From extraction to preservation

The extractive lineage

Natural resources were the first great conquest. Forests felled, ores mined, oceans overfished: the industrial revolution turned the planet into a reservoir of raw materials. When physical extraction hit its limits, attention shifted to a new frontier: human labor. Taylor decomposed every gesture into measurable units, turning the worker into a cog whose every second had to be monetized.

With the digital age came a new mutation: human attention became the scarce resource. Herbert Simon had stated it as early as 1971 — "a wealth of information creates a poverty of attention." Web 2.0 platforms industrialized this capture and built empires on the monetization of our digital presence.

We are now entering the final phase of this progression: cognitive extraction. It is no longer merely our time or attention being monetized, but our thinking itself. Large language models ingest billions of pages, absorb our ideas, our reasoning, our creations, to reassemble and redistribute them. Every interaction becomes training data that enriches the model.

And this extraction has a troubling property: it is bidirectional. Humans feed the machine their knowledge, and the machine, in return, shapes human thought — standardizing answers, flattening nuance, reflecting back a homogenized way of thinking.

Having exhausted resources, rationalized labor and monetized attention, we are now extracting the very essence of what defines us as thinking organizations.

Digital transformation: an honest reckoning

Before proposing a new path, let us look honestly at the road traveled. For fifteen years I actively took part in what was called digital transformation — selling and implementing enterprise platforms at PwC, IBM, ServiceNow, Kyriba. That experience demands a nuanced assessment, without complacency but without unfair hindsight.

The transformation answered a real urgency. Organizations of the early 2000s ran on inherited systems, incompatible silos, costly manual processes. Platformizing finance, HR and procurement produced measurable efficiency gains. Nobody mourns the lost diversity of payroll systems.

The problem appeared when platformization reached the differentiating processes: pricing, sales strategy, R&D, decision-making. When every company in a sector adopts the same CRMs with similar configurations, their teams converge toward the same qualification processes, the same pipelines, the same indicators. Methodologies standardize, strategies align. Competitive advantage erodes precisely where it should have been preserved.

This is no malice on the vendors' part — it is structural. The customization that would preserve singularity runs into cost (tunnel effects, technical debt) and risk (version dependencies, unsupported developments). Facing these constraints, most choose the standard. Short-term efficiency wins over long-term differentiation.

And the paradox is cruel: the organizations holding the most accumulated cognitive capital are the ones most pressured to "transform" — to resemble startups that have, precisely, no heritage to preserve. Digital transformation treats their history as a burden. It is an asset.

The cognitive phenotype, a source of durable value

To escape this impasse, we need a structuring concept. The term phenotype comes from biology: the genotype is the inherited genetic material, the phenotype its observable expression in a given environment. Two organisms with identical genotypes develop different phenotypes depending on the conditions they encounter.

Transposed to organizations, the cognitive phenotype designates the unique way a company thinks, reasons, decides and innovates — shaped by its history, its culture, its sector, its experience. It is written in no procedure manual. It shows in a thousand daily details: how teams react in a crisis, the kind of questions asked in a strategy committee, the balance between boldness and prudence, the way failures are interpreted.

These patterns are not arbitrary. They were selected because they worked. They constitute an adaptive advantage forged by experience — and destroying them in the name of standardization means erasing what allowed the organization to survive.

Strategy consultancies offer the most visible example: they literally sell their way of thinking. Each major firm has forged its own epistemology; a consultant internalizes it within five years until it becomes instinctive. Now imagine they all adopt the same generative tools, trained on the same public corpora: proprietary frameworks erode, firms converge toward a median approach, clients stop seeing the difference. This scenario is not hypothetical — it is what happens every time a sector adopts the same platforms without a preservation strategy.

The same goes for banking: millions of lines of legacy code are not merely a technical obstacle — they are the programmatic expression of decades of business reasoning. Every risk-control rule designed in the 1990s was tested, refined, corrected through crises. It survived because it worked. Rewriting everything from scratch on standardized platforms means risking throwing out the knowledge with the obsolescence.

Pierre, 58

Take the example that motivated this reflection. Pierre, an offshore engineer for thirty-five years, holds an expertise forged by thousands of incidents. Facing a particular combination — pressure at 847 bars, rising temperature, unstable weather — his brain computes in three seconds the probability of a major incident and decides to evacuate the platform. That decision saves more than a hundred people. It compresses tens of thousands of hours of experience. It can be neither coded into a manual nor transmitted in standard training. In the traditional model, when Pierre retires, it disappears.

