Summary
Citrini Research argues that agentic finance may be built on blockchain infrastructure and move crypto investing toward a more fundamental, data-driven phase. The thesis also raises unresolved questions about permissions, risk controls and accountability when software agents perform financial tasks on behalf of users.
A thesis about the next phase of crypto investing
Citrini Research’s view is that agentic finance will be built on blockchain infrastructure and could move crypto investing into a phase that is more strongly driven by fundamentals. The argument is best understood as a directional thesis rather than a report about a single product or a fully deployed market structure. The available source material does not indicate that a standardized agentic-finance system already exists, nor does it establish that the model will produce superior returns or remove investment risk.
The idea begins with a change in what financial software is expected to do. Conventional systems typically follow predefined instructions: collect market data, generate a report, submit an order or trigger an action when a condition is met. Agentic systems are designed to handle a broader sequence of tasks. Within an assigned objective and a defined set of permissions, an agent may gather information, compare assets or protocols, identify constraints, formulate a proposed course of action and coordinate subsequent steps.
Blockchain could serve as the execution and settlement layer for that workflow. Instead of an agent merely producing an analysis for a human operator, it could interact with on-chain accounts, smart contracts and data systems. Rules around ownership, authorization and transfer could be expressed in programmable form and recorded in a way that supports later review. That is the infrastructure logic behind Citrini Research’s thesis.
It is also where the distinction between a useful framework and a proven outcome becomes important. More automation does not by itself create better analysis. An agent may process information faster while still relying on incomplete data, inconsistent definitions or flawed assumptions. The investment implications therefore depend not only on what an agent can do, but also on how its information sources, permissions and decisions are constrained.
What a fundamentals-driven market would require
Crypto markets have often been shaped by liquidity conditions, narratives, social sentiment and short-term flows. The relationship between an asset’s price and the underlying use of a network or protocol is not always clear. If agentic finance is to support a more fundamentals-driven approach, its main contribution would not simply be faster order execution. It would need to help participants continuously collect, compare and interpret complex information.
Depending on the asset, potentially relevant indicators could include protocol usage, fee generation, collateral composition, governance arrangements, contract upgrade history and exposure to operational or market risks. For tokenized real-world assets, the analytical picture could also include information about the underlying asset, legal claims, servicing arrangements and reporting quality. An agent may help organize these materials and apply a consistent framework, but it cannot automatically make the data reliable or the comparisons valid.
That limitation matters because “fundamental” is not a single, universally agreed metric in digital assets. A protocol’s fee income may be affected by temporary incentives. Usage can rise without creating sustainable economics. Governance can appear active while decision-making remains concentrated. On-chain activity can be visible but difficult to interpret without off-chain context. A system that summarizes these indicators without explaining their limitations could create an impression of precision that the underlying evidence does not support.
Human judgment therefore remains part of the proposed model. Users and institutions would need to understand which data an agent considered, what assumptions it applied and how it handled uncertainty. If the agent is unable to establish the reliability of a source, that uncertainty should be surfaced rather than hidden behind an automated conclusion. Otherwise, automation could amplify analytical errors instead of reducing them.
Why blockchain matters—and where it stops
Blockchain is attractive to agentic-finance designs for several reasons. It can provide a shared environment for accounts, asset representations, permissions and settlement. Smart contracts can encode certain conditions, while transaction histories can support monitoring and reconstruction of what happened after the fact. For agents operating across multiple financial workflows, standardized on-chain interfaces could reduce some forms of manual coordination.
Those benefits have clear boundaries. A blockchain record can show that a transaction occurred, but it does not independently prove that the input data was accurate, that an oracle functioned correctly, or that a contract’s logic matched the user’s intent. It also cannot eliminate risks arising from stolen keys, malicious contracts, prompt injection, mistaken authorization or insufficient liquidity. If an agent is allowed to call several protocols in sequence, vulnerabilities in one part of the workflow may affect the rest.
As a result, the relevant infrastructure is broader than a model and a blockchain. It includes identity, key management, permissioning, data validation, transaction simulation, monitoring and incident response. An agent that can explain a decision but cannot be restricted may be unsuitable for high-impact financial operations. An agent that can execute transactions without producing a reliable audit trail may be difficult to reconcile with institutional governance.
Implications for institutional wallets and custody
As agents begin to act on behalf of institutions, a wallet may become more than a place where assets are held. It may also function as the boundary that defines what an agent is allowed to observe, propose or execute. A robust design would distinguish between read-only analysis, transaction preparation and final execution. Institutions may also require transaction limits, approved contract lists, time delays, separation of duties and human approval for specified actions.
Custody and wallet operators would need to address several basic governance questions. Who grants an agent its authority? How can that authority be revoked? What records show the instructions, data and approvals behind an action? Who is responsible when a model, data source or policy changes? How should an institution respond when an agent behaves outside its expected pattern?
These questions are relevant to the broader institutional-wallet and digital-asset-custody sector, including providers such as Cobo, but they do not imply that any particular provider has resolved all of the challenges associated with agentic finance. Automation can reduce operational friction while increasing the potential impact of a single configuration error or compromised credential. Institutional assessment therefore has to cover governance and security controls as well as technical capability.
Regulation and accountability remain unresolved
Agentic finance may also complicate existing questions about financial responsibility. If an agent makes a decision using incomplete or misleading information, responsibility could potentially involve the user, developer, operator, data provider or institution that granted authority. When an agent interacts with multiple protocols, it may be difficult to determine how liability should be allocated after a loss. The answers may differ across jurisdictions.
Disclosure and oversight could become important as well. Users may need to know what data an agent used, which constraints applied and when a human was required to intervene. Workflows involving customer assets, payments, certain digital-asset payment instruments or tokenized assets could raise additional issues related to customer identification, anti-money-laundering controls, consumer protection and asset segregation. Blockchain’s cross-border operating model may also sit uneasily with rules that are enforced on a jurisdiction-by-jurisdiction basis.
The legal status of an automated recommendation may differ from that of an automated transaction. So may the responsibilities of a service that supplies infrastructure versus one that makes a decision or controls an account. These distinctions will matter if agentic finance moves from experimentation into regulated financial activity.
From concept to dependable infrastructure
Citrini Research’s thesis offers a useful way to connect two developments: artificial intelligence agents that can manage multi-step tasks, and blockchains that provide programmable assets and settlement. It also suggests why the next debate may focus less on whether agents can automate execution and more on whether they can support disciplined, auditable fundamental analysis.
The distance between that concept and broad adoption remains significant. Progress would require better data standards, clear authority models, stronger security controls, reliable monitoring and defined responsibility when something goes wrong. It would also require users and institutions to treat automated output as conditional analysis rather than unquestionable judgment.
Agentic finance may eventually change how crypto investment processes are organized, but its success cannot be measured only by the number of tasks an agent can complete. The more consequential test will be whether these systems can operate within transparent, reviewable and enforceable boundaries. Until those conditions are established, the idea remains a significant infrastructure and governance direction—not a confirmed replacement for human oversight or a guarantee of fundamentals-driven market behavior.
Source: link