Abstract
Operability — not capability — is what has held tokenized finance back. As financial systems move from static to programmable instruments and from isolated platforms to composable markets, safely overseeing a tokenized workflow outgrows what a human can realistically manage, while today's AI agents and real-world financial infrastructure suffer a structural mismatch: agents work best in structured environments, but even mature tokenized systems are fragmented, brittle, and opaque. This IEI note argues the answer lies in operationalizing context into structured inputs for AI agents and constraining those agents to behave predictably — in ways that remain verifiable, auditable, and interpretable to humans — and previews the Institute's 2026 research agenda on the problem.
Executive summary
The Intelligence Economy Institute has long championed functional scalability through modularity — making financial systems programmable and composable at the architectural level. But a second challenge sits at the operational level: as instruments become programmable and platforms become composable markets, the technical and financial knowledge required to safely oversee a tokenized workflow quickly exceeds what any human agent can muster.
AI is a hopeful answer — but today's AI agents and real-world financial infrastructure are fundamentally mismatched. Agents behave unpredictably unless constrained, yet failure is not an option for financial institutions; and while agents work best in structured environments, even mature tokenized systems carry fragmented data models, brittle standards, opaque smart-contract code, and ad hoc DeFi/TradFi integrations.
We argue the resolution lies in operationalizing context — protocols, code, environments — into structured inputs for AI agents, and in constraining those agents so they behave predictably. Crucially, that operationalization and those constraints must remain, at all times, verifiable, auditable, and interpretable to humans — applied recursively, starting from a core people can understand and audit.
Resolving this clears the path for tokenized finance to scale: agents that can design, operate, and oversee complete workflows with precision and efficiency, without sacrificing accountability. The Institute is making this a research priority for 2026, and an early report — with detailed contexts and initial study results, developed with financial-institution partners — is expected in Q1 2026.