Curse means the tacit knowledge being converted into training data is gone from the organization the moment it's captured — it doesn't return when the model improves. The Missing Junior Loop means the human capacity to verify, already scarce, is being structurally eroded at the same moment verification becomes the most valuable thing a human can do.

These forces don't wait for the organization to notice them. They operate regardless of whether anyone is watching. That is precisely what makes them dangerous.

The organizations that build the function responsible for knowledge integrity now — before the crisis, before the feedback wave, before the board asks how they let this happen — will have something the others won't.

They will know what's actually happening to them.

In the age of agents coordinating agents, that is no longer a nice-to-have. It is the thing that determines whether the organization is still capable of understanding itself when the wave arrives.

The economists are building the model. The practitioners need to build the practice. If you are already doing this work — without the title, without the mandate, inside an organization that doesn't have language for what you're solving — Unmeasured & Unmeasurable (U&U) is a Slack community for people who can see this coming. [Join here.]*

Incontextable is a publication about organizational knowledge, technology design, and why the gap between the two keeps getting wider.

Sometime in the last thirty years, the enterprise software industry developed a habit of naming its products after the organizational capabilities they were supposed to deliver.

Customer Relationship Management (CRM). Enterprise Resource Planning (ERP). Human Capital Management (HCM). Learning Management System (LMS). Supply Chain Management (SC). Knowledge Management (KM).

Every one of those names describes something that, if an organization actually had it, would change how the business performed. Every one of those names was then attached to a system that delivered something considerably more modest and considerably more expensive.

The names outlasted the ambitions. The acronyms became the territory — and everyone forgot there was ever a map.

Take them one at a time.

Customer Relationship Management. The name suggests an organization that understands its customers — what they actually need, how the relationship is developing, when it is deepening or cooling. A system worthy of the name would know a key account is at risk before the revenue signal appears. It would carry the context of every interaction in a way that made the next person as effective as the last.

What the CRM actually does is record what salespeople are willing to type after meetings they have already had. It captures the transaction, not the relationship. It stores the history, not the understanding.

When that person leaves, the relationship doesn’t transfer. The data transfers. The data is not the relationship.

Enterprise Resource Planning. The name suggests an organization that actually plans — that looks across its resources and makes coherent decisions about allocation before commitments become crises.

What the ERP actually does is track resource consumption after decisions have already been made. It is the most expensive rearview mirror ever built. The ERP knows what happened. It rarely knows why. It has no opinion about whether it was a good idea.

Learning Management System. The name suggests an organization that understands what its people know, what they need to know, and whether the investment in development is producing genuine capability.

What the LMS actually does is track completion. It measures whether the course was clicked through, not whether anything changed in how someone thinks or works. It produces a compliance report that proves learning happened. It has no mechanism for detecting whether it did.

The Learning Management System is how organizations prove that learning is happening without having to check whether it is.

Human Capital Management. The name — setting aside the question of whether people should be called capital, which is worth its own essay — suggests an organization that understands the human capabilities it has, how they are developing, where they are underutilized.

What the HCM actually does is store org charts, process payroll, and generate reports on headcount and attrition. The actual person — with specific knowledge, specific relationships, specific judgment that took years to develop — appears in the system as a job title, a salary band, and a performance rating from one annual cycle. The capital is not managed. It is catalogued.

Knowledge Management. The name promises a system for capturing, preserving, and applying what the organization actually knows — the accumulated understanding that shouldn’t evaporate when people leave.

What knowledge management systems actually manage is documents. Wikis that nobody updates. Intranets that nobody visits. The knowledge — the tacit understanding that lives in practice and judgment — was never in the system. No search function will surface what was never captured. No SharePoint implementation will preserve what lives in the experienced judgment of the person who just retired.

The pattern is the same each time.

Each system was designed before anyone fully understood what delivering the capability would actually require. The technology could automate what was automatable. The parts that actually mattered couldn’t be automated. So the automatable parts became the system and the system became the name.

The measurable thing became the system. The name described the goal. The system captured the proxies. The proxies became the metrics. The metrics became the reality.

The approximation got sold as the thing. The thing was forgotten. The approximation accumulated thirty years of configurations, customizations, and integrations.

And here we are.

The tools that could actually close the gap are arriving now — but the question is what they get pointed at.

Not AI as a feature layered onto the existing system — Salesforce Einstein, SAP Joule, the AI assistant bolted onto the product that was already doing the wrong thing. That is the approximation updated with a chatbot. The gap is still there. The thirty years of accumulated configuration are still underneath.

The genuine opportunity is to ask what the name actually promised before deciding what to build. The CRM that knew a relationship was cooling before the revenue signal appeared. The LMS that knew whether capability changed, not whether a module was completed. The Knowledge Management system that made the why as recoverable as the what.

None of this requires magic. It requires asking what the system was named after and building toward that rather than toward the proxy that was technically feasible in the previous era.

Every organization making AI investment decisions is choosing between two approaches. Add AI to the existing systems — the ones named after aspirations they never delivered, optimized for proxies that were never the real thing. Or ask what the names always promised, and build toward that before deciding what the AI layer sits on top of.

The first approach produces faster approximations. The second produces the thing the name was always promising.

One of these is more expensive to start. One is more expensive in the end.

CHEF is not an aspiration. It is a diagnosis — a name for what something became. That is a different kind of name.

Incontextable is a publication about organizational knowledge, technology design, and why the gap between the two keeps getting wider.

The CISO has a budget, a team, a board presentation slot, and a mandate that has expanded with every headline breach and every new compliance requirement. The risk is taken seriously because the failure mode is visible — it has a date, a cost, a notification requirement, and consequences that are immediate and attributable.

For a brief moment in the late 1990s, some organizations also had a Chief Knowledge Officer.

The CKO was trying to solve a different problem — how to capture, preserve, and leverage what the organization actually knew, so that institutional wisdom didn’t walk out the door with every retirement, so that knowledge accumulated rather than evaporated, so that the organization got smarter over time rather than just busier.

The intellectual foundation for this came largely from Ikujiro Nonaka and Hirotaka Takeuchi’s 1995 book The Knowledge-Creating Company, which argued that Western management theory had been thinking about organizational knowledge in the wrong way. The prevailing assumption was that knowledge was explicit — something that could be captured in documents, procedures, systems, best practices. If you could write it down and store it, you could manage it.

What Nonaka and Takeuchi argued was that the most valuable organizational knowledge is t

acit — embedded in practice, in judgment, in the understanding that comes from doing something repeatedly in a specific context. It lives in people and in the relationships between people. It cannot be fully extracted and stored because extraction is precisely what destroys what makes it valuable. The master craftsman’s knowledge of when the metal is ready is not in any manual. The account manager’s understanding of what a particular client actually needs, as distinct from what they say they need, is not in the CRM.

Their SECI model — moving knowledge between tacit and explicit forms through socialization, externalization, combination, and internalization — was essentially a design framework for the conditions under which tacit knowledge becomes shared organizational understanding. Not capture it. Cultivate the conditions for it.

What happened to their ideas is instructive. The explicit half of the model got productized into knowledge management systems, wikis, intranets, and document repositories. The tacit half — the part that was actually the important half — got ignored because it couldn’t be turned into software. The conditions for cultivating tacit knowledge are social, structural, and cultural. They don’t ship in a box.

The Learning Management System is the clearest example of this failure. The LMS tracks completion, not understanding. It measures whether the course was clicked through, not whether anything changed in how someone thinks or works. It produces a compliance report that proves learning happened. It has no mechanism for detecting whether it did. The organization gets a dashboard showing certification rates and hours logged. The tacit knowledge — the judgment that comes from experience, the pattern recognition that can’t be put in a module — continues to evaporate without appearing on any report.

The LMS is how organizations prove that learning is happening without having to check whether it is.

The CHEF ecosystem — the Complex, Hidden-dependency, Expensive, Fragile environment that most organizational technology stacks have become (explored in full here) — is in large part the residue of applying technology to the explicit layer while the tacit layer evaporated without anyone building a system to notice. The CKO role was supposed to hold that distinction. Without it, nobody did.

Then the dot-com bubble burst, budgets contracted, and the CKO disappeared from org charts while the CISO grew into one of the most powerful roles in the enterprise.

That divergence is not accidental. It is a precise record of which risks organizations know how to take seriously.

The Risk That Has a Date

Cyber security commands the attention it does because its failure mode is legible.

A breach happens at a specific moment. It has a scope — this many records, this many systems, this many customers affected. It has a cost that can be calculated — remediation, legal fees, regulatory fines, reputational damage, customer churn. It has parties who are accountable and parties who demand accountability. Insurers price it. Regulators mandate responses to it. Board members ask about it in terms they understand.

