The Next Rung of Reality

Jane Smith got her letter to 2030 right. Here is what was difficult to see from there — and why the rung she stood on was never meant to be the top.

Jane Smith published a letter this week that is really, really worth your time. Written in the voice of a CDAIO passing the baton in 2030, it is funny, honest, and more structurally correct than most things written about the data function in the past decade. The ficus joke is good. The observation about SQL is earned. The diagnosis of what happened to the pipeline builders is accurate.

https://www.linkedin.com/pulse/fall-rise-data-teams-letter-new-chief-ai-officer-2030-jane-smith-ald5e/

I am not writing to disagree with Jane. I am writing because her letter, as good as it is, ends one rung short of where we will likely be in 2030. And in a field where stopping one rung too short has historically been the proximate cause of every major failure, that matters.

What follows is a response from a little further down the road — not 2030, but further. From a vantage point where the events she describes are already in the rearview, and where the thing she could not quite see from 2030 has become the central problem of the decade.

The premise — Jane's argument in brief

The CDAIO function was decimated between 2025 and 2030 because it had organized itself around activity rather than assets. Pipeline builders were replaced by automation. Dashboard developers were replaced by self-serve AI. The middle of the stack was eaten.

What survived — what rose to replace the old function — was the semantic layer: the curated, governed, single source of truth that gave both humans and AI agents unambiguous context. The Semantic Architect became the critical role. Data became an asset class because meaning became the moat.

"Context became the asset," she writes. And she is right.

That argument is correct. It is also incomplete in a way that has consequences.

The rearview mirror, beautifully polished

Here is what the semantic layer actually is, stated plainly: it is a better-organized rearview mirror.

This is not an insult. A well-built rearview mirror is enormously valuable. For decades, organizations made decisions with mirrors that were cracked, fogged, and showing different images to different passengers in the same car. The semantic layer fixed that. It gave everyone the same clear view of what was behind them. The Semantic Architects did work that genuinely mattered.

But a rearview mirror — however precise, however well-governed, however free of ambiguity — is still pointed backward. It tells you what happened. It resolves disputes about the past with remarkable clarity. It is, in the language of epistemology, a system for achieving consensus about prior states of the world.

What it cannot do is tell you why something happened.

What it cannot do is tell you what will happen when you intervene.

What it cannot do is account for the fact that the historical record it is drawing on may no longer represent the world as it currently operates.

That last point is the one that breaks everything. And it was already breaking in 2026, well before Jane's 2030 letter was written.

The collapse that the architecture could not stop

There is a phenomenon that has been underway for years and that almost no one in the data industry discussed with any seriousness until the consequences became impossible to ignore. I have been calling it Correlative Collapse — the accelerating decay of historical data's predictive validity.

The basic mechanism is not complicated. Correlative models — the models that underpin virtually every enterprise analytics and AI system in operation today — work by finding patterns in historical data and projecting them forward. They assume, as a structural matter, that the relationships they have found in the past will continue to hold in the future. When that assumption is sound, they work well. When it degrades, they fail.

The assumption has been degrading for years. The rate of structural change in markets, supply chains, consumer behavior, competitive dynamics, regulatory environments, and geopolitical conditions has been accelerating. Events that were once statistical outliers — the kind that models could safely ignore — began arriving with increasing frequency. The half-life of a pattern found in 2019 data became shorter in 2021, shorter still in 2023, and by 2025 was in some sectors approaching the span of a single planning cycle.

The data was not wrong. The world had simply moved on from it. The data had no continuing relevance. It became an obsolete reflection of a past that no longer exists.

This is the part that makes Correlative Collapse so dangerous: it does not announce itself. The semantic layer tells you what your data says with great precision. It does not tell you that what your data says may have a rapidly shortening relationship with what is actually true in the world today. The better you make the mirror, the more confidently you can read a reflection that is already outdated.

Without necessarily reflecting on it, Jane was building a better rearview mirror in 2030. She was right that it was better, but she could not yet see that the road had bent.

Then the boards got sued

The third revelation — the one that forced the issue into the boardroom and out of the data function — came from a direction that the CDAIO industry had not been watching closely enough: the courts.

The evolution of fiduciary duty law in Delaware had been a slow-moving signal for years. The McDonald's ruling in January 2023 extended officer oversight obligations in ways that, when combined with subsequent decisions, created a legal framework with a specific implication for AI-assisted corporate decision-making. Boards and officers could not simply point to a well-governed analytics system as evidence of due diligence. They had to demonstrate that they had interrogated the causal warrant behind the decisions those systems recommended.

Causal warrant is the operative phrase. Not confidence interval. Not semantic governance. Not a clean data lineage. The question the courts began asking was: what is your basis for believing that the causal mechanism you are assuming in this decision actually operates in the current environment? How do you know that the relationships your model found in historical data still hold today, given the pace of structural change in your market?

