Originally published at Forbes
Many used to think the drive of fintech, in the payments space at least, was simply to create frictionless payments: from chip-and-pin to mobile wallets, contactless to embedded finance. But the next phase is bigger. In addition to removing friction, we’re now addressing who makes the payments.
We’re entering the era of agentic AI: intelligent, autonomous agents that don’t just assist with payments, but initiate and complete them on our behalf. These AI agents can sense a need, decide the best course of action and act—without waiting for a human to click, tap or approve.
We’re on the brink of a leap that may transform how businesses operate, how consumers manage money and how the financial system itself functions.
If done right, it’s a recipe for increasing productivity, efficiency and reducing costs. But if it isn’t, all hell will break loose.
What Agentic AI Is—And Why We Should Use It
Agentic AI are simply software agents that can perceive, reason and act independently, within parameters humans set. They don’t just follow rules—they decide. That’s a profound difference.
For example, imagine a household personal AI assistant that knows your utility providers, budget, usage patterns and payment deadlines and pays providers on the right day, applying loyalty discounts and flagging cheaper alternatives. Or consider a smart factory purchasing AI agent that senses dwindling stock levels, negotiates with pre-vetted suppliers, verifies delivery via IoT sensors and triggers payments via smart contracts. No purchase order, no finance team, no friction.
Of course, we’ve already seen some of this with walk-in/walk-out retail stores, for example. Agentic AI just completes the journey: It removes the human trigger entirely for routine transactions.
And why? Because it offers speed, scale and efficiency and makes for a more enjoyable, frictionless experience, for vendors and customers alike.
What’s Making It Possible
For agentic payments to work, we need more than just smarter AI. We need the financial infrastructure to support fast, secure, nonhuman transactions. And the good news? It’s falling into place.
• Embedded Finance And APIs: The proliferation of banking-as-a-service platforms means software agents can now initiate payments, check balances or execute transfers via secure API calls. By 2030, embedded finance is expected to be a $7.2 trillion market.
• Real-Time Settlement Networks: In 2023, 266 billion real-time payments were made globally—up 42% year over year. Whether through SEPA Instant Credit Transfer, FedNow or stablecoins, AI agents need near-instant rails—and those rails are being laid.
• Multi-Party Computation (MPC): This cryptographic technique lets payment approval be split across devices or servers. It ensures that even if an AI agent initiates a transaction, no single point of compromise can execute it alone.
• Smart Contracts And Programmable Money: Funds can now be conditional and coded, releasing only when certain criteria are met. This gives agents rules to follow, enforcing safety and intent through code.
None of these technologies is perfect yet. But together, they make autonomous payments not just plausible, but increasingly inevitable.
The Barriers: Trust, Liability And The Regulatory Vacuum
As always, the tech is ahead of the regulations. And this time, the gap matters.
If an AI agent pays the wrong vendor, who’s liable: the user, the platform or the agent’s creator? If a rogue bot makes repeat transactions, who intervenes? And how do we prove that a given payment was initiated by a legitimate, authorized AI—not a hacked script?
Regulators are grappling with these questions, but responses vary wildly. The EU has gone all-in with the AI Act, imposing strict controls on how agentic AI can operate. Its Article 5 bans several high-risk AI practices including:
• Subliminal techniques
• Exploiting vulnerabilities (age, disability, socioeconomic status)
• Social scoring
• Biometric-based profiling
Fines for violations can reach 35 million euros or 7% of turnover—a clear signal that Europe is drawing a hard line.
In contrast, the U.K. and U.S. have opted for what’s called a principles-based approach, effectively delegating oversight to sector regulators and encouraging voluntary compliance. It’s light-touch to the point of invisibility, and while that may appeal to innovators, it also creates blind spots. If no one’s watching, how will they spot abuse?
To me, the best path lies between these extremes. We need rules that allow innovation—but not without oversight, transparency and redress.
What Comes Next
Agentic AI won’t arrive all at once. It will creep in through niches: AI financial advisors that move your money automatically. Subscription managers that cancel services you don’t use. Virtual CFOs for SMEs. Procurement bots in supply chains. And yes, machine-to-machine payments: cars paying for fuel, drones paying for airspace and solar panels selling energy to neighbors.
The challenge is to shape this future. The shift is already underway—not from frictionless to faster, but from triggered to autonomous. And when machines start moving money on our behalf, the systems they rely on and the safeguards they follow will matter more than ever.
Comments are closed.