The security stack that organizations have built in response to this risk is genuinely impressive. Zero Trust architecture. Endpoint detection and response. Security information and event management. Multi-factor authentication. Penetration testing. Compliance frameworks. Detailed platform security assessments that evaluate every significant technology decision against data residency, access controls, and encryption standards.

All of it is aimed at the risk that has a date. The risk that, if it materializes, will appear in a headline.

The Risk That Doesn’t

Knowledge integrity fails differently.

There is no breach notification for the moment when an organization’s assumptions about its customers stopped being accurate. There is no incident report for the meeting in which forty slides of data produced no shared understanding and no decision. There is no regulatory requirement to disclose that the institutional knowledge of how to serve a key account walked out the door when the account manager retired and was never captured anywhere the next person could find it.

The failure is real. The cost is enormous — in decisions made on drifted assumptions, in strategies built on models of the market that no longer reflect the market, in customer relationships that degrade because nobody preserved the context of how they were built. But the cost is invisible in the way that hidden dependencies are invisible. It accumulates without a date, without a headline, without a notification requirement, and gets attributed to other causes — market conditions, competitive pressure, talent gaps, execution failures — because there is no forensic process that traces organizational underperformance back to the erosion of knowledge integrity.

So the CISO has a budget and a board slot and a mandate. And nobody has a title or a team or a budget line for the risk that is, in most organizations, considerably more likely to determine whether the business succeeds or fails over the next decade.

What the Security Framing Misses

The platform security assessment your IT team runs before any major technology decision is detailed, serious, and asks exactly the right questions for what it is assessing — compliance certifications, data residency, access controls, encryption standards. It is exactly the right assessment for the risk it is assessing.

But it is assessing the wrong risk.

Any platform that passes the security assessment, regardless of which one you choose, shares the characteristics that actually determine knowledge integrity. Every major platform encodes assumptions about how organizational work should flow. Every one strips context from information to make it portable. Every one creates hidden dependencies between tools, workflows, and data that nobody fully maps. Every one produces outputs that look like knowledge and function as sediment. The security certification provides organizational comfort that the risk has been addressed. The CHEF dynamics get worse regardless of which certified platform you picked.

The security assessment asks: which platform is better at keeping unauthorized people out?

The knowledge integrity question asks: which platform is better at keeping understanding in?

Those are different questions. The second one almost never gets asked, because there is no organizational role with the mandate to ask it.

The Brief Life of the Chief Knowledge Officer

The CKO era deserves more credit than it gets in retrospect.

The people working on knowledge management in the late 1990s were asking the right questions. How does organizational knowledge actually flow? Where does it accumulate and where does it evaporate? How do you design systems that preserve the why alongside the what? How do you measure whether an organization is getting smarter rather than just busier?

Practitioners working at the intersection of lean manufacturing and product development were building the concept of Knowledge Turns — a metric for knowledge health borrowed from inventory turns, designed to measure not how much information the organization had but how efficiently it converted experience into understanding and action. The emphasis was on metrics linked to causal actions, not surface indicators.

The idea faded because the measurement problem was genuinely unsolved. Inventory turns works because inventory is physical and its movement observable. Knowledge isn’t. Documents, training hours, and patents are proxies — they can look healthy while organizational understanding degrades. Without a credible way to measure it, Knowledge Turns stayed conceptually right and practically stranded.

The work was serious and the questions were important. What happened to it is instructive.

The knowledge management discipline got commercialized into products and consulting frameworks. The dot-com contraction eliminated the budget tolerance for anything that couldn’t demonstrate immediate ROI. The CKO role, unable to point to a prevented breach or a compliance certification, couldn’t compete for resources with functions whose risks were legible and whose failures were attributable.

The CISO, meanwhile, got a steady supply of regulatory mandate and catastrophic breach. The Patriot Act and Know Your Customer requirements created serious obligations around identity verification and transaction monitoring that needed real infrastructure and oversight. Then the breach history did the rest — a drumbeat of high-profile compromises that made the case for security investment self-evident to any board and gave the CISO role both budget and urgency that the CKO could never match.

The visible risk won. The invisible risk lost its organizational champion. And in the quarter century since, the gap between how seriously organizations take security and how seriously they take knowledge integrity has grown into one of the most significant and least discussed structural problems in organizational life.

What Knowledge Integrity Would Actually Require

The point is not that organizations should spend less on security. The security risks are real and the investment is largely justified.

The point is that knowledge integrity deserves the same organizational seriousness — a function with a mandate, a budget, a seat at the table, and metrics that actually measure what they claim to measure.

Before getting to what that function looks like, one distinction needs to be made clearly, because the entire backup and recovery industry obscures it.

Saving data is not the same as preserving knowledge integrity.

A perfect backup of every document, every database, every email thread preserves the artifact. It does nothing for the meaning. The backup gives you the what. It never had the why. And the why is what degrades first and restores never. When someone says “we have robust data backup and recovery” as evidence that their knowledge is protected, they are describing a solution to a different problem. The data will survive. The understanding of what it means, why it was created, what problem it was solving, what assumptions were in the room — that was never in the file. No recovery process will restore what was never captured in the first place.

This is Nonaka and Takeuchi’s explicit/tacit distinction applied to backup strategy. Backup systems are extraordinarily good at preserving explicit knowledge. They are structurally incapable of preserving tacit knowledge because tacit knowledge was never in the file to begin with.

So what would genuine knowledge integrity look like organizationally? Two roles, at different levels.

The Knowledge Integrity Officer — call it the KIO — is the organizational peer of the CISO. A function with a mandate to ask the knowledge integrity question before any significant technology decision gets made. Not after implementation, when the hidden dependencies are already baked in and the context is already being stripped. Before. The KIO asks whether a proposed system preserves the reasoning behind decisions or just their outcomes. Whether it makes organizational assumptions visible or buries them in configuration. Whether it creates conditions for genuine understanding or just faster information retrieval. Whether it will make the organization more capable of learning from its own experience five years from now or less.

That role doesn’t exist in most organizations. Nobody has that mandate. Which is why the CHEF environment keeps getting worse with every implementation cycle — each decision locally rational, nobody responsible for the cumulative effect on the organization’s capacity to understand itself.

The Knowledge Steward operates at a different level — within teams and functions rather than across the whole organization. The best analogy is a good board secretary.

A good board secretary doesn’t transcribe everything. They don’t produce a verbatim record that nobody reads. They capture the reasoning behind decisions — the alternatives that were considered and rejected, the context that explains why this choice made sense at this moment, the dissenting view that didn’t prevail but might matter later. Someone reading the minutes a year on can understand not just what was decided but what problem it was solving and what assumptions were in the room.

That’s a skill. It requires judgment about what matters and what doesn’t. It requires enough understanding of the substance to know which disagreement was consequential and which was noise. And it produces something that compounds over time — an organization with good board minutes learns from its own history. An organization without them keeps relearning the same lessons because nobody preserved the reasoning that would have prevented the repetition.

Most organizations apply that discipline at the board level and nowhere else. The decisions that actually shape daily operations — made in project meetings, in vendor negotiations, in the conversations that precede the formal decision — get no equivalent stewardship. They produce action items. The reasoning evaporates.

The Knowledge Steward brings the board secretary’s discipline to the level where most organizational knowledge actually lives and dies. Not documenting everything — that way lies the data lake fallacy applied to meeting notes. Capturing just enough context that the next person can understand not just what was decided but why it made sense, what was traded away, and what would need to change for the decision to be revisited.

This is not an expensive role. It does not require new software. It requires someone with the judgment to know what’s worth preserving and the discipline to preserve it briefly and well. In Nonaka and Takeuchi’s terms, it is the practice that keeps tacit knowledge from evaporating entirely as it moves toward the explicit — the human in the process who holds the context that no system can hold.

The KIO sets the organizational conditions. The Knowledge Steward operates within them. Together they represent the minimum viable answer to the question this piece has been circling: who, exactly, is responsible for what your organization actually knows?

Right now, in most organizations, the honest answer is nobody.

The Question Worth Asking

Your organization almost certainly has a process for evaluating the security posture of every significant technology decision.

It almost certainly does not have an equivalent process for evaluating the knowledge integrity implications of the same decision. Whether the system preserves or destroys context. Whether it creates or obscures hidden dependencies. Whether it makes the organization more or less capable of understanding itself over time.

The platform security assessment will tell you which system is more secure. It will not tell you which platform is less likely to make your organization blind to its own behavior over the next five years.

