Every meaningful technological shift follows a familiar pattern. We build tools that take repetitive work off people’s hands, and in doing so we give ourselves more time to focus on judgment, relationships and decisions that actually move things forward. That has been true from industrial machinery to cloud software, from automation to AI, and I think it is exactly what we are beginning to see again with agentic AI.
Unlike earlier generations of AI that were designed to answer questions or make recommendations, AI agents are built to take action. Within clearly defined permissions they can plan, use software, interact with other systems, complete tasks and report the outcome, which means technology starts becoming something we delegate work to rather than simply something we use.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
Capital markets ripe for innovation
I believe capital markets are one of the places where this shift will become most visible, largely because the industry has already spent decades digitising itself without ever fully connecting the underlying infrastructure.
Today’s financial markets are powered by sophisticated technology, but behind almost every transaction sits a surprisingly fragmented process involving separate systems for issuance, compliance, investor onboarding, custody, settlement, reporting and asset servicing.
Every participant maintains their own records, verifies much of the same information independently and spends a considerable amount of time reconciling data across institutions. We successfully digitised the individual components of the market, but we never really created a shared operating environment where those components could work together seamlessly.
Machine learning has undoubtedly improved this picture by helping institutions detect fraud, model risk and automate countless operational decisions, but those systems have largely remained inside organisational boundaries. They can analyse information extremely well, yet they still struggle to coordinate activity across the broader financial ecosystem because the infrastructure itself remains fragmented.
An AI agent is capable of moving across workflows rather than simply analysing them, but intelligence alone is not enough. Before an agent can execute financial activity with any degree of autonomy, it needs confidence that the information it relies upon is accurate, current and trusted. It needs to understand ownership, permissions, settlement and authority without constantly depending on manual intervention or disconnected databases.
Tokenisation full potential yet to be realised
For years, tokenisation was framed primarily as a way to democratise finance, but I have always believed its greater contribution lies much deeper within market infrastructure. Turning an asset into a token does not automatically create liquidity or investor demand, but it does create a programmable environment where ownership, compliance rules and transfer conditions can exist within a shared system that every authorised participant can rely on.
That shared infrastructure matters because AI agents perform best when they are operating against a common source of truth. If every institution maintains a different version of the same transaction, the agent inherits the same inefficiencies that exist today, spending its time retrieving information, reconciling records and waiting for approvals instead of executing work efficiently.
Blockchain changes that dynamic by giving agents a verifiable view of ownership, permissions and transaction history while smart contracts establish clear execution rules, allowing assets and increasingly digital forms of money to move together through programmable workflows. None of this requires confidential information to live publicly on-chain, but it does allow participants to verify what matters without repeatedly recreating trust from scratch.
AI & blockchain: solving very different problems
This is also why I don’t see AI and blockchain as competing technologies because they solve very different problems. Blockchain establishes trust, ownership and the movement of value, while AI provides the reasoning, coordination and operational intelligence that allows increasingly complex financial processes to happen with far less manual involvement. Neither technology reaches its full potential without the other.
An AI model can already read a prospectus or a legal agreement, but reading information is very different from knowing whether it is authoritative, who issued it, what rights it represents, whether it remains valid and what actions can legally be taken as a result. For agentic capital markets to become practical, assets themselves need to become machine-readable so that intelligent systems can interpret structured rights, permissions, ownership and eligibility with confidence rather than inference.
We are already seeing the foundations emerge through new standards that help AI systems connect with external tools, communicate with one another and exchange value programmatically, alongside blockchain standards that are evolving around identity, compliance and standardised digital asset frameworks. Like most infrastructure, these developments may appear highly technical today, but history suggests the technologies that reshape industries are usually the ones that become almost invisible over time.
Agents to enable humans to focus on judgement, value creation
I do not believe agentic AI will replace human judgment in capital markets because investment decisions, regulation and fiduciary responsibility will always require accountability. Instead, what I do think is that intelligent agents will steadily take responsibility for much of the operational work that surrounds financial products, allowing people to spend less time coordinating fragmented systems and far more time exercising judgement where it creates the greatest value.
Capital markets have spent decades becoming digital while the next chapter is about becoming coordinated, and I believe that will happen through the convergence of machine-readable assets, programmable settlement and delegated intelligence. When those three pieces come together, agentic capital markets will no longer be a future concept.
