The Compounding Enterprise

I

The ground has gone

The senior commercial reader has been told, for a decade, that marketing is becoming more complex. The complexity is real. It is also a distraction. The discipline has been so absorbed in the changing methods of reaching the customer that it has not noticed the ground beneath its own practice turning from rock to mud. Most people are still sitting on the deck admiring the view.

The view, in this case, is the quarterly brand-tracker. Awareness steady. Consideration off a point. Net promoter holding. The deck recommends a refresh, a sharper positioning, a tighter funnel. None of these recommendations are wrong. All of them are being made on the assumption that the foundations of the practice are intact, and that the work to be done is on the visible structure built on top of them.

The foundations are not intact. They have not been intact for some years. AI has accelerated a process that platform pressure began, and the condition the senior reader is now operating in is one the discipline has no settled language for.

The foundations were these. Marketing as a practice was built on ground the era produced for free, and the discipline never had to think about because it was always there. Four conditions made the ground.

These four conditions were not strategy. They were the water the fish swam in. The marketing discipline took them as given and built its entire methodology on top of them. Positioning, segmentation, integrated communications, equity tracking: all of it was the science of how to expose the ground to the market, on the unexamined assumption that the ground would be there next quarter, and the quarter after.

The conditions have ended.

Production is no longer scarce. Aesthetic and craft can be approximated by anyone with a prompt, in seconds, at no cost. Distinctive language can be generated by a competitor on demand, indistinguishable from the original to anyone but the brand team. The cost curve of distinctiveness has not shifted: it has collapsed.

Attention is no longer channelled. The customer asks an AI assistant which running shoe to buy. The assistant returns three options with reasons. The brand is one of the three, or it is not. The assistant does not mention the campaign. It does not register the positioning. It synthesises an answer from the ground of the category and surfaces what it judges fit. The moment of encounter has been pre-empted by an interface the brand cannot buy and cannot directly address.

Meaning is no longer continuous. Cultural metabolism has accelerated to the point that a brand built on a meaning established a decade ago is now operating in a culture that has moved on twice since. The half-life of brand assets is measured in months in some categories, and the enterprise that built its equity on a twenty-year horizon is amortising into a horizon that no longer exists.

Knowledge of the customer is no longer proprietary. The data advantage that took a decade to accumulate can be approximated by a competitor using public data and AI inference, closing the gap that the enterprise spent years opening.

Each of these would be serious on its own. They are happening at once. The practice the discipline has spent thirty years refining is a practice for exposing a foundation that is no longer there.

This is not the platform argument restated. Platform pressure compressed the methods of reaching the customer. The enterprise responded with the moves it knew, spent more, adapted, metabolised. The pressure was real, but it operated on the visible layer. AI is doing something different. AI is reproducing, at near-zero cost, the outputs the enterprise spent decades accumulating the right to produce. The work of building distinctiveness has been industrialised. The cost of looking like you know what you are doing is now approximately zero.

The structural condition this names is plain. The premium positioning the enterprise has been relying on to deliver the margins the market expects was resting on a foundation that has turned, in the space of two or three years, from rock to mud. The visible structure looks the same from the deck. The brand-tracker reports normal weather. The instruments are not built to detect what has actually changed, which is that the conditions that produced the foundation as a free externality have ended, and nothing the enterprise is currently doing is generating a substitute.

This is what the senior commercial reader is actually experiencing. Pricing power softening in categories that used to hold it. Acquisition costs climbing as discovery moves out of channels the brand can buy. Premium positions defending themselves with progressively thinner margins. Share-of-voice losing its predictive relationship with share-of-market. None of these is the problem. All of them are the same problem, expressed in the instruments the discipline knows how to read.

II

The Compounding Enterprise

Last year, on a trip to Shanghai with clients, I spent an afternoon at Douyin's offices. The numbers being walked through in the meeting were the kind that produce silence in a Western marketing room. Hundreds of millions of users; live commerce sessions producing more revenue in a single hour than most brands do in a quarter; an integrated stack in which content, commerce, payment, logistics, and recommendation lived inside one platform and were optimised against one objective, which was the platform's gross merchandise value.

