The Internet of Intelligence (and the dawning AI economy)
- Christopher Lyrhem
- May 21
- 7 min read
In the mid-1990s, most of us had heard about this new hot thing called ‘The Internet’ – but few truly grasped just how revolutionary it was going to be for society. Through standards like HTTP and FTP, the emergence of the Internet of Information forever altered the way we did business.
What we currently stand before is the dawn of something similar, but with a profoundly stronger future impact. With the infusion of AI and standards like MCP (Model Context Protocol) and A2A (Agent-to-Agent) into the digital fabric of society, there is a viable case to be made for the emergence of an Internet of intelligence – that in similar fashion as the worldwide web will forever alter the way we do business.
Enjoy,
Christopher Lyrhem
Chief Future Officer at Sircular
The original internet was built to move information: including documents, emails, and search results. It gave us access to knowledge, communication, and marketplaces. But the intelligence still lived in our heads. We had to read, interpret, decide, and act. Today, that model is changing.
AI is no longer confined to isolated models sitting behind web interfaces. With the rise of Model Context Protocols (MCP) and Agent-to-Agent (A2A) architectures, we’re now moving from simply sharing information to actively sharing intelligence. Similar to HTTP, these emerging standards will not only change how digital services are built, but also how organizations operate, how products are developed, and how decisions are made.
Model Context Protocol (MCP) is an open standard that defines how AI systems connect to external tools, services, and data sources in a consistent, and interoperable way. Much like HTTP standardized how websites communicated with browsers, MCP standardizes how AI models request and use context.
Where today’s AI tools require custom integrations for every external API or database, MCP allows any AI agent that “speaks MCP” to access any MCP-enabled service or dataset. This turns the AI ecosystem from a collection of walled gardens into a flexible, composable, and deeply interoperable network of intelligence.
An Internet of intelligence. In short, MCP enables AI systems to: Discover and invoke external tools using a shared language, maintain context across tools and tasks, and scale across different models and environments
Before MCP, building a powerful AI workflow meant custom-wiring APIs, adapting different data schemas, and maintaining brittle connectors. With MCP, an agent can now access tools the way a browser accesses web pages. A developer can register a service (like a financial news feed or a CRM database) behind a unified interface, and any AI agent can query it securely. This allows developers, companies, and even non-technical teams to create rich, intelligent workflows without rebuilding the plumbing each time.
Furthermore, Google’s Agent-to-Agent (A2A) protocol introduces a new programming paradigm. Instead of central systems handling everything through rigid code, we can now see a future where distributed AI agents (each with specialized functions) can collaborate, delegate, and negotiate across workflows. These agents talk to one another, operate semi-autonomously, and generate knowledge by combining different perspectives.
And this is already being explored in production environments through open-source agent stacks, research prototypes, and commercial implementations. As these two concepts, MCPs and A2A, converge, the old assumptions of what software is, how it behaves, and how it scales can be expected to be rewritten.
SOFTWARE COMPANIES WILL BE REDEFINED
For investors, this poses a major shift. Most of the software companies funded over the past 20 years followed a familiar pattern: build a vertical or horizontal SaaS product, own the data, wrap it in a cool interface, and lock users into a controlled workflow. Integration with other services often required significant effort, and intelligence was hard-coded. But the rise of MCP and A2A will tear these silos right open.
Software is no longer a fixed product; it is becoming a fluid network of context-aware, intelligent functions, stitched together through agents and protocols. What we’re seeing, is the dawn of a new internet with composable intelligence.
In practical terms, this means that many legacy SaaS products, especially those that rely on dashboards, manual data entry, or low-level reporting – will face intense pressure from agent-native alternatives. These new solutions won’t just look better – they’ll behave fundamentally differently.
Let me give you an example; let’s say a user won’t click through multiple menus to create a report anymore. Instead, they’ll instruct an agent: “Summarize our top customer churn risk signals based on the last six months”, and the agent will then autonomously gather data, call APIs, invoke MCP interfaces, and deliver a human-level summary with sources, analysis, and even suggested actions.
The implications are likely to be quite staggering. Software becomes programmable at runtime, not just by developers but by AI agents. Some interfaces will fade away, or become dynamically generated. Products will become orchestration layers between intelligent agents and standard services.
And the core value creation will likely shift from owning a dataset to understanding how to leverage context across systems. The competitive moat is no longer features, but the quality of your agent architecture, your plug-in ecosystem, and your compliance with emerging AI interoperability standards like MCP.
For venture capitalists, this requires a fundamental rethink in due diligence and valuation. Traditional metrics, like user growth, ARR, retention, will no doubt still matter – but they won’t tell the whole story.
