Artificial intelligence is rapidly moving from experimentation into production. Software systems are no longer just assisting humans. They are beginning to make decisions independently. In finance, this shift introduces a unique set of challenges because money movement is irreversible, regulated, and deeply intertwined with real-world systems.
According to fintech operator Zack Shooter, the next major bottleneck for AI adoption will not be intelligence. It will be the financial infrastructure AI systems are expected to operate on.
Shooter spent five years helping build global financial infrastructure at Deel, supporting payments, compliance, and treasury operations across more than 100 countries. That experience gave him early exposure to what happens when automation meets real-world financial rails. As AI capabilities accelerated, he began to see a deeper structural problem forming.
Financial systems, he argues, were never designed for autonomous actors.
A Financial Stack Built Across Three Eras of Technology
Modern financial infrastructure is not a single system. It is a layered stack built over decades.
Legacy banking rails still underpin global money movement. Modern fintech platforms sit on top of those rails with APIs and centralized control. Alongside them, blockchain systems introduce deterministic execution and irreversible settlement.
Each layer operates under different assumptions around trust, latency, reversibility, and human oversight.
“We are asking AI systems to move money across infrastructure that spans decades of technology paradigms,” Shooter explains. “None of it was designed with autonomous actors in mind, and none of it shares a unified control model.”
Human operators have historically absorbed the friction between these layers. AI agents will not.
Why Today’s Fintech Stack Is Not AI-Ready
Most financial systems assume humans are ultimately responsible for decisions. Approvals, exception handling, and risk reviews are built around people being in the loop.
AI systems behave very differently. They operate continuously. They act faster than humans can intervene. They make probabilistic decisions rather than deterministic ones.
“Financial infrastructure today is built around manual checkpoints and human intuition,” Shooter says. “AI does not pause for review. It acts repeatedly, and that exposes weaknesses that were previously manageable.”
In human-driven workflows, delayed reporting or inconsistent error handling can often be corrected manually. Under autonomous operation, those same gaps compound silently and at speed.
The Risk of AI-Initiated Financial Operations
Shooter expects AI agents to increasingly initiate payments, manage liquidity, route transactions across providers, reconcile balances, and interact directly with banks, payment processors, and on-chain systems.
The problem is not whether AI can do these things. It is whether existing systems can safely govern them.
Today’s financial infrastructure offers limited real-time observability, fragmented authorization models, and little ability to explain or reverse automated decisions.
“The failure modes change once software is allowed to move money on its own,” Shooter notes. “Small gaps that humans can compensate for become systemic risks when decisions happen continuously and at scale.”
Closed-Loop Agentic Payment Systems Are Only Early Experiments
Shooter points to emerging systems like X402 and other agentic payment frameworks as important signals of where the industry is heading.
These systems explore how autonomous agents can transact with one another. However, they largely operate in closed-loop environments designed for specific use cases and known counterparties.
They avoid much of the complexity involved in interacting with global banks, regulated payment providers, and legacy financial rails.
“Agentic payment systems like X402 are valuable experiments,” Shooter says. “But they exist in controlled environments. The real challenge begins when AI systems have to interact with banks, regulators, PSPs, and legacy infrastructure all at once.”
Until AI-driven payments can operate safely in open, regulated systems, the hardest problems remain unresolved.
The Identity and Accountability Gap for AI Agents
Beyond infrastructure and governance, Shooter sees a fundamental identity problem emerging.
Financial systems are built around accountable entities. Humans have legal identities. Companies have corporate identities. AI agents have neither.
Existing KYC and KYB frameworks depend on this structure. Without a clear link between an autonomous system and a responsible human or organization, those frameworks cannot function as intended.
“Every financial action ultimately needs to be attributable to someone,” Shooter explains. “Today, AI agents don’t have a clear identity, and financial systems don’t have a way to understand who they represent or who is responsible when something goes wrong.”
Shooter believes this will eventually require new attribution and delegation models that explicitly bind AI systems to accountable entities, whether through delegated authority frameworks, cryptographic credentials, or other verifiable mechanisms.
Until that link exists, autonomous financial systems will remain constrained by design.
Preparing Financial Infrastructure for Autonomous Actors
Shooter believes the next generation of financial infrastructure must treat AI agents as first-class participants rather than edge cases.
That means systems designed with:
- real-time observability
- unified governance across Web1, Web2, and Web3
- clear authorization boundaries for automation
- explicit accountability and auditability
The companies that succeed, he argues, will not be those that automate the fastest. They will be the ones that invest in the foundations required to make autonomy safe.
“AI will change how financial systems operate,” Shooter concludes. “But without the right infrastructure in place, it will expose weaknesses that have been building for years.”