What I remember most clearly is not the scale. It was the brands. Or rather the absence of them. The platform had reduced the consumer's encounter with most categories to a stream of products, prices, demonstrations, and creator endorsements. The brand had not been disintermediated; it had been processed. It had become a small, slightly weighted input into a synthesised feed designed to maximise transactions on the platform's terms, not the brand's. The Chinese brands inside this environment, with rare exceptions, had been flattened before they had the time to develop the institutional depth that might have let them resist the flattening. They had emerged into the platform era and the platform era had eaten them.

The Western marketing reading of this, when the cases get cited at conferences, is that platforms are coming for the West and Western brands need to learn the platform-native tactics that Chinese brands developed in response. This is the wrong reading. Chinese brands did not develop tactics that allowed them to resist the platforms. Chinese brands were largely flattened by the platforms, and the few that have begun to break out are doing so by building exactly the kind of institutional distinctiveness that the Western discipline has spent thirty years failing to take seriously. The China case is not a tactical lesson. It is a precautionary tale that the West is now living through in compressed time, with AI playing the role that the platforms played there, only faster, and across more dimensions of the brand at once.

This matters because the C-suite has already started asking, in private and tip-toeing, whether brand is dead in the AI age. The instinct behind the question is correct. The conclusion is wrong.

What is dead is the brand the discipline has been building for thirty years. The brand-as-projection, constructed by a marketing function on top of an operating model that has been organised around something else, sustained by media spend and channel relationships and the slow continuity of an era that produced its own raw materials for free. That brand is dead. AI has not killed it elegantly; AI has industrialised the production of everything that brand depended on, until the layer the discipline was producing has become the layer everyone can produce, which is by definition the layer that no longer differentiates anyone.

What replaces it is a different thing entirely, and the rest of this piece is about what that thing is, why the discipline has barely understood it, and what the work of building it now consists of.

The structural reason all of this is happening at once is older than the AI age, and it is the reason this argument should sound familiar to anyone who has read the history of technological change.

In the late nineteenth century, factories began to replace steam with electricity. They did this badly. The early electrified factory was a steam factory with an electric motor where the steam engine had been: one large central source of power, distributing mechanical energy through a system of belts and pulleys called a line-shaft, driving fixed stations across the factory floor. The new technology had been adopted. The architecture inherited from the old technology had not been touched. Productivity gains were marginal. Some factories saw none.

The unlock came thirty years later, when a generation of plant managers understood that the electric motor could be small and distributed. Each station could have its own. The line-shaft could be torn out. The factory could be re-laid around the logic of the work itself rather than around the constraint of distributing power from a single source. Production lines, as we now understand them, were the result. Productivity moved by margins that retired the previous architecture inside a decade. The factories that had electrified the line-shaft did not survive the transition. They had adopted the technology and refused the re-architecture. They were not in the same business as the firms that had understood what the technology was for.

Most enterprises today are electrifying the line-shaft. They are buying AI tools and deploying them inside the operating model they already had, to make the existing work marginally faster and cheaper. Marketing uses AI to produce more content. Strategy uses AI to accelerate analysis. The creative function uses AI to generate variations. The line-shaft is intact. The motors are bigger. The work flows through the same channels in the same shape at the same altitude, with AI substituting for human labour at specific points along the line. Productivity moves a few points. Distinctiveness collapses, because the projected layer is now produced by everyone at zero cost and the institution underneath has not been re-architected to produce something AI cannot reach.

Brand-as-projection is what brand looks like inside a line-shaft enterprise. The discipline of producing it is the discipline of polishing the layer while the foundations below are dissolving. The polishing is being done at industrial scale, by competent professionals, in good faith. It is also, structurally, the wrong work.

The brand the AI age requires is not a layer. It is the externalisation of the institution that produces it.

This is the load-bearing claim of the piece, and it deserves to be stated cleanly. Brand is not what the marketing function constructs and projects. Brand is what the institution accumulates and expresses, over time, through the totality of how it operates, judges, decides, hires, builds, and acts. The marketing function does not produce the brand. At its best, the marketing function reports on a brand the institution is producing through how it works. At its worst, which is most of the discipline most of the time, the marketing function is constructing a layer that is fighting the institutional gravity rather than expressing it.

The exceptions have always been visible. The enterprises with the strongest, most enduring competitive position in their categories have, intuitively or by design, been organised around the production of their own distinctiveness top to bottom. Apple under Jobs is not an enterprise with a brand layer. Apple under Jobs is an enterprise whose every operational, product, retail, hiring, and supplier decision is read by the market, over time, as the brand. Patagonia is not a marketing programme. Patagonia is an institution whose externalisation, from the supply chain to the repair policy to the political stance, is what the customer encounters and recognises. Hermès is not a luxury brand in the projection sense. Hermès is an institutional discipline of craft, scarcity, and continuity that the market reads as a brand because it has nowhere else to read it.

