The Knowledge Divide: How AI Will Reorganise Work, Margins, and Power

Editor’s Note: This article is the first in a new Founding Fuel series exploring how artificial intelligence is reshaping organisations and the future of work.

Much of the current conversation around AI remains focused on tools, productivity, and use cases. This series takes a different lens—examining how AI is beginning to reconfigure organisational architecture: how decisions are made, how knowledge flows, and how work itself is organised.

The essays aim to surface the strategic choices leaders now face—and the implications those choices may have for how work, margins, and power evolve over the coming decade.

It is the early sixteenth century on the streets of Florence. We catch a glimpse of Leonardo da Vinci moving between sketches, machines and anatomy. Michelangelo directs one of his chapel masterpieces. But behind their individual brilliance, the studios are alive with activity—apprentices grinding pigments, assistants stretching canvases, sculptors roughing out marble before masters refine the form.

Renaissance Florence was extraordinary not just for its talent, but for how creative intelligence was organised—through systems that allowed mastery to scale beyond individuals.

Today, artificial intelligence is beginning to raise a similar set of questions inside modern organisations—not just about tools or productivity, but about how intelligence itself will be organised, and what role humans will play in that architecture.

The real shift is not just automation. It is how knowledge is structured, embedded, and allowed to move—or not move—across organisations.

The real shift is not just automation. It is how knowledge is structured, embedded, and allowed to move—or not move—across organisations.

The Organising Question

Agents are systems that can read information, make decisions, and act across workflows. A sales agent can assemble proposals using historical deal data. A support agent can interpret queries and retrieve answers. A procurement agent can compare suppliers. A finance agent can flag anomalies.

As these systems mature, the question is no longer whether companies will adopt AI. The deeper question is how intelligence itself will be organised inside the firm.

Over the coming decade, companies will diverge into three distinct architectures of intelligence: organisations where AI agents run most workflows, organisations that rent intelligence from external markets, and organisations where humans remain the primary carriers of knowledge.

Each architecture will reshape how work is performed, where profits accumulate, and how companies rise—or disappear.

Three Emerging Architectures of Intelligence

Archetype 1: Agent-First Organisations

In these organisations, companies train or deeply customise models using proprietary data. These models power agents that operate across workflows—from reconciliations and documentation to complex coordination and decision-making.

Human roles shift sharply. Execution-heavy work declines as agents take over operational tasks. A smaller group of highly skilled individuals designs, trains, and supervises these systems.

In this model, institutional knowledge becomes infrastructure. What once lived inside teams is embedded into systems and made available across workflows. Productivity rises first, followed by speed and operating leverage.

But only a limited number of firms will reach this level. It requires deep data, technical capability, and sustained investment. Those that do may not only transform themselves but also begin supplying agents to others—turning internal intelligence into a revenue stream.

This creates a strategic dilemma: how much proprietary knowledge should a firm expose when those same agents may eventually serve competitors?

Archetype 2: The Agent Economy (Rent and Integrate)

Many firms will not build deeply contextual systems. Instead, they will adopt agents developed by specialised providers or by more advanced firms.

Over time, a marketplace of agents will emerge. Organisations will be able to access systems built for specific functions and industries—sales, support, compliance, and more.

What these agents lack is deep company context. They must be integrated into internal workflows before they become effective.

That integration becomes the central human role. Employees refine context, adapt processes, and supervise outputs where judgement is required. Implementation and orchestration become core capabilities.

Efficiency improves, but differentiation depends on how well firms integrate and adapt these systems.

Archetype 3: Human-Centred Organisations

A third group of firms will continue to operate through primarily human-led systems.

Employees use AI tools individually, but intelligence remains embedded in people rather than integrated systems. Decision-making relies on judgement, experience, and experimentation.

These organisations may remain large employers in the near term but will face pressure. Compared with agent-first firms, they may struggle to match productivity and margin expansion.

Yet they may also become a source of discontinuous innovation. As AI-generated knowledge converges toward the statistical centre, human-led systems may explore ideas that fall outside algorithmic consensus.

Some future industry leaders may emerge from this group—until they themselves scale and begin building agent architectures.

The Knowledge Lock-In Loop

Across these archetypes, a deeper system dynamic begins to emerge.

AI embeds knowledge into systems. As a result, knowledge moves less through people. Firms begin to operate as more closed systems. Divergence across companies increases. Over time, power concentrates in a few firms while many others struggle to keep pace.

In agent-first organisations, knowledge becomes deeply embedded in proprietary systems and a small group of specialised employees. Mobility reduces. What is shared with the market is often a diluted version of internal capability.

In the agent economy, firms build on external systems but enhance them with proprietary context. These improvements are rarely shared back, creating fragmented ecosystems.

In human-centred firms, knowledge continues to move through people. These organisations may remain more adaptable—but face pressure from more efficient competitors.

This creates a recurring loop: knowledge is embedded, mobility reduces, systems become closed, and industries diverge.

AI embeds knowledge into systems. As a result, knowledge moves less through people. Firms begin to operate as more closed systems.

Leadership Imperative: Choosing the Knowledge Architecture

The first responsibility of leadership in the age of AI is to recognise where organisational intelligence resides.

In some firms, knowledge will live inside coordinated AI systems. In others, it will sit partly in external platforms. In many, it will remain within people.

The critical decision is which architecture the organisation will evolve toward. Once chosen, the path becomes difficult to reverse.

This choice will shape cost structures, innovation capacity, and long-term survival.

A New Architecture of Work

Florence’s studios organised artistic intelligence into systems that allowed creativity to scale beyond individuals.

Artificial intelligence may now trigger a similar shift. Companies will organise themselves around different architectures of intelligence, each with its own trade-offs in productivity, innovation, and power.

The firms that win will not simply adopt AI. They will decide where knowledge lives—and how it moves.

That choice will shape how value is created, how power accumulates, and which organisations endure.

(This article opens a series exploring how these architectures will reshape work, margins, and power across industries.)