That question needs someone responsible for asking it.

The CISO role exists because organizations eventually understood that security risk, left unmanaged, compounds into catastrophic failure. The same is true of knowledge integrity risk. The failure mode is slower, the headline less dramatic, the board conversation harder to initiate.

But the organizations that take it seriously before it becomes a crisis will have something the others won’t — the organizational capacity to understand what’s actually happening to them, and to do something about it while there’s still time to choose.

That capacity has a name. It used to have a title.

It’s time to give it one again.

If you are already doing this work — without the title, without the mandate, inside an organization that doesn’t have language for what you’re solving — you are not alone. Unmeasured & Unmeasurable (U&U) is a Slack community for practitioners who can see this coming. [Join here.]

Incontextable is a publication about organizational knowledge, technology design, and why the gap between the two keeps getting wider.

That is not a dramatic claim. It is a description of what most organizational information environments have become — and why the problem is harder to see than it should be.

How It Happened

Nobody designed it this way. It accumulated.

A CRM to manage customer relationships. A financial system to track money. A project management tool. A communication platform. An HR system. A security stack. A data warehouse to connect them. An analytics layer on top of that. AI summaries on top of that.

Each decision made sense at the time. Each addition solved a real problem. The vendor made a compelling case. The competitor was already using it. The demo showed exactly the capability that was missing.

What nobody calculated was the cumulative effect. The complexity that compounds with every new integration. The dependencies that form between systems in ways nobody maps and most people don’t know exist. The total cost that nobody adds up honestly — not just the licenses, but the integration work, the training that repeats every time something changes, the consultants who arrive when the complexity exceeds what the internal team can handle, the workarounds built around systems that almost fit but don’t quite. The fragility that builds as the system gets optimized for normal conditions and loses any capacity to handle abnormal ones.

This is the CHEF model — Complex, Hidden Dependencies, Expensive, Fragile. Not a design philosophy. A diagnosis of what the ecosystem has become.

Nobody Knows How It Works

Leonard Read wrote an essay in 1958 called “I, Pencil” — narrated from the perspective of a pencil describing everything that goes into its own making. No single person on earth knows how to make a pencil from scratch. The knowledge is distributed across thousands of people and processes — logging, mining, refining, manufacturing, shipping — none of whom understand the whole.

The CHEF environment is the same problem applied to organizational information systems. The people who configured the CRM are gone. The consultant who built the integration retired. The vendor who wrote the original software has been acquired twice. The logic behind half the workflows exists only in the memory of people who may or may not still be at the organization.

You cannot redesign what you cannot understand. The CHEF environment, by distributing knowledge of itself across vendors, consultants, and departed employees, makes genuine reform nearly impossible from the inside. The pencil knows more about itself than the organization knows about its own systems.

The Map Is Not the Business

Iain McGilchrist describes a particular failure of the map-making mind — the tendency to mistake the representation of a thing for the thing itself. The map is precise, bounded, manageable. The territory is alive, changing, infinite. The pathology is when the map stops being a tool for navigating the territory and starts being a substitute for it.

CHEF environments do this systematically. The dashboard is not the business. The configured workflow is not the process. The best practice encoded in a system is a description of what worked somewhere, under conditions that no longer fully exist, frozen at the moment someone decided it was worth capturing.

This produces two specific traps.

The first is the End of History Bias. The CRM was configured for the business as it was. The dashboard was designed to measure what mattered then. The assumption baked into every configuration decision is that current conditions are roughly permanent. They aren’t. But nothing in the CHEF environment is designed to notice when they change.

The second is the Best Practices Are Forever Bias. What worked in one context, at one moment, for one organization gets codified as a universal principle. Best practices are always historical. The moment they get embedded in systems and certified against, they start drifting from the reality that produced them. The practice outlives the conditions that justified it. The system keeps measuring compliance with it anyway.

Both biases share the same root — the representation gets treated as more real than the reality it was meant to represent. The system’s model of the business displaces the business itself. David Deutsch calls knowledge growth the beginning of infinity — unbounded, never complete, always provisional. The CHEF environment is the organizational attempt to close that infinity down, to declare the important questions answered and the relevant variables captured. It works until reality produces something the system wasn’t designed to represent. Which it always eventually does.

What It Costs

The most visible cost is financial — the licenses nobody uses, the integrations nobody maintains, the consultants nobody planned for. Organizations that audit their actual SaaS utilization against what they’re paying for consistently find a significant proportion of features that nobody uses but everybody pays for. That number compounds annually.

The less visible cost is what the system does to organizational learning.

In the 1990s, practitioners working at the intersection of lean manufacturing and product development tried to build a Knowledge Turns metric — organizational knowledge health measured the way manufacturing measures inventory turns. Not how much information the organization has, but how efficiently it converts experience into understanding and action.

The idea faded because the measurement problem was genuinely hard. Inventory turns works because inventory is physical and its movement observable. Knowledge isn’t. You can count documents, training hours, patents. None of those are knowledge turning. They’re proxies that can look healthy while organizational understanding degrades.

The CHEF environment produces exactly that — reports full of numbers that tell you something is happening and nothing about why. The causal map, the connection between what the organization does and what results, gets distributed across systems until nobody can trace it. The system is good at producing information. It is poor at producing learning. And because the first is measurable and the second isn’t, the reporting always looks fine.

If You Were Starting Today

Here is a clarifying question worth asking about any established information environment: if you were building this from scratch today, would you build it this way?

Almost certainly not.

Nobody designing an organizational information system from scratch would say — let’s start with thirty years of accumulated vendor decisions, bundle economics, and integration debt, add a security layer on top, and call that our knowledge infrastructure. The CHEF environment isn’t an architecture. It’s a residue. And residue is hard to defend once you say it plainly.

The honest answer to the starting-from-scratch question also reveals what the system is actually optimized for — which is usually vendor retention, license revenue, and the path of least resistance at each decision point. The organization’s needs shaped the early decisions. The vendor’s needs shaped everything after.

The Only Question That Gets at the Root

You cannot fix a CHEF environment by adding more tools to manage the complexity, more integrations to bridge the hidden dependencies, more reporting to make the expense visible. That is the logic that built it.

The upstream question is the only one that matters: what does this organization actually need to understand, and what is the minimum system needed to support that understanding?

It is also the question that almost no technology evaluation process ever asks — because it requires honest assessment of what the organization can actually do, not what the vendor’s demo suggested was possible. And because asking it honestly usually reveals that a significant proportion of what the current environment produces is not connected to any genuine organizational question. Data captured because it can be. Reports generated because the tool generates them. Meetings scheduled because the calendar made it easy.

Organizations built systems to capture what they know. The systems got good at that. What they didn’t build — and what the systems actively work against — is the capacity to know something new.

That capacity doesn’t come from a better system. It comes from asking better questions of the systems you already have. Starting with: if we were building this today, what would we leave out?

Every productivity platform, every workflow automation, every scheduling tool, every AI assistant is built on the same premise: friction is waste. Remove the steps, reduce the effort, eliminate the awkwardness, smooth the path.

It’s a reasonable assumption in a lot of contexts. But it’s destroying something that most organizations don’t have a name for. Because friction isn’t always waste. Sometimes friction is a filter. And when you optimize away a filter, you don’t just save time. You lose the selection mechanism the friction was providing — and you don’t notice, because the loss is invisible and the efficiency gain is measurable.

————————

What Friction Actually Does

Consider the Calendly link.

Calendly is a scheduling tool that eliminates the back-and-forth of arranging a meeting. You send someone a link, they pick a slot, it appears in both calendars. No negotiation, no awkward emails, no “does Tuesday work for you.” Clean, fast, frictionless.

It’s also, in a precise sense, indiscriminate.

The old process of arranging a meeting — the email, the phone call, the ask — contained information. The effort required to initiate a meeting was roughly proportional to the perceived value of having it. Which meant that meetings which weren’t quite worth the effort of arranging often didn’t get arranged. The friction filtered them out before they consumed anyone’s time.

Calendly removes that filter. The CEO and the cold outreach vendor get the same frictionless access to your calendar. The meeting that would have died in the scheduling process — because actually, when forced to articulate why it needed to happen, neither party could quite do it — now happens, because the path of least resistance made it easy.

This is not a criticism of Calendly specifically. It’s an observation about what happens when you remove friction from a place where friction was doing useful work. You don’t just save time in the scheduling process. You transfer the cost — now paid in the meeting itself, and in every subsequent meeting that results from it, and in the relationships that form around conversations that probably shouldn’t have started.

