Embracing the Agentic Era with a Reliable Network

AI is now making its presence felt on the trading floor, not merely as a tool, but as a participant. These agents are capable of analyzing markets, negotiating deals, setting terms, and transferring capital across decentralized channels where settlements are final. For institutional crypto trading desks, this translates into quicker transactions, better products, and novel exposures. Imagine a scenario where two agents are negotiating a derivatives contract, but they record different figures. One agent logs $100 million, while the other logs $120 million. In the event of a discrepancy that leads to failures or investigations, who bears the responsibility? This is the harsh reality of the agentic era, where AI operates, negotiates, and acts within financial systems, and even minor discrepancies can pose systemic risks. A pressing concern is emerging: agents may act on false or unverifiable data, leading to real-world consequences. For instance, an AI system used by the UK's national healthcare provider misdiagnosed a patient, citing a non-existent 'Health Hospital' with a fake postcode. As we move beyond basic automation, it is essential that we develop systems that prioritize verifiability and accountability. Just as the web needed HTTPS, the agentic web requires a trusted network. Without a shared memory, also known as a ledger, agents may diverge, resulting in conflicting records that lead to failures. Moreover, without audit trails, they become opaque, unaccountable, and untrustworthy, rendering them unsuitable for enterprise use. This is not a distant scenario; the infrastructure gap already exists. To navigate the agentic era successfully, we need a foundation built on three core layers: Decentralized networks must serve as the cornerstone of this stack. Agents require systems that can handle thousands of transactions per second, identity frameworks that can operate across borders, and logic that enables them to collaborate and work together seamlessly, rather than merely exchanging data. To operate effectively in shared environments, agents need three key elements: In the absence of these elements, agents can behave unpredictably across disconnected systems. Since they are always operational, they must be designed to be sustainable and trustworthy. To address this challenge, enterprises must build on systems that are transparent, auditable, and resilient. Policymakers must support open-source networks as the backbone of trusted AI. Ecosystem leaders and builders must integrate trust into the foundation from the outset, rather than attempting to add it later. The agentic era will not only be automated but also negotiated, composable, and accountable – and trustworthy, if we choose to build it that way.