Investment thesis
Owning the Control Layer
of the AI Economy
A Quiet Shift Beneath the Noise
Most technological revolutions are misread in their early stages. They are interpreted through the lens of what is visible: new tools, new interfaces, new capabilities. This is where we are today with artificial intelligence. The conversation is dominated by models, copilots, and applications that appear to augment human work. These are impressive, but they are not where enduring value will ultimately accrue.
What is actually unfolding is far more structural. Artificial intelligence is not simply improving software; it is collapsing the cost of cognition itself. Tasks that once required scarce human expertise — analysis, reasoning, planning, synthesis — are becoming abundant and, increasingly, near-instantaneous. This is not a feature upgrade. It is a shift in the fundamental economics of decision-making.
History suggests that when a core input becomes abundant, the locus of value moves. When energy became cheap, value shifted to systems that organised production. When computation became cheap, value shifted to software and, later, to the infrastructure that made software scalable and reliable. When information became cheap, value shifted to platforms that curated, ranked, and distributed it.
Now that cognition is becoming cheap, the same pattern will repeat. The first wave builds intelligence. The second wave will build the systems that make that intelligence usable. The third wave will dominate the economic landscape.
We believe we are at the boundary between the first and second waves.
The Illusion of the Current Market
Today's AI ecosystem gives the impression that the primary opportunity lies in building better models or more refined applications. This mirrors earlier periods in technological history, where the initial excitement clustered around visible innovations: early websites in the 1990s, standalone software in the 1980s, or electrical appliances in the early 20th century.
These layers, while important, share common characteristics. They are highly competitive, relatively easy to replicate, and tend toward commoditisation over time. The long-term winners rarely emerge from this layer alone. Instead, they arise from the infrastructure that stabilises and scales the underlying technology.
Companies such as Amazon Web Services or Stripe did not invent the internet or digital payments. They built the systems that made those capabilities reliable, programmable, and accessible at scale. Their value lies not in novelty, but in necessity.
The current generation of AI companies, for the most part, are still operating in the "novelty layer." They demonstrate what is possible. They do not yet solve what will become unavoidable.
When Intelligence Becomes Abundant
For most of human history, cognition has been the primary constraint on progress. Organisations have been limited by the number of people who can think, analyse, and decide. Even the most sophisticated companies are fundamentally shaped by this constraint: decision-making is slow, expertise is scarce, and strategic exploration is limited by human bandwidth.
Artificial intelligence removes this constraint.
In an AI-native organisation, it will be possible to generate thousands of strategic options where previously there were only a handful. Marketing strategies, pricing experiments, operational optimisations, and product variations will no longer be constrained by human capacity. They will be generated continuously, in parallel, and at scale.
At first glance, this appears to be purely beneficial. But abundance has consequences.
When a resource becomes abundant, it ceases to be the bottleneck. New bottlenecks emerge in its place. In the case of AI, the abundance of cognition creates a set of entirely new constraints. Organisations will struggle not to generate ideas, but to determine which ideas are correct. They will struggle not to act, but to coordinate actions across thousands of autonomous processes. They will struggle not to produce knowledge, but to ensure that knowledge remains coherent, accurate, and useful over time.
In other words, the problem shifts from producing intelligence to managing intelligence.
The Transformation of the Firm
This shift has deeper implications for the structure of organisations themselves. The traditional firm is designed around the limitations of human cognition. Hierarchies exist to manage decision-making. Processes exist to standardise work. Planning cycles exist because analysis is slow and expensive.
As AI removes these constraints, the internal structure of the firm begins to change.
The economist Ronald Coase famously argued that firms exist because they reduce the cost of coordination compared to open markets. AI dramatically lowers these coordination costs internally. It allows organisations to operate with a level of fluidity and responsiveness that was previously impossible.
The result is that companies begin to resemble not static hierarchies, but dynamic systems. Instead of employees executing predefined roles, there are networks of human and artificial agents continuously interacting, optimising, and adapting in real time. Decisions are no longer discrete events; they are ongoing processes. Strategy is no longer periodic; it is continuous.