The efficiency gain is real and measurable. The filter loss is real and invisible.

————————

The Productivity Industry’s Blind Spot

The entire productivity optimization industry is built around a single question: how do we do this faster and with less effort?

It is almost never built around the prior question: should we be doing this at all?

Friction, when it exists in the right places, answers that prior question automatically. The cover letter that requires genuine effort filters for candidates willing to do the work before the interview begins. The budget approval process that requires written justification filters for projects whose proponents believe in them enough to articulate why. The sales process that requires multiple conversations filters for customers who actually need what you’re selling rather than customers who were frictionlessly onboarded and will churn in three months.

In each case, the friction is not an obstacle to value. It is the mechanism by which value gets selected for.

When you automate it away, you don’t eliminate the selection problem. You just move it downstream, where it’s more expensive and more disruptive to resolve.

————————

Toyota Knew This

The most counterintuitive quality control insight in the history of manufacturing was the andon cord.

In a Toyota plant, any worker on the production line could pull a cord to stop the entire line if they spotted a problem. This was deliberate, designed friction introduced into a system that every other manufacturer was trying to make as frictionless as possible. The assumption everywhere else was that stopping the line was waste. Toyota’s assumption was that stopping the line was the mechanism by which quality was preserved.

The friction was the product.

What the andon cord created was a system in which problems surfaced rather than got optimized around. In a frictionless production system, a defect gets passed down the line, incorporated into subsequent steps, and discovered — expensively — at the end. In Toyota’s system, the friction of stopping everything forced the problem to be addressed at the point where it was cheapest to fix.

The insight transfers directly to information and organizational decision making. The meeting that requires a written agenda before it gets scheduled is an andon cord. The approval process that requires someone to articulate the rationale for a decision is an andon cord. The pause before deploying an AI output that requires a human to verify the reasoning is an andon cord.

These are not inefficiencies. They are the mechanisms by which organizational quality gets preserved. And the productivity optimization impulse, applied without judgment, pulls them out one by one.

————————

The Mimicry Problem

Friction also gets eliminated for reasons that have nothing to do with efficiency — through mimicry.

A tool spreads through a professional community not because it solves a problem everyone has but because enough visible people use it that using it becomes a signal of belonging. Calendly is partly this. The bundle of SaaS features nobody uses but everyone pays for is entirely this. The AI summary appended to every document regardless of whether the document needs summarizing is this.

When a tool spreads through mimicry rather than genuine need, it carries whatever frictions it eliminates along with it — into contexts where those frictions were actually serving a purpose. The professional who needed Calendly because they were genuinely managing hundreds of inbound meeting requests is a real person. The professional who got a Calendly link because everyone in their network had one, and now sends it to everyone including people they’re trying to build a relationship with for the first time, has imported a solution to a problem they didn’t have — and exported a filter that was worth keeping.

The question is not whether to use scheduling tools. It’s whether the friction they eliminate was doing useful work in your specific context — or whether you adopted the tool because your peers did, without asking.

————————

Where Friction Belongs

None of this is an argument for making things unnecessarily hard. Friction in the wrong places is genuinely wasteful — the approval process that exists because nobody removed it when the organization changed, the report that gets generated because it always has been, the meeting that happens weekly because it was scheduled to recur.

The discipline is knowing the difference. And that requires asking a question that almost no technology evaluation process ever asks:

What is this friction currently filtering, and do we want to keep filtering for that?

If the friction is filtering for nothing — it’s just inertia, bureaucracy, legacy process — eliminate it. That’s genuine efficiency.

But if the friction is filtering for effort, for seriousness, for genuine need, for the kind of commitment that separates the conversations worth having from the ones that would consume time without creating value — then optimizing it away is not efficiency. It is the removal of a selection mechanism you will miss when it’s gone and won’t be able to name when you’re trying to understand why things feel faster but somehow worse.

————————

The Design Question

This is ultimately a design question, and it belongs upstream of any technology decision.

Design for Manufacturability — the discipline that preceded lean manufacturing — taught engineers that the decisions made earliest in a product’s development had the most leverage over everything downstream. A design that ignored manufacturability created problems that no amount of effort on the factory floor could fully resolve. The leverage was always in the early decisions.

The same principle applies to organizational information design. The decision about where to preserve friction and where to eliminate it is an early, upstream, high-leverage decision. Made well, it shapes what kinds of conversations happen, what kinds of decisions get made carefully, what kinds of relationships form with sufficient investment to be real. Made badly — or not made at all, because the default is always to eliminate friction wherever it appears — it produces organizations that are very fast and very smooth and gradually, invisibly, unable to filter for what matters.

That question requires judgment. Judgment requires pausing. Pausing is, in a small way, friction.

Which is exactly the point.

Interfaces are redesigned to remove clicks.

AI tools are built to remove pauses.

Workflows are optimized to move from prompt to output as quickly as possible.

Speed feels like progress.

But there is a strategic tradeoff embedded in that assumption.

What Friction Used to Do

Before automation smoothed everything out, many forms of friction were unavoidable.

• Drafting a memo required thinking through the argument.

• Building a spreadsheet required deciding what variables mattered.

• Preparing a presentation required choosing what to leave out.

Some of that friction was waste. Some of it was interpretation.

When the process required effort, interpretation was harder to skip.

When the process becomes frictionless, interpretation becomes optional.

That distinction matters.

The Default Problem

Frictionless systems are efficient at one thing: accepting defaults.

If an AI can summarize a meeting instantly, most teams will use the summary as written.

If a model can draft a strategy in seconds, its framing becomes the starting point.

If a dashboard updates automatically, its categories define what gets attention.

None of this is irrational. It’s efficient.

But when defaults become frictionless, they also become invisible. And invisible defaults tend to compound.

Where This Connects to Convergence

In a world where everyone uses similar tools, frictionlessness has a side effect.

It reduces the moments where organizations impose their own logic.

Shared infrastructure already compresses variance.

Frictionless workflows accelerate that compression.

Not because anyone intends it. Because ease makes adoption automatic.

The faster ideas move from model to action, the fewer opportunities exist to question whether the model’s framing fits your context.

As speed increases, differentiation quietly narrows.

Friction as Signal

This is not an argument against efficiency.

It is an argument for recognizing where efficiency has replaced judgment.

Some decisions benefit from being fast Others benefit from resistance. Resistance forces articulation.

It forces tradeoffs.

It forces ownership.

The absence of friction is not neutral.

It signals that no additional interpretation was required — or that none occurred.

The Practical Question

Instead of asking, “How do we remove friction?” You might want to ask:

Where in our organization should friction never fully disappear?

• Pricing decisions?

• Risk acceptance?

• Hiring?

• Escalations?

• Strategic positioning?

These are not workflow problems. They are judgment problems.

When judgment becomes frictionless, it becomes standardized. And standardized judgment is difficult to differentiate.

The Cost

Frictionless systems feel productive.

• They move quickly.

• They generate output.

• They reduce visible effort.

But if every important decision flows with the same smoothness as a routine task, something has shifted.

Not necessarily toward error.

But toward sameness.

In a shared technological environment, the path of least resistance often leads to the industry average.

And the industry average is rarely where leverage lives.

The difference between a utility and a competitive advantage.

In my last post, I discussed Strategic Convergence: the process by which organizations, by adopting the same algorithmic models and “best practices,” gradually eliminate the differences that make them competitive.

There is a structural driver behind this trend.

I call it the Renting Trap.

Most organizations today are renting their intelligence. They integrate generalized AI models into their operations, capture a baseline efficiency gain, and mistake that for strategy.

But if you are using the same models as your competitors in the same way, you are not building a defensible position.

You are adopting infrastructure.

And infrastructure does not differentiate.

The Limits of Generalized Models

When your strategy relies entirely on centralized, third-party models, you accept three structural realities:

Performance Parity

If the model’s best output is available to every competitor at the same price, that output becomes the industry floor. It creates competence, not advantage.

The Information Gap

Generalized models are trained on public data and broad patterns. They do not understand your internal tradeoffs, historical failures, informal decision logic, or the subtle constraints that shape how your organization actually works.

Operational Dependency

If core workflows depend on infrastructure you do not control, your “strategy” shifts whenever your provider changes pricing, policies, or model behavior.

None of these are technical flaws. They are economic facts.

The Unseen Risk: When Rented Intelligence Feels Like Yours

Earlier waves of SaaS felt rented.

Cloud felt rented. Platforms felt rented. AI is different.

Microsoft Copilot for 365 is the cleanest example of how subtle the Renting Trap has become.

• .

It feels like your intelligence. But it isn’t.