What increasingly will matter is how “agent-compatible” a company’s product is. Does it expose MCP-compatible endpoints? Does it allow external agents to reason over its data and functions? Can its workflows be delegated to autonomous agents? And is the company building for a future where humans are no longer the primary interface?
EXPECTATIONS GOING FORWARD
In my view, over the next 12 months, we should expect early-stage startups to emerge that are natively built on top of MCP and A2A. These companies won’t build rigid software. They’ll build agent ecosystems, open toolchains, and microservices that advertise themselves in a language models can understand.
In some ways, they will resemble marketplaces of intelligence, where models and tools collaborate through protocols. These will look strange at first, barely being applications even, more like programmable agents with API backends and modular permissions.
And in the coming couple of years or so, we should expect major incumbents, like Microsoft, Salesforce, SAP – to move beyond isolated AI copilots and shift toward full AI orchestration layers, exposing their functionality through MCP-like standards. A lot of software will no longer be experienced in applications, but in conversations, workflows, and outcomes.
And this will entail new job roles emerging, such as agent architects, context engineers, orchestration designers. The integration effort that once required armies of consultants will be reduced to dynamic agent instruction pipelines.
And in the next three years (who knows), we’ll likely see the emergence of platform-agnostic AI assistants – like a browser for the “Internet of Intelligence”. These assistants, powered by your preferred model (or a mix), will be able to carry context across domains, interact with hundreds of services, and build complex chains of reasoning and execution.
Furthermore, in this future scenario – your AI assistant may begin in your email client, jump to your finance dashboard, pull data from a competitor research platform, call your CRM via MCP, and deliver a synthesized plan, without ever breaking stride. The lines between applications will dissolve. What matters will be the coherence of the agent layer and the richness of the protocol ecosystem.
THE DAWNING AI ECONOMY
When the Internet of Information materialized, because it rewrote the logic of how information travelled around the world – it completely rewired our economy. Predominantly because the marginal cost of information reached a near-zero state.
So, if the Internet of Intelligence rewrites the logic of how intelligence travel the world software and lowers the marginal cost of it to a near-zero state – this would likely also create the foundation of a new type of economy. An “AI Economy”.
This future economy will not simply be a digitized version of the one we know. It will be driven by new fundamentals – new forms of interaction, new rules of engagement, and perhaps new definitions of what counts as productive work, competitive advantage, or even a business entity.
In the AI economy, companies will not just use AI tools to become more efficient. They will be made of AI. They will be structured around autonomous agents as a core operating system, with workflows, decisions, transactions, and communications increasingly handled by intelligent systems that don’t wait for instructions. They will take action on their own.
This economy will run on agents that conduct business with each other, across systems, networks, and markets. Agents communicating with other agents. They will negotiate contracts, book services, assess compliance, compare strategies, and even price each other’s services dynamically.
Much like websites once evolved from static brochures of sorts, into fully interactive services – these agents will evolve from task executors into proactive business actors. And this changes what a “company” actually is. No doubt, strange times lie ahead...
Some business entities will still be made of people. Others will be entirely digital – agent-native “microbusinesses” (for lack of a better word) that buy and sell services autonomously. Some will be hybrid, composed of a founder, a few people, and a core team of AI agents who manage everything from finance to marketing to operations. These agents will remember everything, operate across time zones, never burn out, and continuously learn.
Furthermore, we’ll likely see new forms of entrepreneurship emerge – where starting a business is less about building an app, and more about launching and orchestrating a swarm of agents designed for a specific vertical use case. And because these agents will speak in context-aware protocols like MCP and operate through A2A architectures (and many others of course), they won’t be trapped inside one product. They will form networks, connect with other agents, and carry out cooperative strategies.
So, in this dawning new world, being first to market with a cool feature set won’t be enough. Companies will need to ask questions like: how composable is our business? How adaptable is our system to agent orchestration? Can our core operations be replicated or reassembled by a smarter, faster swarm of agents tomorrow? If not, the moat might not be as deep as it looks..
CONCLUSIONS
In this new economy, the unit of scale might not (only) be the number of users. Instead, it might (also) be the number of agents. And companies won’t grow by adding employees or writing more code, but by deploying swarms of intelligent processes that can adapt in real time. Competitive advantage will come from how well your systems reason, how quickly they connect, and how natively they participate in a global protocol-based network of intelligence.
For investors, it means rethinking what value looks like. For example, ask not only what the product does, but what its agents know, and how well they move across networks. Understand protocol readiness, agent composability, and vertical intelligence infrastructure.
The next breed of unicorns might not be "isolated" apps or other digital organism, they might instead be economies of thought, built on MCP and other standards, powered by agents, and scaled on top of the internet of intelligence.
Until next time,
Christopher Lyrhem
Comments