For much of the internet’s history, technology has supported human transactions. Increasingly, it’s executing them directly.
That shift is the focus of this episode of the FCAT Crypto Brief, which examines agentic payments: transactions executed by autonomous AI systems. Based on pre-programmed directives, these agents discover services, access resources, and settle payments without hands-on human involvement.
This episode’s discussion centers on what becomes possible when technology participates in the economy as an actor, and what new obligations emerge for the people designing—and overseeing—those systems.
Technology with agency
Many common AI interactions remain query-based and reactive. Agentic systems operate‑ differently. They are purpose-built to pursue goals across multiple steps—identifying tools, calling APIs, consuming data, and paying for what they use along the way.
“We’re at an inflection point where we’re seeing the agent as the customer. Not the developer. Not the human.” — Parth Gargava, Managing Director, Fidelity Labs
When agents transact, they require access to credentials, spending authority, and settlement mechanisms. They also tend to introduce scale, as one agent becomes many, operating continuously rather than intermittently.
Most current implementations still route these activities through human-linked‑ payment methods, such as credit cards, or leverage preloaded usage credits. Those approaches support early experimentation, but they become operationally complex and potentially risky as agent activity expands. Managing permissions, limits, and reconciliation across many autonomous processes places stress on systems designed around individual users. Additionally, prioritizing outcomes may sideline other factors.
Where blockchain payments enter the picture
Blockchains were designed to move value directly between participants, with settlement embedded within the transaction itself. In some cases, that characteristic maps naturally to agent activity, where authorization and execution occur programmatically.
So far, this has surfaced in several practical patterns, including:
- Agents paying for compute or data access at the moment it’s consumed
- Session-based‑ settlement that aggregates many small interactions
- Delegated accounts allowing agents to operate independently within defined rules
Stable-value payment instruments are often discussed in this context because they allow these interactions to occur without constant repricing. The emphasis remains‑ on programmability and policy enforcement rather than on price exposure or trading activity.
Efficiency introduces new risks
Autonomous spending can change the pace at which transactions are made, but they can also accelerate the speed at which costs accumulate. Agents optimize for task completion, not budget sensitivity. In these instances, small experiments can lead to unexpectedly high usage charges once spending was automated.
However, this behavior is somewhat predictable given how agents are designed, highlighting the importance of embedding constraints in advance. This might include:
- Spending ceilings tied to purpose
- Time‑bound authorizations
- Clear ownership over agent behavior and outcomes
In short, effective controls help establish the guardrails that allow autonomous systems to operate reliably as they scale.
Credentials, governance & responsibility
Agentic payments require mechanisms to express authority clearly. Delegated wallets and sub‑accounts provide one approach, allowing a primary account to define policy while agents execute within it. Activity can be monitored and audited directly through transaction records.
In all cases, accountability remains a central consideration. As non‑human actors participate directly in financial flows, questions of responsibility, liability, and oversight remain somewhat unanswered, with compliance and governance frameworks ever-evolving alongside the technology itself.
Why this matters more broadly
Many people encounter digital assets through foundational questions about access, custody, or taxation. Agentic payments sit upstream from those concerns, and they speak to how transaction systems behave when technology initiates activity on its own.
As autonomous systems take on tasks like procurement and resource allocation, payments begin to shift from a back‑office function to an architectural strategy, and this ongoing conversation provides a lens into how assumptions about control, trust, and responsibility change once transactions no longer hinge on human approval.
Want to learn more? Tune in to the latest episode of the FCAT Crypto Brief.
Digital assets are speculative and highly volatile, can become illiquid at any time, and are only for those investors willing to risk losing some or all of their investment and who have the experience and ability to evaluate the risks and merits of an investment.
This information is for informational purposes only and is not intended to provide investment or any other advice and should not be construed as an offer to sell, a solicitation of an offer to buy, or a recommendation for any security or other assets. The opinions provided are those of the authors and not necessarily those of the Fidelity Center for Applied Technology (FCAT) or its affiliates. FCAT does not assume any duty to update any of the information. FCAT and any other third parties are independent entities and not affiliated. Mentioning them does not suggest a recommendation or endorsement by FCAT. Third-party trademarks and service marks are the property of their respective owners. All other trademarks and service marks are the property of FMR LLC or its affiliated companies.
© 2026 FMR LLC. All Rights Reserved. 1040156.117.0