The reasoning layer belongs to Microsoft.

The behavioral defaults were written by Microsoft.

The model updates happen on Microsoft’s timeline.

The training data reflects the global average, not your specific history.

You can configure access. You can adjust permissions. You can prompt it. You cannot redefine how it fundamentally reasons.

And because it operates inside familiar, trusted software, automation bias kicks in.

• The output looks authoritative.

• It sounds reasonable.

• It appears inside your workflow.

• So we accept it faster.

• We interrogate it less.

Over time, something subtle happens. The model’s framing becomes the organization’s framing. The model’s omissions become institutional blind spots.

The model’s tone becomes cultural tone.

The intelligence feels owned because it operates inside your system.

Structurally, it is shared infrastructure. That is the unseen risk.

When rented capability feels personal, convergence accelerates m.

Capability Is a Utility. Direction Is Not.

Electricity is essential. It does not make you unique.

AI is becoming similar: necessary for participation, insufficient for advantage.

If every competitor can access the same model, the same API, and the same outputs, then relative advantage depends on a different layer entirely:

• Your context.

• Your constraints.

• Your tradeoffs.

• Your judgment.

• Not your subscription.

Owning intelligence does not mean training your own frontier model. It means owning direction.

When capability is shared, differentiation depends on what you decide to automate and what you refuse to automate.

It depends on what assumptions you encode and which ones you challenge. It depends on how deliberately you apply the tool.

Rented intelligence can execute. It cannot decide what is worth executing.

That distinction is strategic.

The Diagnostic

Ask yourself:

If your AI provider materially changed tomorrow, would your competitive position remain intact?

If your differentiation disappears when the API changes, it was never your differentiation.

You weren’t building strategy. You were configuring a utility.

The Shift

The Algorithmic Age is not a race to adopt tools faster than your competitors.

It is a discipline of maintaining differentiated direction in a world of shared capability.

The Renting Trap is not about technology. It is about confusing access with advantage.

When intelligence becomes abundant and cheap, the scarce resource is judgment. And judgment cannot be rented.

There is a particular kind of confidence that comes from having access to a lot of processed information. It feels like knowledge. It presents itself as knowledge. But there is a distinction that most organizations are no longer equipped to make — between knowledge and the residue of knowledge. Between understanding and its fossil record.

This distinction is becoming one of the most consequential strategic questions of our time, and almost nobody is asking it seriously.

What Algorithms Actually Do

Start with what an algorithm is, stripped of mythology.

An algorithm is a set of rules applied to existing data to produce an output. It is, in a deep sense, always looking backward. It has no access to the present moment except through proxies — inputs that were themselves captured at some prior moment, cleaned, structured, and fed into a system designed to find patterns in what has already happened.

This is not a flaw. It is the design. And for many purposes — logistics, fraud detection, recommendation, pricing — it works extraordinarily well precisely because the past is a good predictor of the near future within a stable system.

The problem is that strategy is not a stable system. Strategy is the attempt to act well in conditions that are changing, contested, and not yet fully legible. And for that purpose, the backward-looking quality of algorithms is not a minor limitation. It is a structural disqualification.

David Deutsch, the physicist and philosopher, makes an argument that cuts deeper than most AI criticism manages to reach. Knowledge, he argues, is not information. Information can be stored, copied, retrieved. Knowledge is something different — it is an explanation that actually reaches beyond the data it was built on. Real knowledge makes predictions that could be wrong. It is always a conjecture, always provisional, always capable of being refuted by the world. That is not a weakness of knowledge. It is what makes it knowledge rather than mere pattern.

Algorithms do not conjecture. They interpolate. They find the most probable path through a space defined by what they have already seen. They cannot step outside that space, because stepping outside requires proposing an explanation that the data does not yet support — which is precisely what human thinking, at its best, is able to do.

The Quantum Problem of Information

There is a further problem that goes largely unremarked, and it comes from the same direction — from physics, from the nature of information itself.

Information is not a stable object. It is a moment. Every insight, every piece of organizational knowledge, every customer understanding was generated at a particular time, by particular people, asking particular questions, in a particular context. That context is not separable from the knowledge. It is part of the knowledge. The moment you strip the context to make the information portable — to put it in a database, a model, a report — you have already begun to lose it.

This means that the more an organization relies on algorithmically mediated knowledge, the more it is operating on a kind of intellectual fossil record. The data represents the world as it was understood at the time of capture. The model represents the patterns that existed in that world. But the world has moved. The context has shifted. And the system has no way of knowing this, because it has no access to the present except through new data that will itself arrive already slightly out of date.

Nassim Taleb would recognize this as a hidden fragility. The system appears to be functioning. The outputs are coherent. The confidence intervals look reasonable. But the system is quietly accumulating a gap between its model of the world and the world itself — a gap that is invisible until it isn’t.

The organization that runs on algorithmic mediation is not just converging with its competitors. It is also, slowly, losing its grip on the present.

The Organizational Brain and Its Asymmetry

Iain McGilchrist’s work on the divided brain is usually received as a contribution to neuroscience or philosophy of mind. But it has a remarkably precise application to how organizations think and fail.

McGilchrist’s argument, condensed roughly, is this: the left hemisphere of the brain is extraordinarily good at working within known systems. It categorizes, sequences, optimizes, and executes. It takes the world as it already understands it and works efficiently within that understanding. The right hemisphere, by contrast, holds the capacity for context — for grasping the whole, for sensing that something doesn’t fit, for the kind of open attention that allows genuinely new understanding to arrive.

The danger McGilchrist identifies is not that the left hemisphere is wrong. It is that the left hemisphere, given sufficient institutional power, tends to mistake its map for the territory. It optimizes the representation and loses the thing being represented.

Algorithms are left-hemisphere artifacts at organizational scale. They are extraordinarily powerful within a defined problem space. But they do not hold context. They do not sense that something doesn’t fit. They cannot tell you that the question you are asking is the wrong question. That capacity — to step back, to feel the inadequacy of the current frame, to propose a different one — requires exactly the kind of open, contextual, conjectural thinking that algorithmic systems structurally cannot provide.

When organizations progressively route their decisions through algorithmic mediation, they are not just automating tasks. They are, gradually, choosing left hemisphere over right at an institutional level. They are building organizations that are very good at executing within known parameters and increasingly unable to question those parameters.

The Surrender That Doesn’t Feel Like Surrender

There is an obvious objection to everything argued above: surely leaders know what they are trading away. Surely the decision to route judgment through a system is a conscious one, made with clear eyes about the tradeoffs.

It isn’t. And understanding why it isn’t is central to understanding the problem.

No organization decides to surrender its interpretive capacity. What organizations decide, repeatedly and locally, is to adopt tools that are faster, cheaper, more consistent, and easier to audit than the human judgment they replace. Each of those decisions is rational. Each one produces a visible gain. The loss — the quiet erosion of the organizational capacity to question its own frame — is invisible at the moment of adoption and accumulates only in aggregate, over time, across many such decisions.

This is precisely the structure of a Talebic hidden fragility. The system does not look like it is becoming more brittle. It looks like it is becoming more capable. Outputs are faster. Reporting is cleaner. Decisions are more defensible. The competence being lost — the ability to notice that the question is wrong, to conjecture beyond the available data, to interpret rather than retrieve — leaves no gap in the reporting. There is no metric for what you can no longer ask.

By the time the frame breaks — when the environment shifts in a way the model did not anticipate, when the customer need moves outside the categories the system was built to serve — the organization discovers that it has optimized away the capacity it most needs. Not through a single bad decision. Through a thousand good ones.

How Cloud Software Homogenized Failure

The strategic convergence story usually focuses on strategy — positioning, differentiation, competitive moves. But something more pervasive happened when organizations moved their core operations to standardized cloud platforms, and it has received almost no serious attention.

The promise of cloud software was democratization. Small companies could now access the same CRM, the same ERP, the same HR systems, the same ticketing infrastructure that previously only large enterprises could afford to build. That promise was largely kept. Capability spread. Implementation costs fell. Integration became easier.

But the promise came with a hidden symmetry. When every organization runs the same Salesforce instance, the same Workday configuration, the same ServiceNow workflows, they do not just gain the same capabilities. They inherit the same assumptions. The same categories. The same definitions of what a customer is, what a ticket is, what a performance review is, what an exception looks like.

Software encodes a theory of how work should flow. That theory was built for the median case — the most common version of the problem the software was designed to solve. For most situations, most of the time, it is good enough. But organizations are not only their median cases. Their character, their distinctiveness, their actual competitive advantage often lives in how they handle the situations that don’t fit the workflow. The edge cases. The exceptions. The moments where someone has to decide what to do because the system doesn’t know.