These are not questions a semantic layer can answer. They are not questions any correlative system can answer. They require something that the data industry had been systematically avoiding building: genuine causal infrastructure.

The semantic layer told the machine what words meant. The courts wanted to know what was actually causing what — and whether the organization could demonstrate it knew the difference. That was the question no dashboard could answer because the Reality was in between the data points.

This was the forcing function. Not a technology trend. Not a vendor roadmap. Not a visionary keynote. A fiduciary obligation, adjudicated in court, that created discoverable liability for organizations that had substituted correlative confidence for causal understanding.

Modern decisioning systems did not emerge from a LinkedIn article about the future of data. It emerged from a board meeting where the general counsel explained what the latest Delaware decision meant for the decisions they had made the prior quarter using their best-in-class semantic layer.

The three revelations as a single arc

Let me pull the thread through all three, because they are not separate problems. They are one problem, seen from three angles.

The semantic layer solved the meaning problem. Before it, organizations had multiple conflicting sources of truth, ambiguous definitions, and no stable foundation for either human analysis or AI agents. Solving that was necessary and real. Jane is right to celebrate it.

But meaning without mechanism is just a precise description of a pattern you do not understand. You know what the numbers say. You do not know why those numbers are what they are, what forces are producing them, or whether those forces are stable. You have resolved the ambiguity of language. You have not resolved the ambiguity of causality.

Correlative Collapse is what happens when the mechanism changes and the meaning system does not know it. The semantic layer continues delivering clear, well-governed, unambiguous answers about a world that has quietly shifted beneath it. The organization acts on those answers with increasing confidence. The confidence is not warranted because it is not connected to a live understanding of what is actually causing what in the current environment.

The fiduciary exposure is what happens when that gap — between correlative confidence and causal understanding — produces a bad decision large enough to attract legal scrutiny. At that point, the question is not whether your data was clean. The question is whether you had a basis for believing your causal assumptions were valid. And if you built your entire decision infrastructure on a semantic layer, you do not have a good answer to that question.

"Meaning without mechanism is still just a faster way to be confidently wrong."

The next rung on the ladder — the one that Jane's 2030 vantage point could not yet see clearly — is causal infrastructure. Not as a replacement for the semantic layer. The semantic layer is necessary. It is the foundation. But a foundation is not the building.

What this means for the organization you are running today

I am aware that this essay is not purely historical. The Correlative Collapse is not a future problem. It is happening now, in your organization, in your data, in your decisions. The question is whether you have built the infrastructure to see it.

The practical implication is this: somewhere in your analytics stack, there is a set of causal assumptions that have never been made explicit. Your models assume that the relationships they found in historical data still hold. Your decisions assume that the mechanisms producing your key metrics are the same mechanisms that were operating when the training data was collected. Your forecasts assume that the structural conditions of the market are stable enough that pattern-projection is a valid epistemic strategy.

Some of those assumptions are probably still sound. Some of them are not. The semantic layer cannot tell you which is which. Only causal analysis can do that — analysis that goes upstream of the data to ask what is actually causing what, tests whether those causal structures are stable, and flags when the mechanisms have shifted faster than the historical record can represent.

This is not a technology pitch. It is an epistemological obligation. Organizations that are making significant decisions — capital allocation, market entry, product strategy, risk assessment — on the basis of correlative systems, however well-governed, are making a bet that the world has not moved faster than their data. That bet is becoming harder to win.

A word about Jane

I want to be clear about something before I close. Jane's letter is a really important contribution. The insight that data is an asset class with a limited life — that the function had to be rebuilt around value creation rather than ticket fulfillment — is correct and important. The prediction that the semantic layer would become the surviving moat was accurate. The observation that Logic Assurance and critical thinking would replace dashboard literacy was prescient.

She got her moment right. That is pretty damn hard to do. A lot of people do not get their moment right.

They are either early, which looks like you're wrong, or late, which means that you're wrong.

What I am arguing is that Jane's moment was a rung on a ladder, not the top of the ladder. This is the nature of genuine progress — each real insight reveals the next real question. The semantic layer reveals the causal question precisely because it solves the meaning question so well. You cannot clearly see that you are missing causal infrastructure until you have built the semantic infrastructure that shows you the shape of what is still missing.

What Jane wrote was a prerequisite for what comes next. The ficus is thriving. The next generation of the function is being asked a question the old one never had to answer.

The question is not what does your data say.

The question is what is actually causing what and how do we intervene effectively to change it — and do you have a defensible basis for that answer?

That question has a legal address now. It has a fiduciary address. It has a board address. And it will not be answered by a better rearview mirror, however polished, however well-governed, however beautifully architected.

It will be answered by organizations that build the next rung.

Jane Smith's original post — "The Fall and Rise of Data Teams: a letter to the new Chief Data & AI Officer in 2030" — was published on LinkedIn on April 28, 2026. It is worth reading in full before reading this response. The argument here is built on top of hers, not against it.