New York correspondent Amrit Kang examines the rise of agentic finance in the US, as AI agents increasingly execute transactions and drive a shift towards machine-native, autonomous financial systems.
This month has been a whirlwind in the U.S. fintech ecosystem we’ve had it all.
Stripe Sessions sets the tone for an agentic economy
At Stripe Sessions 2026, the industry got a glimpse into what may be the next era of
financial infrastructure.
Payments are becoming autonomous.
Stripe unveiled a sweeping set of updates that point toward a future where AI agents
are not just assisting transactions but executing them:
- AI agent–driven purchases
- AI agents issuing and managing payments
- Agent wallets (via Link) enabling delegated spend
- Deep investment in machine-native payment infrastructure
- Over 280+ product launches focused on AI-native commerce
Stripe is effectively positioning itself as the economic layer for AI, partnering with
companies like Google to embed payments directly inside AI environments such as
Gemini.
This is a fundamental shift: payments are no longer user-initiated they are becoming
intent-executed by machines.
Are agents taking over fintech?
Short answer: they’re rapidly becoming core infrastructure.
Across the U.S., there is a surge in agentic AI startups and enterprise deployments,
with fintech at the centre.
Recent developments show how quickly this is materialising:
- Major banks like Citigroup have launched internal agent platforms, Arc to allow employees to deploy AI agents across workflows from portfolio analysis
to risk simulations. - AI-native financial tools are emerging, such as Citi’s upcoming AI wealth
advisor Citi Sky, designed to scale personalized financial advice. - Companies like Anthropic are releasing specialized financial AI agents
capable of drafting credit memos, closing books, and even assisting
compliance teams.
Even more critically, these agents are not just internal tools they are increasingly
external-facing economic actors.
Payments giants are embracing AI agents
The shift is not limited to startups.
- Visa is actively preparing for a world where AI agents transact using cards,
predicting new forms of B2B and microtransaction flows driven by
autonomous systems. - Stripe’s agent wallets and tokenized payment flows reinforce this trend,
ensuring secure delegation of financial authority to AI systems.
This signals a broader evolution:
Payments infrastructure is being redesigned for machines as first-class users.
Startups powering the agentic fintech stack
Alongside incumbents, a new wave of U.S. fintech startups is building the
infrastructure layer for this shift:
- Increase API-first banking infrastructure enabling programmable money
movement and machine-readable transaction states. - SoFi expanding into hybrid fiat + crypto banking and launching business
banking tools that unify programmable finance. - Emerging agent-focused startups (e.g., AI-native commerce, autonomous
procurement tools) are integrating directly with Stripe and similar platforms to
enable “agent checkout” experiences
These companies are laying the groundwork for machine-to-machine finance, where
APIs replace interfaces and logic replaces manual workflows.
The rise of agentic commerce
All of this feeds into a broader concept gaining traction: agentic commerce.
This model enables AI systems to:
- Discover products
- Make decisions
- Execute transactions
- Optimize outcomes over time all without human intervention.
Technically, this requires real-time payments, tokenization, APIs, and continuous risk
monitoring areas where U.S. fintech is heavily investing.
Stripe’s moves, combined with Visa’s strategy and enterprise adoption, suggest that
agentic commerce is no longer theoretical it’s entering production.
Key takeaway: fintech is becoming machine-native
What’s happening in the U.S. right now is not just incremental innovation it’s a
structural shift:
- From user-driven → agent-driven finance
- From interfaces → APIs and autonomous execution
- From SaaS tools → embedded financial intelligence
AI agents are evolving from copilots into economic participants.
AI level 0 of 5: researched, written and edited by Amrit Kang without generative AI; only everyday tools such as spelling and grammar checkers were used. What the levels mean