What standardized cloud software did, quietly and at scale, was make organizations incompetent in identical ways. The customer whose problem doesn’t map to the ticket categories gets the same runaround everywhere. The employee situation that doesn’t fit the Workday workflow hits the same wall at every company. The strategic decision that doesn’t appear in the ERP’s reporting structure goes unmeasured across entire industries.

Organizations did not just converge on their strategies. They converged on their dysfunctions. They became, in a precise sense, different instances of the same system — running the same logic, failing at the same edges, with no particular organization better positioned than another to recognize it, because the tool that would tell them something is wrong is the same tool that is wrong.

Convergence Is the Symptom, Not the Disease

Gary Hamel warned thirty years ago about strategic convergence — the tendency of industries to homogenize around best practices until strategy becomes imitation with slight variation. His warning was apt. SaaS accelerated it. Cloud standardized it. Now algorithms industrialize it at a speed and scale Hamel could not have anticipated.

But convergence, as serious as it is, is the symptom. The disease is the loss of the capacity to diverge.

Convergence can in principle be reversed by a sufficiently original strategic insight — a new way of seeing the competitive landscape, a genuinely different theory of what customers need. What makes the algorithmic moment different from previous waves of standardization is that the tools organizations are now deploying actively suppress the conditions under which such insights arise.

Original insight requires slack — space in the system for thinking that does not immediately justify itself in output. It requires dissent — people willing to say that what the numbers show is not what is actually happening. It requires conjecture — the willingness to propose an explanation that goes beyond current evidence. And it requires a culture in which the question “what assumptions are encoded in this system?” is not only permitted but actively valued.

Algorithmic infrastructure, over time, tends to crowd out all of these. Not through malice. Through convenience. It is simply faster and less uncomfortable to run the model than to question it.

The accountability that diffuses when decisions become “what the system recommended” is not only an ethical problem. It is an epistemological one. When no one owns the interpretation, no one is positioned to notice that the interpretation is wrong.

What Deliberate Differentiation Requires

The response to all of this is not to reject algorithmic tools. That is neither possible nor desirable. The response is to govern them — to design organizations that use these tools without becoming captured by them.

This requires, first, an honest audit of where algorithmic mediation has quietly replaced judgment. Not where automation has replaced labor — that is mostly fine — but where the interpretive function itself has been outsourced. Where the question of what does this mean has been handed to a system that can only answer what pattern does this match.

It requires, second, the deliberate preservation of what might be called interpretive capacity — the organizational ability to question encoded assumptions, to notice what the model cannot see, to maintain human agency at the points where genuine understanding matters. This is not a call for inefficiency. It is a call for knowing which kind of thinking a given decision actually requires.

And it requires, third, a recognition that knowledge is not a stock that accumulates. It is a practice that must be continuously exercised. The organization that stops conjecturing, stops criticizing, stops generating genuine explanations — and instead retrieves and recombines — is not becoming more intelligent. It is becoming more brittle in ways that will not be visible until the environment shifts in a way the model did not anticipate.

The scarcest resource in the algorithmic age is not data, not processing power, not even talent in the conventional sense. It is the institutional capacity to ask whether the frame is right — and to mean it.

The Question Nobody Is Asking

Most of the conversation about AI in organizations is about adoption, capability, and risk in the narrow sense — bias, hallucination, job displacement. These are real concerns. But they are not the deepest concern.

The deepest concern is this: as organizations defer progressively to algorithmic systems, who remains positioned to question them? Not to audit them for technical errors, but to ask whether the questions they are answering are the right questions. Whether the world they model is still the world that matters. Whether the optimization they are performing is in service of the right objectives.

This is not a technical question. It is a strategic and philosophical one. And it requires exactly the kind of knowledge — contextual, conjectural, moment-bound, human — that the algorithmic age is quietly making scarce.

The organizations that recognize this early will not necessarily resist automation. But they will treat the preservation of genuine human understanding as a strategic asset rather than an inefficiency to be engineered away. They will design around their tools rather than into them.

That is the counter-move. Not resistance. Architecture.

• Not just the Information Age.

• Not simply the AI Age.

We are entering the Algorithmic Age.

And in this age, strategic differentiation is quietly becoming harder. Not because companies lack tools. Because they are using the same ones.

This Is Not a New Warning

In the 1990s, Gary Hamel warned of strategic convergence.

As industries adopted similar best practices, benchmarking intensified, and consulting frameworks spread, companies began to resemble one another.

Strategy became imitation with slight variation.

For decades, that warning proved accurate.

SaaS accelerated it.

Cloud standardized it.

“Best practice” globalized it.

Now algorithms industrialize it.

What Is the Algorithmic Age?

The Algorithmic Age is defined by mediation.

Decisions are increasingly:

• Scored

• Routed

• Filtered

• Recommended

• Automated

• Predicted

• Summarized

By systems.

The algorithm is no longer a tool. It is infrastructure.

• Recommendation engines shape attention.

• Dashboards define performance.

• SaaS platforms encode operating models.

• AI copilots generate language.

• Workflow systems route decisions.

• Agents begin acting autonomously.

• Algorithms compress complexity into output.

They optimize within parameters. They do not question the parameters.

The Expansion of Convergence

What Hamel saw in strategy has now spread across the entire organization.

Strategic differentiation is no longer just about positioning.

It now applies to:

• Design — how products are conceived.

• Operations — how work flows.

• Marketing — how messaging is framed.

• Sales — how conversations are structured.

• Procurement — how vendors are selected.

• Management — how decisions are routed.

• Finance — how performance is measured and capital is allocated.

When every function runs on similar systems, similar KPIs, and similar models, convergence becomes systemic.

Not tactical. Structural.

The Illusion of Intelligence

The Algorithmic Age produces extraordinary information density.

Dashboards glow. Feeds update. Summaries appear instantly. Insights are auto-generated.

It feels intelligent.

But more information does not automatically produce more insight.

Algorithms optimize averages.

• Strategy often requires deviation from averages.

Algorithms reinforce pattern.

• Strategy sometimes requires breaking pattern.

When leadership over-relies on algorithmic mediation, optimization replaces design.

Adoption replaces interpretation.

Convenience replaces deliberation.

Depersonalization and Diffused Responsibility

There is another shift beneath the surface.

As decisions become algorithmically mediated:

• Ownership softens.

• Accountability diffuses.

• Judgment feels outsourced.

• Action feels pre-validated.

• “It’s what the system recommended.”

• “It’s what the dashboard showed.”

• “It’s what the model predicted.”

This does not eliminate responsibility.

It merely obscures it.

That is strategically dangerous.

The Real Scarcity

In the Algorithmic Age, automation is abundant.

Optimization is abundant.

Data is abundant.

The scarce resource is deliberate differentiation.

Not differentiation as branding.

Differentiation as design choice.

The organizations that stand out will not be those that automate fastest.

They will be those that:

• Constrain automation intentionally.

• Question encoded assumptions.

• Design non-default operating models.

• Preserve interpretive capacity.

• Maintain human agency at strategic nodes.

• Introduce friction where thinking matters.

Strategic differentiation must now be practiced across the entire enterprise.

Not just in the strategy deck.

The Counter-Move

The response is not rejecting algorithms.

It is designing around them.

Leaders must ask:

• Where are we converging unintentionally?

• What assumptions are encoded in our systems?

• Where has optimization replaced thinking?

• Where must interpretation remain human?

• What tradeoffs are we accepting without naming them?

• Which functions are drifting toward sameness?

This is not resistance.

It is governance.

The Storm Forming

The Algorithmic Age is still forming.

Like a tropical system over warm water, its structure is visible before its impact is fully felt.

Signals are everywhere:

• AI-generated communication flattening voice.

• Standard SaaS stacks shaping similar workflows.

• KPI dashboards driving similar behavior.

• Automation bias influencing confidence.

• Agent systems reducing interpretive pauses.

The water is warm. Convergence increases.

Differentiation becomes harder. And more valuable.

The Thesis

Gary Hamel warned of strategic convergence three decades ago.

The Algorithmic Age makes convergence scalable.

Algorithms optimize for sameness. Strategy requires difference.

To achieve strategic differentiation in the Algorithmic Age, you need think carefully about what really matters to real humans and not just the algorithm across every function of

the organization.

It is not about resisting the algorithm, but designing beyond it.

