A major financial institution just published research suggesting that AI is preparing to do something far more significant than summarize earnings reports or screen for trade ideas.
The research describes a future in which AI systems act inside financial markets independently, without waiting for a human to decide.
Fidelity Digital Assets argues in its latest research that AI agents could become an entirely new class of participants in financial markets, executing trades, arranging loans, managing portfolios, and processing payments with limited human direction.
The report is careful to add something the broader AI-in-finance conversation often skips: More activity does not automatically mean more value for the systems supporting it.
That distinction could define which parts of the financial industry actually benefit from the shift.
What Fidelity’s research says about AI agents in financial markets
The core argument in the Fidelity research is that AI agents will not just assist financial professionals. They will begin operating as participants in their own right, processing market information continuously and acting on it without stopping to check in with a human at each step.
Fidelity identifies several areas where this shift could have the most immediate impact. Trading, lending, and portfolio management all involve high volumes of decision-making, access to large pools of capital, and the generation of significant fees.
Payments, while potentially the highest-volume activity, produce narrower margins and may generate less lasting economic value for the infrastructure supporting them.
Related: Google DeepMind prepares for risk of AI agents going rogue
The report also raises a less comfortable question for investors watching the AI-in-finance space closely.
If AI agents can shift between financial platforms quickly, optimizing for cost and execution quality at each step, then higher overall activity may not translate into durable value for any single platform or institution.
The competitive advantage could belong to the platforms that give AI agents the most reliable access to liquidity, data, and settlement, not necessarily the ones generating the most raw volume.
Why AI agents in trading and lending could reshape capital management
Fidelity places AI-driven trading, lending, and portfolio management ahead of payments in terms of economic impact. That sequencing reflects where AI already has measurable advantages.
Evaluating risk, synthesizing market signals, identifying pricing inefficiencies, and rebalancing positions across multiple assets are areas where AI systems can operate at a speed and scale that no human team can match.
Logan Xie, leader of KuCoin AI Lab, told TheStreet in an interview that the most significant opportunity is a structural one. “The greatest near-term value will come from AI turning capital from something that is periodically allocated into something that can continuously interpret markets, manage risk, and act within defined mandates.”
That shift would be significant for institutional and retail investors alike.
A hedge fund running an AI agent on its long-short book does not pause for a bank holiday. A fixed-income desk using an AI system to evaluate duration risk does not take a week off between reviews. Capital that currently sits idle between investment decisions could instead be continuously managed within a framework of predefined objectives.
The portfolio manager is not replaced by a single AI decision. It is replaced by an ongoing process that never pauses for weekends, holidays, or human availability.
The lending market offers a parallel opportunity. AI agents could continuously evaluate creditworthiness, match borrowers to lenders, price risk in real time, and adjust loan terms based on changing conditions.
For financial institutions, that could mean faster deployment of capital, lower defaults from better risk assessment, and reduced operational costs from fewer manual reviews.
Cravetiger/ Getty Images
Why more AI activity in finance does not automatically mean more value
Fidelity’s warning is about the gap between activity and value. More transactions do not automatically mean more lasting economic benefit for the financial infrastructure behind them. This is the part of the AI-in-finance thesis that receives the least attention.
GoMining CEO Mark Zalan told TheStreet the distinction matters. “What actually accrues to a network is settlement demand, and the thing that generates settlement demand at a scale nobody has seen before is machines paying machines.”
That points to a category of financial activity the current system was not built to handle. AI agents transact with one another, buying computing resources, data access, and services from other automated systems in amounts too small for traditional banking infrastructure to process.
Coinbase, Stripe, and Visa are all actively building infrastructure for machine-to-machine payments, Seeking Alpha reported.
These transactions do not appear in standard models of trading or lending volumes. They represent a different kind of demand altogether.
“The asset that ends up doing machine settlement is the one that wins this era, and right now that layer gets far less attention than the trading story,” Zalan added.
Fidelity identifies a similar dynamic in its analysis of what happens when AI makes financial software faster and cheaper to build.
If AI lowers the cost of replicating the technology behind financial platforms, the technology itself stops being the competitive advantage. What remains is the accumulated weight of real economic activity: the users, the liquidity, and the trust built over time.
Scott Dykstra, co-founder of Space and Time, told TheStreet that verification becomes the critical issue as AI takes on more financial responsibility. “AI agents need trustworthy inputs and auditable execution.”
That requirement extends to the institutions deploying autonomous systems. If an AI agent makes a bad lending decision or misreads a market signal, the financial institution behind it needs a record of what the system was told, the data it used, and the logic it followed.
Without that trail, the accountability question is unanswerable.
Related: AMD CEO doubles down on AI and the stock market
“Code can be copied quickly, but liquidity, users, and established economic activity are much harder to recreate,” Dykstra added.
An autonomous system making lending or investment decisions needs reliable data. The financial institutions and customers it serves need ways to confirm that the agent followed the strategy it was supposed to follow.
As automation increases, the systems that verify what happened inside those decisions could become as important as the systems carrying out the decisions themselves.
What the shift to autonomous AI means for financial infrastructure
The Fidelity research points to a broader transformation in what financial infrastructure will need to look like when the customer is increasingly not a person. AI agents require identities, permissions, access to assets, and mechanisms to audit their actions. Those are governance and infrastructure problems, not just technology problems.
For banks, asset managers and financial technology firms, the question is whether they build infrastructure suited for human customers and then adapt it, or design systems from the start around the idea that the primary user may be automated.
The firms that solve that problem earliest could gain access to a category of financial activity that did not previously exist.
The same dynamic applies to asset managers and payments providers. Institutions that adapt their infrastructure to serve AI agents as clients, rather than treating automation as an internal efficiency tool, open themselves to a new category of revenue that does not require acquiring a single additional human customer.
Fidelity is specific about one potential outcome. AI could accelerate the development of financial technology broadly, lowering the cost of building platforms and enabling new entrants to compete, Benzinga reported.
That competition would benefit AI agents optimizing for cost and performance. But it could also concentrate activity on the platforms with the deepest liquidity, most reliable execution, and strongest existing user base, because those are the factors AI systems will optimize toward when choosing where to transact.
Xie put it plainly: “AI will commoditize code, but it will not commoditize network effects. The networks that have accumulated deep liquidity, strong user trust, and regulatory clarity will become even more valuable as AI floods the market with new entrants.”
The financial institutions that understand that distinction and build around it may be the ones best positioned when AI agents become active participants in markets.
Fidelity’s warning is not that AI will disrupt finance. It is that much of the disruption may end up benefiting infrastructure players nobody is currently watching closely.
Related: Microsoft makes a controversial decision that changes its AI story
