We are entering the age of uncapitalized intelligence

It has never been easier to produce strategic analysis. An investment thesis comes together in a morning, an executive committee memo in an hour, a dynamic market sizing in a few prompts. AI has compressed production time to a point few organizations have truly absorbed. Cycles have sped up, documentation has thickened, deliverables multiply.

We produce more strategic intelligence in a quarter than most firms produced in a year five years ago.

And no one has become any smarter collectively.

The comfortable lie of analytical productivity

In organizations with high decision density — funds, strategy committees, executive teams — a story has taken hold. AI frees up analysts' time. Seniors focus on judgment calls. Quality rises. So does velocity.

This story is false in its most important part.

Quality doesn't rise. Quantity rises. Leaders confuse the two systematically because they consume deliverables, not trajectories. From the top, every file looks better than the last. Over three years, the organization rebuilds the same reasoning with slightly different assumptions, slightly updated comparables, slightly redefined segmentations. Engagement after engagement. Committee after committee.

Individuals learn. Files move forward. The institution starts over.

The finding is no outlier. The MIT NANDA report, The GenAI Divide: The State of AI in Business 2025, documents that of the $30 to $40 billion U.S. companies have invested in GenAI, 95% of pilots produce no measurable P&L impact, and only 5% of organizations manage to integrate AI tools into production at scale. The problem, according to the authors, isn't technological. It's organizational — a learning gap between what the models produce and what organizations know to do with it.

Why knowledge management never solved this

The objection comes fast. The subject isn't new. Organizations have invested tens of millions in document repositories, tagging systems, dashboards, internal wikis, best-practice portals.

None of it produced cumulative strategic memory. It produced searchable archives, which is the exact opposite of a living institutional memory. An archive is an organized cemetery. A strategic memory is a structure that changes how the next analysis gets built.

The distinction isn't semantic. It's operational. A team that consults a 2023 memo before producing the 2026 one is doing internal benchmarking. A team whose 2026 memo can't be produced without the 2023 one being explicitly re-examined, versioned, and either confirmed or invalidated — is capitalizing.

The first model is what 95% of sophisticated organizations practice today. The second barely exists.

AI worsens the very thing it claims to solve

The obvious argument holds that AI, by multiplying production, also multiplies the raw material of institutional memory. More notes, more summaries, more traced reasoning. So more potential capital.

The reasoning is backwards.

An organization that produced ten analyses a year could, with discipline, make them a coherent body of work. The same organization producing a hundred and fifty mechanically can't make anything but an ocean of competing versions. Assumptions diverge silently. Segmentations drift. Comparables contradict each other from one file to the next without anyone seeing it, because no one has time to read it all.

The flaw isn't a tool flaw. It's a flaw in the law of composition. Ten coherent units make a system. A hundred and fifty juxtaposed units make noise.

The financial numbers confirm the scale of the drift. According to the 2025 State of AI Cost Management Report from Benchmarkit and Mavvrik, 80% of companies overshoot their AI infrastructure forecasts by more than 25%, and 84% report significant gross-margin erosion tied to AI workloads, with 26% taking a hit of 16% or more. McKinsey's State of AI 2025 rounds out the governance picture: 51% of organizations report at least one negative AI incident over the past twelve months — output inaccuracies, compliance breaches, reputational damage — and only a third say they have truly scaled AI across the enterprise, the other two-thirds stuck in what analysts call pilot purgatory.

Most organizations don't steer AI as an asset. They endure it as an opportunity cost that accumulates faster than the value it generates.

What leaders confuse: insight and capital

An insight is a one-off perception. Strategic capital is a perception that changes the system that will produce the next one.

This distinction is invisible in the deliverables and decisive in the trajectories. Two organizations can produce the same memo with the same conclusion, and one will have built capital while the other simply burned time. The difference doesn't show in the document. It shows in what happens once the document is filed away.

In the first organization, the memo has just written into the shared grammar a hypothesis that will be tested, compared, refined by the analyses that follow. In the second, the memo joins a file no one will reopen except by accident, and the next analysis will start from a near-blank page, shaped by the new author's mental habits rather than by the inheritance of the one before.

The first organization accumulates. The second performs. Over ten years, the gap is dizzying, and most leaders will never have seen it coming, because year after year their deliverables will have looked excellent.

The real bottleneck isn't analytical

For twenty years, the strategic bottleneck was producing insight. Entire firms were built on that scarcity. AI erased the advantage in under three years.

The new bottleneck isn't decision-making either, contrary to what the consultants now selling decision intelligence claim. Decision-making will stay constrained by governance, accountability, authority — by the organization's social body, not its cognitive machinery. No technology will durably speed up the pace at which an investment committee commits five hundred million euros.

The new bottleneck sits between the two. It's in the layer that turns abundant analytical output into usable institutional patrimony. That layer doesn't exist as a market category. It has no canonical name. It appears in no IT budget. And that is exactly why it will become the strategic differentiator of the next ten years.

The competitive advantage no one measures yet

The organizations that will dominate their sector in ten years won't be the ones that produced the most analyses. They'll be the ones whose thousand-and-fifth analysis is structurally superior to their hundred-and-fifth, because it was composed on the nine hundred and ninety-five before it rather than written beside them.

This superiority won't show in any single deliverable. It will show in the speed of reaction to a reversal, in the finesse of a judgment call in the grey zone, in the ability to recognize a weak signal that resembles one already met three years earlier and correctly tagged at the time.

It's a superiority of trajectory, not of production. The financial markets, which price trajectory better than operations, will end up pricing it. The leaders who haven't built this layer will wonder why their multiple eroded while their quarterly results stayed solid.

The answer will be in what they never measured: the share of the intelligence they produced that actually compounded.

The question that should keep executive committees up at night

Most leaders I meet know how to assess the quality of a deliverable, the ROI of an investment, the performance of a team.

Almost none can answer this one: what share of the strategic intelligence my organization produced this year will still be actively shaping decisions two years from now?

If the honest answer is below 20%, the organization isn't capitalizing. It's consuming.

And in a world where production has become trivial, consuming your intelligence without capitalizing it is no longer an operational inefficiency.

It's a silent decapitalization of the one asset that will stay rare.