The Tirole Theory That Became the Foundation of Platform Strategy®
For the past quarter century, the greatest source of competitive advantage in the digital economy has been the network effect.
The more users a service attracts, the more valuable it becomes. That increased value, in turn, attracts even more users. This self-reinforcing mechanism has been the driving force behind the rise of dominant companies across digital markets, including search engines, social networks, e-commerce platforms, ride-hailing services, and online travel booking services.
The theory that explained this phenomenon was the theory of Two-Sided Markets, developed by economist Jean Tirole and his collaborator Jean-Charles Rochet. Tirole later received the 2014 Nobel Prize in Economic Sciences for his contributions to the analysis of market power and regulation.
They defined a platform as an intermediary that connects distinct groups of customers.
Consider a shopping mall. The more visitors it attracts, the more retailers want to open stores there. The more retailers join, the more attractive the mall becomes to shoppers. The same dynamic applies to online marketplaces. More buyers attract more sellers, and more sellers attract more buyers.
One of the key insights of Tirole and Rochet was that these two sides of the market are mutually dependent. A platform can stimulate growth by charging one side little or nothing while monetizing the other side. In practice, many digital companies achieved explosive growth through this asymmetric pricing strategy.
As platforms expand, they eventually reach what is known as a critical mass. This is not merely a break-even point. Rather, it is the threshold at which network effects become self-sustaining and self-reinforcing.
Once this point is reached, users attract more users, competitive advantages strengthen rapidly, and markets often evolve toward a Winner-Take-All structure.
As a result, network effects became one of the most important concepts in modern business strategy. Executives focused on growing membership numbers and monthly active users, while investors viewed the strength of network effects as a key driver of corporate value.
Today, however, the emergence of AI agents is beginning to challenge this fundamental assumption.
What AI Agents Change
The important point is not that network effects will disappear. They will continue to exist. What changes is where those network effects emerge.
In the traditional internet era, humans visited platforms directly. Consumers used search engines, compared products across e-commerce sites, and opened applications to choose services. In other words, platforms controlled the customer interface.
As AI agents become increasingly capable, this structure begins to change.
Instead of manually searching and comparing options, a user may simply instruct an AI agent:
“Book the best hotel for my Kyoto business trip next week.”
The AI evaluates prices, locations, reviews, past travel history, schedules, and personal preferences before presenting the optimal recommendation.
In this scenario, the customer is no longer interacting with a hotel booking platform. The customer is interacting with an AI agent.
Control of the customer relationship shifts from the platform to the agent.
This does not invalidate the Rochet–Tirole framework. However, it may alter one of its fundamental assumptions. Their theory assumes that consumers and suppliers meet through a platform.
AI agents change that equation.
They can compare multiple platforms and services simultaneously and make decisions on behalf of users.
As a result, the importance of gathering users onto a single platform diminishes.
Network effects do not disappear, but the center of gravity shifts—from user-centric networks to AI-agent-centric networks.
From Economies of Scale to the Economics of Inference
This transition also changes the nature of competition itself.
Historically, companies competed on scale. Membership counts, engagement time, downloads, and transaction volume were the primary drivers of corporate value.
In the AI era, scale alone may no longer be sufficient.
AI does not care who is popular. It cares who delivers the best value.
When an AI agent searches for products, for example, it evaluates price, quality, inventory availability, delivery speed, and review reliability. Advertising budgets and brand image become less influential.
Humans are influenced by advertising.
AI is influenced by data.
This distinction is profound.
As a result, the basis of competition shifts from user volume to inference quality.
Competitive advantage in the AI era will be built through a new feedback loop:
Customer Understanding → Better Inference → Higher Recommendation Quality → Greater Trust → Deeper Customer Understanding
I call this the Economics of Inference.
If network effects represented the economics of scale, then AI represents the economics of inference.
The Innovator’s Dilemma Returns
Perhaps even more interestingly, this transformation may recreate the conditions described in Clayton Christensen’s famous theory of the Innovator’s Dilemma.
History shows that incumbent firms rarely fail because they lack technological capability. More often, they fail because optimizing for existing customers prevents them from adapting to emerging markets.
Mainframe manufacturers underestimated personal computers.
Mobile phone makers reacted too slowly to smartphones.
Video rental chains overlooked the potential of streaming services.
In each case, management decisions were rational from the perspective of existing customers. Yet that very rationality delayed adaptation to disruptive change.
The same danger now confronts today’s platform giants.
Companies serving hundreds of millions—or even billions—of users devote enormous resources to improving existing user interfaces, maximizing advertising revenue, maintaining loyalty programs, and increasing time spent within their applications.
Yet in the age of AI agents, the critical function is not maximizing engagement. It is decision-making on behalf of users.
The key metric shifts from screen time to API connectivity and MCP integration. The objective shifts from ad impressions to inference quality.
MCP (Model Context Protocol) is an emerging open standard that enables AI models to connect securely with external systems and data sources. It is often described as the “USB-C of AI.” Once an organization creates an MCP server, multiple AI systems can access enterprise data and tools through a common interface.
The challenge is that successful companies often struggle to disrupt their own business models.
This is the essence of the Innovator’s Dilemma.
The greater the success, the more difficult adaptation becomes.
The Limits of Lock-In Strategies
In this environment, traditional lock-in strategies must be reconsidered.
Loyalty points, membership tiers, proprietary payment systems, and exclusive member benefits have long served as powerful mechanisms for reducing customer churn.
AI agents, however, operate through calculation rather than emotion.
Even if one company offers generous reward points, an AI agent may switch providers if the total cost elsewhere is lower.
A famous brand will not be selected if quality is inferior.
Human users may be creatures of habit.
AI agents are not.
Of course, switching costs will not disappear entirely. Learning histories, personal data, identity systems, and authentication infrastructures will remain important assets.
Nevertheless, their effectiveness is likely to weaken.
Companies must shift their thinking from:
“How do we lock customers in?”
to
“How do we ensure AI agents continue choosing us?”
A Warning for Business Leaders
The meaning of network effects is changing dramatically.
The traditional formula for success was simple:
Attract users.
The new formula is different:
Be selected by AI agents.
In this new environment, open connectivity becomes more important than closed ecosystems. Interoperability becomes more important than lock-in. Inference quality becomes more important than membership size.
The Two-Sided Market theory developed by Professor Tirole provided a powerful framework for understanding the platform era. Yet the age of AI agents requires us to build an additional layer on top of that framework.
We are standing at a historic transition point—from the Age of Network Effects to the Age of Inference Effects.
In the next chapter, we will explore how companies can build sustainable competitive advantages in this new environment through what I call the First Connection Strategy.
Platform Strategy® is a registered trademark of NetStrategy Inc.
