Artificial intelligence has become remarkably good at generating answers. What it still struggles to do is act.
Today’s AI assistants can recommend products, compare services, search databases and write code, but when a transaction is required, they usually stop at the point where a human reaches for a credit card or approves a payment.
MoonPay believes that is about to change.
In a one-to-one interview, Kevin Arifin, Head of Agentic AI at MoonPay, joined Justin Roberti to discuss why payments are becoming one of the next major battlegrounds for AI. Rather than viewing AI simply as a conversational interface, MoonPay sees autonomous agents becoming economic participants capable of holding value, authenticating themselves and moving money across digital networks.
The conversation explored whether blockchain provides capabilities that traditional payment infrastructure cannot, why wallets may become as important to AI agents as large language models themselves, and how MoonPay is embedding crypto payments directly into leading AI platforms including Claude, ChatGPT, Grok and Perplexity.
From On-Ramp to Operating System
For years, MoonPay has been known primarily as a fiat-to-crypto gateway. Arifin argues that description is becoming increasingly outdated. Instead, MoonPay now sees itself as building what he describes as an “operating system for value.” The shift reflects a much broader change in computing itself.
Historically, operating systems were built for people sitting behind keyboards. The next generation, Arifin believes, will increasingly be designed for software agents.
Rather than humans manually moving funds between bank accounts, wallets and exchanges, autonomous agents will eventually perform many of those tasks themselves—provided they have secure ways to authenticate, transact and manage assets.
Why Blockchain Fits AI
Much of today’s discussion around AI focuses on model capability. MoonPay is more interested in what happens after the model produces an answer.
Arifin argues that blockchain naturally complements AI because it provides three capabilities agents require.
- First, it acts as a globally accessible public database that AI can read from.
- Second, it gives agents the ability to write back through transactions that permanently alter state.
- Third—and perhaps most importantly—it enables authentication through cryptographic wallets.
Instead of relying on usernames, passwords or platform-specific APIs, AI agents can prove ownership and authorise actions using public and private keys.
As Arifin explained, blockchain allows agents not only to read information but also to write to shared infrastructure and securely authenticate themselves in order to take action.
Natural Language Becomes the Interface
One of the interview’s most compelling demonstrations showed how MoonPay is attempting to remove much of crypto’s traditional complexity.
Rather than manually opening multiple wallets, bridging assets between chains or navigating decentralised exchanges, users simply describe what they want. The AI agent handles the routing.
During the live in camera demonstration, the assistant researched a newly launched blockchain ecosystem, identified a trending asset, located available liquidity, executed a cross-chain swap from an existing wallet and confirmed the updated portfolio—all through conversational prompts.
Behind the scenes, MoonPay manages wallet discovery, chain abstraction and transaction routing while leaving users’ private keys securely stored on their own devices rather than under MoonPay’s control.
For Arifin, this represents a fundamental shift. Instead of asking users to understand blockchain infrastructure, AI simply makes blockchain behave like another native capability.
AI as Crypto’s Missing Interface
Blockchain has often struggled with usability. Wallet management, multiple networks, bridge protocols and fragmented applications remain major barriers for mainstream adoption.
Arifin believes AI finally solves that problem.
“I think AI is the interface layer that blockchain has always needed,” he explained.
Rather than forcing users to learn dozens of applications and workflows, large language models can translate natural language directly into financial actions. Instead of memorising protocols, users simply explain their intent.
The agent determines the required tools.
Winning Over Mainstream Users
The discussion also addressed perhaps the biggest obstacle facing agentic finance: trust.
Most consumers remain cautious about giving software permission to move money autonomously. Arifin acknowledged that sophisticated technology alone will not drive adoption. Success depends on combining familiar financial experiences with intuitive AI interactions while embedding robust safeguards beneath the surface.
MoonPay’s approach includes KYC processes, transaction guardrails, funding controls and increasingly conservative behaviour for inexperienced users, while allowing more advanced users to unlock greater automation as confidence grows.
The objective is not complete autonomy, but graduated trust.
Your Personal Financial Agent
Looking further ahead, Arifin believes AI agents will democratise financial expertise itself.
Historically, investors relied on brokers who researched markets, monitored opportunities and helped execute trades. Future AI assistants may perform many of those functions continuously.
Rather than replacing financial judgement, they augment it—tracking both traditional and tokenised assets, understanding a user’s risk profile and helping manage increasingly complex portfolios through natural conversation.
As Arifin described it, everyone could eventually have access to a personal financial broker that understands both on-chain and traditional markets, making sophisticated financial tools available far beyond today’s professional investors.
The Bigger Picture
MoonPay’s vision extends beyond making crypto easier to buy. It positions payments as the missing infrastructure layer for autonomous AI.
Large language models have become exceptionally capable at reasoning and planning, but economic participation requires more than intelligence. Agents must be able to own assets, authenticate themselves securely and settle transactions across multiple financial systems.
Whether blockchain ultimately becomes the dominant payment rail remains an open question.
What seems increasingly clear, however, is that if AI agents are to move beyond answering questions and begin acting independently, they will need a financial operating system capable of moving value as seamlessly as today’s models move information.
