We are witnessing an explosion of AI agent platforms, each promising to revolutionise how businesses operate. But there is a fundamental problem: we are forcing these AI agents to interact through infrastructure designed for humans.
It is like we have invented cars but are still making them drive on dirt paths meant for horses.
Consider what happens when an AI agent needs to book a dental appointment. Today, it must navigate websites, parse business listings, potentially make phone calls or async emails, and handle payment systems all designed for human interaction. Instead, we need a protocol where AI agents can discover local dental practices, verify their availability, negotiate appointment times, and execute bookings and payments directly at machine speed.
Most concerning is how platforms create isolated agent ecosystems, like having phones that can only call other phones from the same manufacturer.
The fundamental mismatch
The most impactful mismatch is speed. AI agents can process information in microseconds, but our infrastructure forces them to operate at human timescales. In our dental example, an agent could evaluate all possible appointment times across multiple practices instantly, but instead must wait for each practice's system to respond individually. It is like having a supersonic jet but forcing it to follow traffic lights.
When thousands of AI agents are making millions of decisions, human centric security measures become massive bottlenecks. We need cryptographic security that operates at machine speed, allowing agents to verify and trust each other instantly. The volume of interactions compounds this problem. AI agents can participate in thousands of parallel negotiations, but current infrastructure was not built for this scale.
These mismatches are fundamental barriers holding back autonomous AI agents. To move forward, we need infrastructure that matches how AI agents actually operate, with protocols designed for machine speed interaction and standardised ways for agents to communicate regardless of their implementation.
Key requirements for AI-native infrastructure
What we need is infrastructure designed from the ground up for AI agent interactions. This means moving beyond traditional request response cycles to a fluid architecture where agents communicate asynchronously and in parallel. The protocol must handle millions of simultaneous interactions without bottlenecks, similar to how financial markets handle millions of trades per second.
Identity and verification must be built into the protocol layer itself, with every agent having a cryptographic identity that can be instantly verified without central authorities. For instance, a dental practice's agent could instantly verify a patient's insurance coverage and payment capability without manual verification steps.
These foundational requirements enable something far more transformative.
The power of a universal protocol
By establishing these requirements in a universal protocol, we create a fundamental communication layer that any agent can use regardless of implementation. Like email protocols allowing messages between different providers, this enables innovation across the entire ecosystem. Various platforms could implement the protocol differently, optimising for specific use cases like high frequency trading or supply chain coordination, while maintaining seamless interaction between all agents.
The network effects would transform how business operates. Businesses could have their agents continuously scanning for opportunities and executing transactions at machine speed. As more businesses join, each new agent adds capabilities to the network, creating more opportunities for collaboration. What once took days of phone calls and emails to book a dental appointment could happen in milliseconds through agent-to-agent communication. Just as the internet enabled ecommerce and social media, a universal agent protocol will enable business models we have not yet imagined.
This is where we are headed: a future where AI agents operate at their full potential, making and executing decisions at machine speed whilst maintaining security and trust.