The Solana Foundation, AMINA Bank, TensorX, APEX:E3 and the Cardano Foundation have jointly released the Agentic Finance Report, examining how artificial intelligence, blockchain settlement and tokenized assets are converging into a new model for institutional finance. The co-authoring group spans regulated banking, blockchain foundations and market-infrastructure firms, placing the report at the intersection of the industries that would need to interlock for agentic systems to operate. According to the report, AI agents are moving beyond advisory functions and acquiring the ability to execute financial transactions within mandates defined by people.

The report defines "Agentic Finance" as a system in which software can continuously monitor financial conditions, make decisions and execute transactions on-chain without requiring human approval for every individual action. Its authors argue this could remove operational constraints created by human involvement, including overnight processing, periodic portfolio rebalancing and batch-based collateral management. The distinguishing feature of this model is that a single system both decides and settles, closing the gap between a decision and its execution.

Three developments are presented as the foundation for this shift: AI models have become capable enough to operate within controlled financial workflows; financial assets such as stablecoins and tokenized money-market funds can increasingly be accessed programmatically; and blockchain networks provide continuous, low-cost settlement. The report identifies Solana as particularly suited to agentic applications, citing its continuous operation, low transaction costs and ability to settle transactions within the decision cycle.

The report estimates that more than $4.8 trillion in corporate treasury, wealth-management cash, asset-manager buffers and collateral could be addressable by Agentic Finance over five years. It also cites modelling by APEX:E3 suggesting that continuous intraday allocation could generate an uplift of roughly 200 basis points — two percentage points — for a $500 million treasury, while noting that these figures are model outputs rather than realised results.

Governance Remains the Critical Constraint

Despite the emphasis on autonomy, the report argues that removing people from routine execution does not remove institutional accountability. Instead, responsibility shifts toward mandates, identity systems, compliance controls and regulated financial infrastructure.

The proposed architecture separates activities according to risk. Agents could independently handle routine treasury operations and payments within predefined limits, while larger allocation decisions could require human confirmation. Strategic decisions, including mandate design and risk governance, would remain exclusively human. The report stresses that these boundaries must be technically enforceable and revocable rather than relying solely on instructions embedded in an AI prompt.

AMINA Bank's contribution identifies regulated financial institutions as a key trust layer for the emerging system. Banks can provide custody, fiat conversion, compliance monitoring, asset segregation and accountability for autonomous activity, while blockchain infrastructure supplies settlement. The report also highlights verifiable Legal Entity Identifiers — standardized identifiers already used across traditional finance to establish which legal entity stands behind a transaction — as a mechanism for linking an AI agent to the legal entity on whose authority it operates.

The report concludes that the core infrastructure for Agentic Finance is increasingly available, but adoption is likely to proceed through phased mandates rather than an immediate transition to fully autonomous finance. Institutions are advised to begin with narrowly defined use cases, such as idle-cash management, and expand an agent's authority only as governance, monitoring and risk controls are demonstrated to be effective. On that roadmap, the practical markers to watch are narrow pilots in areas such as idle cash, permission boundaries that prove enforceable and revocable in practice, and the emergence of standards tying agent activity to verifiable legal identities.