In the 1980s, GM Chairman Roger Smith had a vision. He was going to leapfrog the Japanese by “automating” the competition into oblivion. He spent roughly $90 billion—more than the market cap of Toyota and Honda combined—on high-tech robotics and the acquisition of EDS (Electronic Data Systems) and Hughes Aircraft.

​He didn’t just want robots; he wanted to buy a “technological brain” for GM. It was the most expensive “Placebo Integration” in history.

We are witnessing the rise of what William Langewiesche calls the 'orphans of automation.' From the cockpit to the assembly line, the human element is being sidelined by systems we no longer fully command. As Sandy Munro often reminds engineers, the goal should be to 'simplify, then automate'—because an automated system built on a flawed human process is destined for failure​.

Today, we are building a new generation of orphans. We call them AI.

​The “Lulu” of a Deal

​Roger Smith famously called the purchase of Ross Perot’s EDS a “lulu of a deal.” He believed that grafting a world-class software firm onto a legacy car manufacturer would magically unify GM’s fragmented data.

​But instead of integration, he got an organ rejection. The “Information Systems” guys at EDS spoke a different language than the “Manufacturing” guys on the floor. Because the Information Design hadn’t been fixed, the technology was just an expensive layer of paint over a crumbling structure.

​The robots were “Orphans” because they had no Context. A robot arm would perfectly weld a door frame, even if the frame was misaligned by two inches. It performed its “Function” flawlessly while destroying the “Value Stream.”

​The 2020s Glitch: The AI Orphan

​We are currently repeating this at the Nth Power.

Companies are dropping “AI Orphans” into their departments:

• ​The Finance AI that can generate a report in seconds but has no context of why the customer is churn-risk.

• ​The Support AI that summarizes a call perfectly but can’t see the engineering defect that caused the problem in the first place.

​Just like Roger Smith’s robots, these AI tools are isolated islands of efficiency. We are asking an AI to “optimize” a silo, forgetting that the silo itself is the design flaw.

​Information Design is the Parent

​An “Orphan” is what happens when you have Tooling without Architecture. If GM had used the logic of Lean or Theory of Constraints correctly, they would have redesigned the flow of the factory before buying the robots. They would have ensured the “Information” (the measurements, the quality cues) was integrated into the system’s DNA.

​In the modern corporation, the only way to prevent “AI Orphans” is to treat Information as a Design Problem. * Stop asking: “How can AI do this task?”

• ​Start asking: “How do we redesign the work so the AI has the context it needs to actually be useful?”

​The In-Context-able Conclusion

​Roger Smith proved that you can’t buy your way out of a bad design. You can’t automate incoherence. Whether it’s a robot arm in 1985 or a Large Language Model in 2026, if the information is siloed, the tool is an orphan.

​We don’t need faster orphans. We need a better architecture.

If your organization has more data than ever but less sense than ever, you aren't imagining it.

The problem isn't information. Organizations have more of it than they can process. The problem is that almost none of it completes a Knowledge Turn.

A Knowledge Turn is what happens when understanding actually moves — from the person who has it to the person who needs it, at the right time, in a form that changes what they know and therefore what they decide. It is the difference between information stored and knowledge used. Most organizations have the first and almost none of the second.

Every tool in the modern enterprise was designed around storing and retrieving information. None of them were designed around whether understanding actually moves. The filing cabinet was the original knowledge management system. SharePoint is a more expensive filing cabinet. The AI layer being sold right now is a faster filing cabinet with better search.

The knowledge flow problem stays invisible because it doesn't show up on any dashboard. The metrics measure what was stored, what was sent, how fast it arrived. Not whether anyone understood anything differently as a result.

That is what this publication is about.

What You'll Find Here

Each issue examines one aspect of the knowledge flow problem — where it breaks down, what organizations have done to redesign for it, and what a different set of questions would produce.

A recurring feature — K-TIP: Knowledge Turns in Practice — looks at one specific organization and one specific practice that redesigned how knowledge flows rather than how information is stored.

The argument runs across every industry and every kind of organization. The tool vendors won't make it because it questions the premise of what they're selling. That's reason enough to make it here.

Years ago Joel Orr told me something I didn’t

fully understand at the time:

Structure is destiny.

Orr was an early CAD pioneer and systems thinker who spent decades watching organizations adopt powerful new tools and struggle with the same old structural problems. He wasn’t being philosophical. He was being literal.

Structure decides outcomes.

And most organizations aren’t designing their structure anymore.

They’re inheriting it.

Inherited structure produces inherited destiny.

That’s the quiet danger of the platform era: not bad tools, not weak technology — but the rise of default thinking.

And default thinking isn’t lazy thinking.

It’s the absence of thinking.

The default is running the organization

We now operate inside a world of ready-made systems:

• default software

• default workflows

• default templates

• default dashboards

• default best practices

Each one removes friction. Each one promises speed. Each one says: you don’t have to decide this.

That’s useful — until it isn’t.

The moment the default becomes a substitute for judgment, design disappears. Work keeps moving, but no one is shaping it. The organization isn’t choosing how it behaves. The system is.

From that moment on, destiny is inherited, not intentional.

And inherited destiny always converges toward sameness.

Platforms don’t create outcomes

Substack proves this brutally.

Every writer gets the same infrastructure:

• same publishing system

• same delivery rails

• same monetization tools

• Yet outcomes spread from invisible newsletters to cultural institutions.

The platform didn’t pick winners.

It exposed structure.

The successful publications have a designed container:

• a clear promise

• a recognizable lens

• consistent framing

• intentional sequencing

• a sense of direction

That’s not tooling. That’s structure expressing itself.

An unknown publication isn’t hidden because Substack failed. It’s still searching for the structure that gives the work gravity.

Platforms amplify design.

They don’t replace it.

Infrastructure isn’t strategy

Operational platforms tell the same story:

• Microsoft 365

• Google Workspace

• Salesforce

• HubSpot

• ERP stacks

Their purpose is stability, not uniqueness. They industrialize the commodity layer so you don’t waste energy reinventing plumbing.

That’s great up to a point.

The mistake is expecting advantage to emerge from infrastructure.

• Which CRM will save us?

• Which integration will unlock productivity?

• Which dashboard will create clarity?

Those are design questions disguised as purchases.

Infrastructure can support design. It cannot supply it. Structure is still a decision.

Where to standardize, where to design

Dissing the default doesn’t mean rejecting platforms.

It means refusing to let them think for you.

Standardize aggressively where uniqueness has no strategic value:

• infrastructure

• commodity workflows

• coordination hygiene

Design intentionally where advantage lives:

• decision-making

• priorities

• interpretation

• customer experience

• organizational behavior

Confuse those layers and you scale mediocrity. Separate them and you scale intelligence.

The platform era didn’t remove the need for design. It removed the excuses for avoiding it.

You can inherit structure. Or you can design it. One path produces intentional destiny. The other produces whatever the defaults happen to allow.

Defaults are not neutral. They are decisions you didn’t make.

Most organizations think their biggest risks are visible.

• Cybersecurity threats.

• System outages.

• Bad actors.

• Regulatory failure.

• AI gone rogue.

Those are real risks. But they’re not the most dangerous one.

The most dangerous risk is quieter — and far more normalized.

It’s the moment when everyone knows the information is partially wrong, incomplete, or out of context… and everyone keeps building on it anyway.

That’s the Black Swan hiding in plain sight. The risk isn’t ignorance. It’s shared denial.

In almost every modern organization, people can feel it:

• Dashboards don’t quite line up

• Reports lag reality

• Metrics mean different things to different teams

• AI summaries sound confident but feel hollow

• Assumptions are old, but still operational

Context has leaked out of the system. None of this is secret.

Everyone knows. But no one wants to be the person who says:

• “I don’t think this is actually true.”

• “We’re stacking decisions on shaky assumptions.”

• “This only works if nothing important changes.”

• “We’ve confused visibility with understanding.”

So people go along to get along. Not because they agree — but because questioning the system feels riskier than trusting it.

That’s not alignment. That’s epistemic appeasement.

How Black Swans are quietly manufactured

Black Swans aren’t just random events.

They are often the result of small inconsistencies that were tolerated for too long.

Here’s the pattern:

• Minor data inconsistencies are ignored

• Local workarounds become normal

• Confidence remains high because outputs look clean

• Automation and AI amplify existing assumptions

• Reality eventually pushes back — hard

When that happens, the postmortem always sounds the same:

“No one could have seen this coming.”

But that’s rarely true. Plenty of people saw it coming.

They just didn’t have a safe way to surface doubt — or a shared way to resolve it.

The danger of “it’s good enough”

Modern information systems are incredibly good at producing plausible coherence.