The discipline has treated these as exceptional, charismatic, inimitable. They are not. They are what brand looks like when an institution is organised to produce it. They have been visible the entire time. The discipline has treated the greater model as a curiosity and organised itself around the lesser one, because the era forgave the lesser one and the lesser one was easier to staff, scale, and sell.

The era no longer forgives it. The greater model is the only model that compounds in the AI age, because every other model is in the business of producing layers that are now produced by everyone at zero cost.

This argument has to survive a reasonable objection, which is that many of the enterprises that face this question operate portfolios of brands rather than single ones. Procter & Gamble is not a brand; it is a holding architecture under which Tide, Pampers, Gillette, Olay, Crest, and several dozen others operate. General Motors runs Buick, Chevrolet, GMC, Cadillac. Volkswagen Group runs VW, Audi, Skoda, Seat, Porsche. Are we suggesting that brand distinctiveness across categories collapses into a single corporate institution?

We are not. The principle holds at the level of the institution that produces the brand, which is not always the corporation that owns the brand. Tide is the externalisation of the Tide institution: the operating unit, the R&D programme, the marketing organisation, the seventy years of accumulated decisions about laundry, fabric, and the consumer who does it. P&G is the externalisation of the P&G institution, which is a different institution operating at a different altitude and read by different audiences. Both are real. Neither is reducible to the other. The portfolio enterprises that have managed their brands well over decades have, intuitively, run their category brands as semi-autonomous institutions with their own accumulated context, judgement, and operational distinctness. The corporate centre provides scale infrastructure; the category institutions produce the brands.

The portfolio enterprises that have struggled, particularly over the last fifteen years, have done the opposite. They have centralised brand production at the corporate centre, consolidated marketing into shared services, harmonised agency rosters, rationalised category leadership, and discovered, after a few cycles, that the brands have flattened. The diagnosis is exact: the corporate centre has hollowed out the category institutions and replaced them with centrally-produced projection layers, which AI deflation now flattens into commodity. The brands look the same on the balance sheet for a few years, then begin to lose pricing power, then begin to lose share, then arrive in the strategy review framed as a category headwinds problem.

This is also why most acquisitions in consumer goods, beverage, and adjacent categories have been dilutive over the long run. The acquirer prices the projected layer and disperses the institution that was producing it, almost as a matter of policy, in the name of integration synergy. Unilever paid roughly a billion dollars for Dollar Shave Club in 2016. Seven years later it sold a majority stake on undisclosed terms, in a transaction widely read as value-destructive. The post-mortem framings emphasised market saturation and the maturing of the direct-to-consumer model. The structural explanation is that Unilever bought an institution and integrated it into a corporation that was not organised to preserve what made the institution distinctive. The mechanism predicts the outcome with depressing reliability, and the cases stack up across decades. L'Oréal's portfolio is the contrasting positive case: Lancôme, Kiehl's, La Roche-Posay, and others have been preserved as institutions inside a corporate architecture that knows what it bought and protects it. The compounding has continued. The acquisitions have paid out.

The work of building the Compounding Enterprise consists of five things. They are not five workstreams; they are five lenses on a single re-architecture project, and any competent engagement integrates them rather than running them in parallel.

This is the work the AI age requires. It is not a marketing programme. It is not an operating model review. It is the deliberate construction, protection, and compounding of the institution that produces the brand, against the structural pressures that flatten everything resting on the projection layer.

It is CEO and COO work, with the marketing function, the operating leadership, and the talent function as participants in a project they are not equipped to deliver alone. It is the foundational architecture of competitive advantage in the AI age, addressed to a question the discipline has not yet learned to ask, at an altitude most consulting work has not yet learned to operate at.

The line-shaft is being polished in most enterprises right now, by competent people doing what they were trained to do. The factories that survived the last transition were the ones whose plant managers stopped polishing the line-shaft and started designing the line that would replace it. The work is the same. The technology is different. The enterprises that understand this are the enterprises whose institutions will compound. The rest will be doing the work of a discipline that no longer answers the question the market is asking.