Amplification proposes something else — and it begins where extraction never does: with consent. With Pierre's agreement and under his control, his reasoning is crystallized: not what he does, but how he thinks. The variables he weighs in critical situations, the patterns he recognizes, the mistakes he has learned to avoid. And Pierre remains its owner — it is his reasoning, under his name, whose reach he decides. Nothing to do with a clone: no avatar, no imitation — a documented piece of reasoning, attributed and dated, like a work of authorship.

Recorded in verified, contextualized, traceable knowledge units, that reasoning remains consultable: in 2028, a young engineer facing a similar situation does not get a ready-made answer, but the thought process that led Pierre to his. He learns to think like Pierre, not merely what to do — and every consultation carries Pierre's name.

An expert can train only a few dozen people in a career. A preserved cognitive phenotype can be consulted by thousands, simultaneously.

That is the difference between the two logics. Disruption rests on replacement: the old is destroyed to make room for the new — a binary logic that can only work a finite number of times. Amplification rests on preservation and multiplication: the existing is not destroyed but made more powerful, more accessible. We move from a model where technology replaces the human to a model where it multiplies the human.

Cognitive sovereignty

This distinction takes on a geopolitical dimension. The widely adopted generative systems operate on an extractive principle: every conversation enriches proprietary models, controlled outside Europe, subject to non-European regulation. When all of an organization's employees use the same external system for their analyses, they converge toward a median style of thought, statistically optimal according to the model's training corpora.

Europe has partially lost its digital sovereignty — cloud, operating systems, social networks. We now risk losing our cognitive sovereignty by delegating our organizational reasoning capabilities to systems that, by construction, homogenize rather than preserve. The answer is not autarky — AI technologies are global by nature. It is a strategy of sovereign preservation: keeping one's cognitive phenotypes on one's own infrastructure, under one's own rules, without extraction by third parties. On-premise or European sovereign cloud deployment, auditable algorithms, a phenotype preserved in structures specific to the organization rather than diluted into a global model.

Update, June 2026. The demonstration is no longer theoretical. On June 12, 2026, one of the most widely used frontier models in the world was made inaccessible overnight, everywhere, following an export control directive decided in the United States — with no notice and no recourse for the organizations that had built their workflows on it. Access was restored three weeks later. Whatever the stated reasons: the dependence has been demonstrated. A reasoning capability whose switch a third party can flip is not an infrastructure — it is a revocable subscription.

Cognitive sovereignty is the natural extension of the data sovereignty that the GDPR and then the AI Act established. This is not a technical question — it is a question of organizational survival in a world that homogenizes.

Why now

If preservation is theoretically superior to extraction, why is it only emerging now? Because several momentums are converging, precisely in this period.

The demographic tsunami. Baby boomers are retiring massively. In offshore energy, 45% of the workforce is over 50; Goldman Sachs projects 510,000 hires needed in energy by 2030; the Dallas Fed documents the departure of roughly 30% of qualified electricians while training one takes three to five years. Every day, thousands of experts leave, taking decades of know-how with them. The window is narrow: within five years, a significant share of this expertise will be gone for good.

The need for sovereignty. Since the Snowden revelations, accelerated by geopolitical tensions, sovereignty — energy, technology, food — has again become a strategic priority. Cognitive sovereignty must be part of it: an organization that does not control its own reasoning capabilities is as vulnerable as one whose supply lines can be cut.

The search for meaning after slop. Informational pollution has produced generalized cognitive fatigue. The return to verified, traceable knowledge, grounded in real experience rather than algorithmically generated, becomes a considerable trust advantage. Preservation is the antidote to slop: every knowledge unit carries the expert's identity, the decision's context, the reasoning's sources.

The technological momentum. The required capabilities did not exist ten years ago. Architectures descended from Transformers, knowledge graphs and agentic systems are now mature; costs allow viable on-premise deployment; European sovereign clouds offer credible alternatives. These technologies have mostly served an extractive logic — training giant models that homogenize. We propose to employ them in the opposite logic: crystallizing specific cognitive phenotypes — with their bearers' consent — without diluting them.

The civilizational alignment. Preserving the expertise of the best practitioners and making it accessible at scale directly serves quality education (SDG 4), decent work in transitioning sectors (SDG 8) — an offshore oil expert whose reasoning serves offshore wind — and resilient innovation infrastructure (SDG 9). This is not cosmetic: it is what makes preservation a societal necessity, not merely a business opportunity.

The organizations that dominate the next decade will not be those that best imitated their competitors, but those that preserved and amplified what makes them unique.