Dashboards reconcile numbers.

• Reports smooth contradictions.

• AI fills in gaps.

• Executive summaries compress complexity.

• The system looks stable.

But stability built on unexamined assumptions isn’t resilience.

It’s latent fragility.

The more an organization rewards:

• Speed over sensemaking

• Output over interpretation

• Agreement over understanding

…the more it suppresses the very signals that would prevent catastrophe.

This is why failures feel inevitable in hindsight

After the collapse, everyone suddenly remembers:

• The metric that never made sense

• The number everyone stopped trusting

• The report that was always “directionally right”

• The assumption that was never revisited

• The warning that felt politically expensive to raise

The Black Swan wasn’t hiding. It was normalized.

Why this isn’t a tooling problem

This isn’t about bad software.

You can swap:

• Dashboards

• BI tools

• AI copilots

• CRMs

• Reporting stacks

• …and the risk remains.

Because the risk isn’t in the tools.

It’s in the mindset:

Information exists to justify action, not to challenge understanding.

When that becomes the default posture, systems stop being instruments of learning and become instruments of reassurance.

The real question leaders should be asking

Not:

• “Do we have good data?”

• “Do we have enough visibility?”

• “Do we trust our tools?”

But:

Where are we quietly going along with information we don’t fully believe — and what are we building on top of it?

That question is uncomfortable.

Which is exactly why it matters.

The Black Swan isn’t coming from outside It’s being assembled internally.

• Piece by piece.

• Dashboard by dashboard.

• Assumption by assumption.

Not because people are careless — but because modern organizations are structurally good at suppressing doubt while accelerating action.

That combination is the real risk.

This site exists to challenge the idea that confidence equals understanding, and to surface the hidden fragility created when organizations stop questioning their own information stories.

Because the most dangerous failures aren’t caused by what you don’t know.

They’re caused by what everyone half-knows — and quietly agrees not to question.

The information landscape has shifted—and most organizations are still operating in the last era.

We have more data than ever.

• More tools.

• More dashboards.

• More AI.

And somehow, organizations are drowning in information that leaves everyone feeling overwhelmed.

• Decisions feel arbitrary.

• Metrics don’t agree.

• AI outputs sound confident but feel wrong.

• Teams talk past each other while insisting they’re aligned.

This isn’t because technology failed.

It’s because the problem changed, and our mental model didn’t.

Cloud computing solved the last era’s problem.

It is not enough for the one we’re in now.

The Cloud Era Solved Access: The New Era Demands Meaning.

Cloud computing was built for a world where the hard question was:

How do we store, share, and access information at scale?

And it solved that brilliantly.

• Elastic storage.

• Global access.

• Shared systems.

• Centralized data.

• Permissioned controls.

For a long time, that was modernization.

But today, the hard question is different:

How do we ensure information still means the same thing across time, teams, and machines?

Cloud platforms were never designed to answer that question.

And in the AI era, that gap has become impossible to ignore.

The New Information Landscape

What comes after cloud computing isn’t better infrastructure— it’s better epistemology.

We are no longer operating in a single-layer information environment.

The modern organization now depends on three distinct capabilities, each addressing a failure mode the cloud alone cannot handle:

AI — acceleration and generation

Shared Intelligence (SI) — collective sensemaking

Knowledge Integrity (KI) — preservation of meaning over time

Together, they form the ASK model.

Not a product stack.

Not a maturity model.

A description of what information must do in the new era.

A = AI

AI Accelerates Information Generation, But not Collective Understanding

AI radically changes the economics of information.

It can:

• generate content instantly

• summarize vast archives

• infer patterns

• answer questions fluently

But AI has a critical blind spot:

It does not know when information is obsolete.

Example: The Obsolete Email

An AI assistant is asked to draft a policy summary, customer response, or strategy brief.

It pulls from:

• archived emails

• old threads

• legacy documentation

Buried in that corpus is an email containing:

• an assumption that no longer holds

• a policy that was quietly reversed

• pricing that changed last quarter

• a customer agreement that expired

The AI has no way to know this.

It lacks:

• temporal judgment

• decision lineage

• awareness of superseded assumptions

So it produces output that is:

• fluent

• confident

• internally consistent

• …and subtly wrong.

This isn’t hallucination. It’s context decay at machine speed.

The cloud preserved the email perfectly.

AI used it perfectly.

Nothing in the system knew it should no longer be trusted.

That’s an AI failure caused not by bad models—but by missing Knowledge Integrity.

K = Knowledge Integrity (KI)

When Systems Force Certainty Where None Exists

Let’s start with something mundane.

• CRM systems.

• Ticketing systems.

• Risk registers.

• Workflow tools.

They all contain required fields:

• Reason Code

• Customer Intent

• Priority

• Risk Level

• Confidence Score

These fields must be filled in.

But reality is often:

• ambiguous

• provisional

• still unfolding

• not yet understood

So people comply.

They enter:

• best guesses

• placeholders

• socially acceptable answers

• whatever keeps the workflow moving

Now the system contains:

• structured

• validated

• non-factual information

Nothing is technically wrong.

But the meaning is corrupted at the moment of capture.

The cloud preserves that corruption faithfully.

Dashboards aggregate it.

AI consumes it.

Decisions rely on it.

This is not a data quality problem.

It’s a Knowledge Integrity failure.

Only now do we get to the principle:

Knowledge Integrity asks whether information still carries the meaning it originally had—and whether that meaning was ever true in the first place.

Without KI:

• documentation becomes dogma

• dashboards become ideology

• AI becomes confidently wrong

• decisions lose legitimacy over time

Cloud platforms enforce completeness. They do not protect truth.

S = Shared Intelligence (SI)

When Everyone Is Right and the Organization Is Wrong

Cloud computing centralized customer data.

It did not centralize customer understanding.

Marketing sees the customer as:

• a segment

• a funnel

• a conversion rate

Sales sees:

• an account

• a deal stage

• a forecast

Support sees:

• a ticket history

• churn risk

• satisfaction scores

Finance sees:

• lifetime value

• margin

• cost exposure

All of this data is accurate. All of it is in the cloud.

All of it is permissioned correctly.

And none of it adds up to a shared reality.

When Shared Intelligence is absent, you get the opposite: everyone is smart, everyone is rational, and nothing adds up.

Not because people are incompetent—because they’re operating in parallel realities.

This is what low SI feels like:

• meetings where everyone agrees on the facts but not the meaning

• decisions that make sense locally and collide globally

• teams optimizing rationally and undermining each other unintentionally

When SI does exist, it feels different:

• teams act independently without drifting

• decisions reinforce rather than contradict each other

• fewer alignment rituals are needed because interpretation is already shared

Shared Intelligence isn’t collaboration.

It isn’t consensus.

It’s the ability for many people to interpret reality compatibly, even when acting separately.

The cloud enables shared access. It does not produce shared sensemaking.

Why Cloud Computing Can’t Close These Gaps

Cloud computing assumes:

• information is stable

• meaning is implicit

• interpretation lives in people

• systems store facts

That assumption is outdated.

In the new landscape:

• AI interprets information

• artifacts outlive their rationale

• meaning decays faster than data

• decisions propagate without shared context

Cloud platforms were never designed to:

• preserve reasoning

• track assumption drift

• maintain shared models of reality

• signal when information should no longer be trusted

So organizations end up with:

• secure data

• modern infrastructure

• powerful AI

• …and rising information insecurity.

The ASK Model, Properly Ordered

Layer

What It Protects

Failure Without It

Knowledge Integrity (KI)

Meaning over time

Decisions detached from reality

Shared Intelligence (SI)

Collective interpretation

Local rationality, global incoherence

AI

Speed and synthesis

Fast propagation of bad meaning

Cloud computing sits underneath all three.

It remains necessary.

It is no longer sufficient.

What This Means

This does not require ripping out infrastructure.

It requires new practices around:

• preserving decision rationale, not just outcomes

• making assumptions visible and revisable

• allowing ambiguity instead of forcing false precision

• maintaining shared models of “what matters and why”

• treating meaning as something that must be actively maintained

• The shift is subtle but profound:

From managing information to stewarding interpretation.

The Key Takeaway

Cloud computing gave us access. The AI era demands understanding.

Without Knowledge Integrity, AI becomes fast amnesia.

Without Shared Intelligence, insight becomes fragmentation.

The organizations that win won’t be the ones with the best cloud stack.

They’ll be the ones that can still answer:

• What does this mean now?

• Compared to what?

• Based on which assumptions?

• And who else sees it the same way?

That is the new information landscape.

And cloud computing alone can’t get you there.