This transformation introduces a new kind of complexity. The organisation itself becomes a system of interacting intelligences. And like any complex system, it requires structure, constraints, and control.
The Emergence of the Control Layer
What is missing today — and what will inevitably be built — is a new layer of infrastructure designed to manage this complexity. We refer to this as the control layer of the AI economy.
This layer is not about generating intelligence. It is about making intelligence usable within real-world organisations. It encompasses the systems that determine whether AI outputs can be trusted, how multiple agents coordinate their actions, how decisions are recorded and audited, how knowledge is preserved, and how the entire system is protected from failure or manipulation.
In many ways, this layer is analogous to the institutional frameworks that underpin other complex systems. Financial markets rely on clearinghouses, regulatory bodies, and risk management systems. Software ecosystems rely on operating systems, databases, and security protocols. These layers are rarely visible to end users, but they are essential to the functioning of the system as a whole.
The same will be true for AI.
Without a control layer, organisations will not be able to safely deploy AI at scale. They will encounter failure modes that are difficult to predict and even harder to diagnose. Decisions will become opaque. Accountability will erode. Coordination will break down. Over time, this will limit the adoption and impact of AI itself.
The control layer is therefore not optional. It is inevitable.
What This Means for Investment
From an investment perspective, this shift reframes the opportunity set. The most valuable companies in the AI era will not necessarily be those that build the most advanced models or the most polished applications. They will be the companies that solve the structural problems that emerge when intelligence becomes abundant.
These problems share several important characteristics. They are universal, affecting every organisation that adopts AI. They are deeply embedded in core workflows, making them difficult to displace once adopted. They improve with scale, as more data and interactions strengthen the system. And they are often invisible until they fail, at which point they become indispensable.
This creates the conditions for highly defensible, high-margin businesses with strong network effects and significant switching costs.
We are particularly interested in companies that position themselves at critical points of control within the organisation. These are systems that sit in the path of decision-making, coordination, and execution. If they are removed, the system ceases to function effectively. This is the defining characteristic of infrastructure.
The Shape of the Emerging Companies
The companies that build this layer will not resemble traditional AI startups. They will not be defined by a single use case or department. Instead, they will operate horizontally across the organisation, integrating deeply into multiple workflows and systems.
Over time, they will accumulate a unique form of data: not just user behaviour or transactions, but the reasoning, decisions, and outcomes that define how an organisation operates. This data is inherently proprietary and difficult to replicate. It forms the basis for continuous improvement and increasing differentiation.
As these systems mature, they will begin to exhibit characteristics similar to operating systems. They will define how other tools and agents interact. They will set standards and protocols. They will become platforms upon which additional functionality is built.
At scale, they will not just support organisations. They will shape how organisations function.
Why This Moment Matters
The timing of this opportunity is critical. The underlying capabilities of AI are now sufficiently advanced to be deployed in meaningful ways within enterprises. At the same time, the infrastructure required to manage these capabilities has not yet been fully developed.
This creates a window in which foundational systems can be built and established before the market consolidates.
As organisations begin to scale their use of AI, the limitations of existing tools will become increasingly apparent. Failures in coordination, trust, and accountability will surface. Regulatory scrutiny will increase. The need for robust control systems will move from theoretical to urgent.
This is the point at which new categories are formed and category leaders emerge.
Investing in the Invisible
The defining companies of the next decade will not necessarily be the most visible. They will not always have the most recognisable consumer brands or the most immediate user engagement. Instead, they will operate beneath the surface, enabling the systems that others rely on.
They will determine which decisions are trusted, how actions are coordinated, how knowledge is preserved, and how risk is managed. They will form the backbone of the AI-driven economy.
In previous eras, these roles were played by operating systems, cloud infrastructure, and financial networks. In the AI era, a new set of control systems will emerge to perform analogous functions.
Our strategy is to identify and invest in the companies that build these systems.
Because while intelligence may become abundant, the ability to control and harness it will remain scarce. And scarcity, as always, is where value accumulates.